From eb3cfd4ff78d350b6ce9c8d0668c3b23df6ccb20 Mon Sep 17 00:00:00 2001 From: NetterSilver <133535268+Silvertree2010@users.noreply.github.com> Date: Tue, 3 Nov 2026 16:08:30 +0100 Subject: [PATCH 01/75] Ignore dictionary key order when checking signed pages for changes Fixes #5955 Co-Authored-By: Claude Opus 5.5 --- crates/sign/src/pdf/changes.rs | 113 ++++++++++++++++++++++++++++++++- 1 file changed, 111 insertions(+), 2 deletions(-) diff --git a/crates/sign/src/pdf/changes.rs b/crates/sign/src/pdf/changes.rs index 3a10a74e41..28c96633cc 100644 --- a/crates/sign/src/pdf/changes.rs +++ b/crates/sign/src/pdf/changes.rs @@ -427,6 +427,35 @@ fn without(x: &Dict, keys: &[&[u8]]) -> Vec<(Vec, Object)> { v } +/// Whether two page dictionaries are the same apart from `/Annots`. Dictionaries are compared +/// without regard to key order at every depth: a writer may re-serialise a page's inline +/// resources in another order without changing what the page shows. +fn same_but_annots(d: &Dict, b: &Dict) -> bool { + same_entries(d, b, &[b"Annots"]) +} + +/// Whether two dictionaries hold the same entries apart from `skip`: both sorted by key, then +/// compared pairwise, so neither side can hide a key the other lacks. +fn same_entries(a: &Dict, b: &Dict, skip: &[&[u8]]) -> bool { + fn sorted<'a>(x: &'a Dict, skip: &[&[u8]]) -> Vec<(&'a [u8], &'a Object)> { + let mut v: Vec<(&[u8], &Object)> = x.iter().map(|(k, o)| (k.as_slice(), o)).filter(|(k, _)| !skip.contains(k)).collect(); + v.sort_by(|p, q| p.0.cmp(q.0)); + v + } + let (x, y) = (sorted(a, skip), sorted(b, skip)); + x.len() == y.len() && x.iter().zip(&y).all(|((k, v), (l, w))| k == l && same_object(v, w)) +} + +/// Object equality where dictionaries (and stream dictionaries) ignore key order; arrays keep it. +fn same_object(a: &Object, b: &Object) -> bool { + match (a, b) { + (Object::Dict(a), Object::Dict(b)) => same_entries(a, b, &[]), + (Object::Stream(a), Object::Stream(b)) => a.raw == b.raw && same_entries(&a.dict, &b.dict, &[]), + (Object::Array(a), Object::Array(b)) => a.len() == b.len() && a.iter().zip(b).all(|(v, w)| same_object(v, w)), + _ => a == b, + } +} + /// Whether a dictionary was a signature object: a signature or timestamp value, a signature /// reference or its transform parameters, or a signature form field. Used to decide from the /// signed revision, so a later relabel of some other object cannot pose as a signature change. @@ -521,7 +550,7 @@ fn signed_page_resources(old: &Document, pages: &HashSet) -> HashMap) -> &'static str { match before { - Some(b) if without(d, &[b"Annots"]) != without(b, &[b"Annots"]) => kind::PAGE_CONTENT, + Some(b) if !same_but_annots(d, b) => kind::PAGE_CONTENT, Some(b) => { let was = refs(old, b.get(b"Annots")); let added: Vec = refs(doc, d.get(b"Annots")).into_iter().filter(|r| !was.contains(r)).collect(); @@ -584,7 +613,7 @@ fn change_kind(doc: &Document, old: &Document, d: &Dict, before: Option<&Dict>) } Some(b"Page") => { return match before { - Some(b) if without(d, &[b"Annots"]) != without(b, &[b"Annots"]) => kind::PAGE_CONTENT, + Some(b) if !same_but_annots(d, b) => kind::PAGE_CONTENT, Some(b) => { let was = refs(old, b.get(b"Annots")); let added: Vec = refs(doc, d.get(b"Annots")).into_iter().filter(|r| !was.contains(r)).collect(); @@ -617,3 +646,83 @@ fn change_kind(doc: &Document, old: &Document, d: &Dict, before: Option<&Dict>) } kind::OTHER } + +#[cfg(test)] +mod tests { + use super::*; + + fn font(order: &[&str], base: &str) -> Object { + let mut d = Dict::new(); + for k in order { + match *k { + "Type" => d.set(b"Type".to_vec(), Object::name("Font")), + "Subtype" => d.set(b"Subtype".to_vec(), Object::name("Type1")), + "BaseFont" => d.set(b"BaseFont".to_vec(), Object::name(base)), + _ => { + let mut w = Dict::new(); + w.set(b"Differences".to_vec(), Object::Array(vec![Object::Int(32), Object::name("space")])); + w.set(b"Type".to_vec(), Object::name("Encoding")); + d.set(b"Encoding".to_vec(), Object::Dict(w)); + } + } + } + Object::Dict(d) + } + + fn page(fonts: Vec<(&str, Object)>, annots: Vec) -> Dict { + let mut f = Dict::new(); + fonts.into_iter().for_each(|(k, v)| f.set(k.as_bytes().to_vec(), v)); + let mut r = Dict::new(); + r.set(b"Font".to_vec(), Object::Dict(f)); + let mut p = Dict::new(); + p.set(b"Type".to_vec(), Object::name("Page")); + p.set(b"Resources".to_vec(), Object::Dict(r)); + p.set(b"Annots".to_vec(), Object::Array(annots)); + p + } + + #[test] + fn reordered_inline_resources_are_not_page_content() { + let doc = Document::new_empty(); + let order = ["Type", "Subtype", "BaseFont", "Encoding"]; + let before = page(vec![("F1", font(&order, "Times-Roman")), ("F2", font(&order, "Helvetica"))], vec![]); + let unchanged = page_change_kind(&doc, &doc, &before, Some(&before)); + let reordered = page( + vec![ + ("F2", font(&["Encoding", "BaseFont", "Type", "Subtype"], "Helvetica")), + ("F1", font(&["BaseFont", "Subtype", "Encoding", "Type"], "Times-Roman")), + ], + vec![], + ); + assert_eq!(page_change_kind(&doc, &doc, &reordered, Some(&before)), unchanged); + assert_eq!(change_kind(&doc, &doc, &reordered, Some(&before)), change_kind(&doc, &doc, &before, Some(&before))); + + // A changed value inside the nested dictionary still is a page content change. + let changed = page(vec![("F2", font(&order, "Helvetica")), ("F1", font(&order, "Courier"))], vec![]); + assert_eq!(page_change_kind(&doc, &doc, &changed, Some(&before)), kind::PAGE_CONTENT); + assert_eq!(change_kind(&doc, &doc, &changed, Some(&before)), kind::PAGE_CONTENT); + } + + #[test] + fn a_swapped_nested_key_is_page_content_both_ways() { + let doc = Document::new_empty(); + let order = ["Type", "Subtype", "BaseFont"]; + let signed = page(vec![("F1", font(&order, "Times-Roman"))], vec![]); + // Same entry count, but /Subtype was dropped and /Encoding added in its place. + let swapped = page(vec![("F1", font(&["Type", "Encoding", "BaseFont"], "Times-Roman"))], vec![]); + for (now, then) in [(&swapped, &signed), (&signed, &swapped)] { + assert_eq!(page_change_kind(&doc, &doc, now, Some(then)), kind::PAGE_CONTENT); + assert_eq!(change_kind(&doc, &doc, now, Some(then)), kind::PAGE_CONTENT); + } + } + + #[test] + fn reordered_arrays_are_still_page_content() { + let doc = Document::new_empty(); + let mut before = page(vec![], vec![]); + before.set(b"MediaBox".to_vec(), Object::Array(vec![Object::Int(0), Object::Int(0), Object::Int(612), Object::Int(792)])); + let mut after = before.clone(); + after.set(b"MediaBox".to_vec(), Object::Array(vec![Object::Int(0), Object::Int(0), Object::Int(792), Object::Int(612)])); + assert_eq!(page_change_kind(&doc, &doc, &after, Some(&before)), kind::PAGE_CONTENT); + } +} From 715c310f5b770e2621c6c2e0f969455dd481209f Mon Sep 17 00:00:00 2001 From: mtlcl410 Date: Sat, 7 Nov 2026 10:25:41 +0800 Subject: [PATCH 02/75] M9: a signed string respelled in hexadecimal is no change after signing (sign.changes-after-signing) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit A string object is its bytes; literal and hexadecimal are two ways to write them (ISO 32000-2 §7.3.4.1-7.3.4.3). The modification check compared objects with the derived equality, which also compares how a string was written, so a later save that rewrote the catalog's /Lang from (en-US) to <656E2D5553> made an untouched signature read as a document structure change. same_object now compares strings by their bytes, and every comparison in changes.rs goes through it: the unchanged-object skip, the catalog, its /AcroForm and /Perms, developer extensions, signed widgets, page resource containers, the security store and the timestamps check. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn --- crates/sign/src/pdf/changes.rs | 57 +++++++------- .../src/pdf/revision_classification_tests.rs | 77 +++++++++++++++++++ parity/acrobat-features.toml | 4 +- 3 files changed, 109 insertions(+), 29 deletions(-) diff --git a/crates/sign/src/pdf/changes.rs b/crates/sign/src/pdf/changes.rs index f6c5b286c4..b0d082a366 100644 --- a/crates/sign/src/pdf/changes.rs +++ b/crates/sign/src/pdf/changes.rs @@ -168,7 +168,7 @@ pub(super) fn classify_changes(doc: &Document, old: Option, p: Option< let r = ObjRef { num, generation: doc.generation(num) }; let new = doc.get(r); let before = old.try_get(num).ok().filter(|o| !matches!(**o, Object::Null)); - if before.as_deref() == Some(&*new) { + if before.as_deref().is_some_and(|b| same_object(b, &new)) { // The same content at a new place. That is true for the renderer as well only when // the entry locates the object itself: where this reader had to search for the // object, another reader may arrive at a different one. @@ -238,7 +238,7 @@ fn only_validation_data(allowed: &mut Vec<&'static str>, disallowed: &mut Vec<&' // Compared by content: an indirect /Perms rewritten in place keeps its reference (and, shared // with the trailer's /Info, would otherwise pass as a metadata change). let perms = |d: &Document| d.root().and_then(|r| d.get(r).as_dict().and_then(|c| c.get(b"Perms").map(|p| (*d.resolve(p)).clone()))); - if perms(doc) != perms(old) && !disallowed.contains(&kind::STRUCTURE) { + if !same_entry(perms(doc).as_ref(), perms(old).as_ref()) && !disallowed.contains(&kind::STRUCTURE) { disallowed.push(kind::STRUCTURE); } } @@ -269,7 +269,7 @@ fn only_invisible_timestamps_added(doc: &Document, old: &Document) -> bool { continue; } // Unchanged since signing: the certification already covered it. - if old.try_get(num).ok().filter(|o| !matches!(**o, Object::Null)).as_deref() == Some(&*new) { + if old.try_get(num).ok().filter(|o| !matches!(**o, Object::Null)).is_some_and(|o| same_object(&o, &new)) { continue; } let value = if is_value { Some(d.clone()) } else { d.get(b"V").map(|v| doc.resolve(v)).and_then(|v| v.as_dict().cloned()) }; @@ -330,7 +330,7 @@ impl ChangedObjectClassifier<'_> { } fn keeps_previous_entries(&self, dict: &Dict, before: Option<&Object>) -> bool { - before.and_then(Object::as_dict).is_some_and(|previous| previous.iter().all(|(key, value)| dict.get(key) == Some(value))) + before.and_then(Object::as_dict).is_some_and(|previous| previous.iter().all(|(key, value)| same_entry(dict.get(key), Some(value)))) } fn is_signed_annotation_array(&self, r: ObjRef, before: Option<&Object>) -> bool { @@ -520,14 +520,14 @@ pub(crate) fn only_grew(doc: &Document, old: &Document, before: &Object, now: &O let refs = |a: &[Object]| a.iter().map(Object::as_ref).collect::>>(); match (refs(b), refs(n)) { (Some(b), Some(n)) => b.is_subset(&n), - _ => b.len() <= MAX_COMPARED && n.len() <= MAX_COMPARED && b.iter().all(|x| n.contains(x)), + _ => b.len() <= MAX_COMPARED && n.len() <= MAX_COMPARED && b.iter().all(|x| n.iter().any(|y| same_object(x, y))), } } (Object::Dict(b), Object::Dict(n)) => b.iter().all(|(k, v)| match n.get(k) { - Some(nv) => nv == v || (depth < 4 && only_grew(doc, old, &old.resolve(v), &doc.resolve(nv), depth + 1)), + Some(nv) => same_object(nv, v) || (depth < 4 && only_grew(doc, old, &old.resolve(v), &doc.resolve(nv), depth + 1)), None => false, }), - _ => before == now, + _ => same_object(before, now), } } @@ -651,13 +651,7 @@ fn acroform_kind(doc: &Document, old: &Document, now: &Dict, before: Option<&Dic if lost_entries(doc, old, before.get(b"Fields"), now.get(b"Fields")) { return kind::OTHER; } - let strip = |d: &Dict| -> Vec<(Vec, Object)> { - let mut v: Vec<(Vec, Object)> = - d.iter().filter(|(k, _)| !matches!(k.as_slice(), b"Fields" | b"SigFlags")).map(|(k, o)| (k.to_vec(), o.clone())).collect(); - v.sort_by(|a, b| a.0.cmp(&b.0)); - v - }; - if strip(now) != strip(before) { + if !same_entries(now, before, &[b"Fields", b"SigFlags"]) { return kind::FORM_FILL; } let old_fields = refs(old, before.get(b"Fields")); @@ -665,13 +659,6 @@ fn acroform_kind(doc: &Document, old: &Document, now: &Dict, before: Option<&Dic if added.is_empty() || added_annots(doc, &added) == kind::SIGNATURE { kind::SIGNATURE } else { kind::FORM_FILL } } -/// Dictionary entries in a stable order, without `keys`. -fn without(x: &Dict, keys: &[&[u8]]) -> Vec<(Vec, Object)> { - let mut v: Vec<(Vec, Object)> = x.iter().filter(|(k, _)| !keys.contains(&k.as_slice())).map(|(k, o)| (k.to_vec(), o.clone())).collect(); - v.sort_by(|a, b| a.0.cmp(&b.0)); - v -} - /// Whether two page dictionaries are the same apart from `/Annots`. Dictionaries are compared /// without regard to key order at every depth: a writer may re-serialise a page's inline /// resources in another order without changing what the page shows. @@ -692,8 +679,13 @@ fn same_entries(a: &Dict, b: &Dict, skip: &[&[u8]]) -> bool { } /// Object equality where dictionaries (and stream dictionaries) ignore key order; arrays keep it. +/// A string is its bytes (ISO 32000-2 §7.3.4.1): written as a literal or as hexadecimal, it is the +/// same string, so a later save that respells one changes nothing it says. References compare by +/// object and generation, and stream data byte for byte as stored. The recursion follows the +/// parsed nesting, which the parser bounds. fn same_object(a: &Object, b: &Object) -> bool { match (a, b) { + (Object::String(a), Object::String(b)) => a.bytes == b.bytes, (Object::Dict(a), Object::Dict(b)) => same_entries(a, b, &[]), (Object::Stream(a), Object::Stream(b)) => a.raw == b.raw && same_entries(&a.dict, &b.dict, &[]), (Object::Array(a), Object::Array(b)) => a.len() == b.len() && a.iter().zip(b).all(|(v, w)| same_object(v, w)), @@ -701,6 +693,15 @@ fn same_object(a: &Object, b: &Object) -> bool { } } +/// [`same_object`] for an entry that may be absent: both absent, or both present and the same. +fn same_entry(a: Option<&Object>, b: Option<&Object>) -> bool { + match (a, b) { + (Some(a), Some(b)) => same_object(a, b), + (None, None) => true, + _ => false, + } +} + /// Whether a dictionary was a signature object: a signature or timestamp value, a signature /// reference or its transform parameters, or a signature form field. Used to decide from the /// signed revision, so a later relabel of some other object cannot pose as a signature change. @@ -798,7 +799,7 @@ fn extensions_kept(doc: &Document, old: &Document, d: &Dict, b: &Dict) -> bool { let was = b.get(b"Extensions").map(|e| old.resolve(e)).and_then(|e| e.as_dict().cloned()); match (was, now) { (None, _) => true, - (Some(was), Some(now)) => was.iter().all(|(k, v)| now.get(k) == Some(v)), + (Some(was), Some(now)) => was.iter().all(|(k, v)| same_entry(now.get(k), Some(v))), (Some(_), None) => false, } } @@ -806,18 +807,18 @@ fn extensions_kept(doc: &Document, old: &Document, d: &Dict, b: &Dict) -> bool { /// A change to the signed catalog `b`, now `d`. fn catalog_kind(doc: &Document, old: &Document, d: &Dict, b: &Dict) -> &'static str { let keys: &[&[u8]] = &[b"AcroForm", b"DSS", b"Perms", b"Metadata", b"NeedsRendering", b"Extensions"]; - if without(d, keys) != without(b, keys) { + if !same_entries(d, b, keys) { kind::STRUCTURE } else if !extensions_kept(doc, old, d, b) { // Developer extensions (ISO 32000-2 §7.12) only declare what the file uses. A writer // adding a /DSS also declares the ESIC extension it belongs to, so a declaration may be // added, but none that was signed may change or go. kind::STRUCTURE - } else if d.get(b"AcroForm") != b.get(b"AcroForm") { + } else if !same_entry(d.get(b"AcroForm"), b.get(b"AcroForm")) { // Compare the forms themselves, inline or not. let now = acroform(doc).unwrap_or_default(); acroform_kind(doc, old, &now, None) - } else if b.get(b"Perms") != d.get(b"Perms") { + } else if !same_entry(b.get(b"Perms"), d.get(b"Perms")) { kind::SIGNATURE } else { kind::METADATA @@ -869,7 +870,7 @@ fn change_kind(doc: &Document, old: &Document, d: &Dict, before: Option<&Dict>) // so relabelling the widget now does not change that. if before.is_some_and(|b| { b.name(b"Subtype") == Some(b"Widget") && b.name(b"FT") == Some(b"Sig") && b.get(b"V").is_some_and(|v| old.resolve(v).as_dict().is_some()) - }) && before != Some(d) + }) && !before.is_some_and(|b| same_entries(d, b, &[])) { return kind::PAGE_CONTENT; } @@ -883,7 +884,9 @@ fn change_kind(doc: &Document, old: &Document, d: &Dict, before: Option<&Dict>) } // Where a signed widget sits and whether it shows are what the page shows: filling the form // changes its value and appearance, never its rectangle or its flags. - if before.is_some_and(|b| b.name(b"Subtype") == Some(b"Widget") && (d.get(b"Rect") != b.get(b"Rect") || d.get(b"F") != b.get(b"F"))) { + if before.is_some_and(|b| { + b.name(b"Subtype") == Some(b"Widget") && (!same_entry(d.get(b"Rect"), b.get(b"Rect")) || !same_entry(d.get(b"F"), b.get(b"F"))) + }) { return kind::PAGE_CONTENT; } match ty { diff --git a/crates/sign/src/pdf/revision_classification_tests.rs b/crates/sign/src/pdf/revision_classification_tests.rs index b5e72fcae0..3b63953a0d 100644 --- a/crates/sign/src/pdf/revision_classification_tests.rs +++ b/crates/sign/src/pdf/revision_classification_tests.rs @@ -282,3 +282,80 @@ fn a_repaired_slot_rewritten_cleanly_is_judged_by_its_references() { let now = signed.clone().revision(&[(4, Row::Obj(0, 0, &filled))]); assert!(matches!(verdict(&now.open(), signed.open(), Some(2)), Modification::Disallowed(_))); } + +/// A signed catalog and text field whose strings a later save respells, given the catalog `extra` +/// entries and the field's `/T` as written in each revision. +fn respelled(signed_extra: &str, now_extra: &str, signed_t: &str, now_t: &str) -> (Document, Document) { + let catalog = |extra: &str| format!("<< /Type /Catalog /Pages 2 0 R /AcroForm << /Fields [5 0 R] /DA (/Helv 0 Tf 0 g) >> {extra} >>"); + let field = |t: &str| format!("<< /Type /Annot /Subtype /Widget /FT /Tx /T {t} /V (Ann) /Rect [20 20 120 40] /P 3 0 R >>"); + let signed = File::new().revision(&[ + (1, Row::Obj(0, 0, &catalog(signed_extra))), + (2, Row::Obj(0, 0, PAGES)), + (3, Row::Obj(0, 0, &page("/Annots [5 0 R]"))), + (4, Row::Obj(0, 0, &page("/Annots []"))), + (5, Row::Obj(0, 0, &field(signed_t))), + ]); + let now = signed.clone().revision(&[(1, Row::Obj(0, 0, &catalog(now_extra))), (5, Row::Obj(0, 0, &field(now_t)))]); + (now.open(), signed.open()) +} + +/// A string is its bytes (ISO 32000-2 §7.3.4.1): a later save that writes the catalog's /Lang in +/// hexadecimal instead of as a literal changes nothing, under every permission level. +#[test] +fn a_string_respelled_in_hex_is_no_change() { + for p in [Some(1), Some(2), Some(3), None] { + let (doc, old) = respelled("/Lang (en-US)", "/Lang <656E2D5553>", "(Name)", "(Name)"); + assert_eq!(verdict(&doc, old, p), Modification::Allowed(Vec::new()), "P={p:?}"); + } +} + +/// The same holds for strings nested in dictionaries and arrays (a name tree's keys), with the +/// entries also in another order, and for a signed field's /T. +#[test] +fn nested_strings_respelled_and_reordered_are_no_change() { + let (doc, old) = respelled( + "/Names << /Dests << /Names [(intro) [3 0 R /Fit]] >> >> /Lang (en-US)", + "/Lang <656e2d5553> /Names << /Dests << /Names [<696E74726F> [3 0 R /Fit]] >> >>", + "(Name)", + "<4E616D65>", + ); + assert_eq!(verdict(&doc, old, Some(1)), Modification::Allowed(Vec::new())); +} + +/// Different bytes are a different string however they are written: a catalog whose /Lang (or a +/// string deep in its name tree) now says something else is still a structure change. +#[test] +fn a_string_whose_bytes_change_is_still_a_change() { + for (signed, now) in [ + ("/Lang (en-US)", "/Lang <64652D4445>"), + ("/Names << /Dests << /Names [(intro) [3 0 R /Fit]] >> >>", "/Names << /Dests << /Names [<6F75747472> [3 0 R /Fit]] >> >>"), + ] { + let (doc, old) = respelled(signed, now, "(Name)", "(Name)"); + assert_eq!(verdict(&doc, old, Some(2)), Modification::Disallowed(vec![super::kind::STRUCTURE]), "{now}"); + } +} + +/// A save that adds a signature field and respells the catalog's and the form's strings on the +/// way is judged by the field it added: the respelled /Lang and /DA are not a structure or form +/// change. +#[test] +fn a_signature_added_with_respelled_strings_is_signing() { + let signed = File::new().revision(&[ + (1, Row::Obj(0, 0, "<< /Type /Catalog /Pages 2 0 R /Lang (en-US) /AcroForm << /Fields [5 0 R] /DA (/Helv 0 Tf 0 g) >> >>")), + (2, Row::Obj(0, 0, PAGES)), + (3, Row::Obj(0, 0, &page("/Annots [5 0 R]"))), + (4, Row::Obj(0, 0, &page("/Annots []"))), + (5, Row::Obj(0, 0, TEXT_FIELD)), + ]); + let catalog = + |lang: &str, da: &str| format!("<< /Type /Catalog /Pages 2 0 R /Lang {lang} /AcroForm << /DA {da} /Fields [5 0 R 6 0 R] /SigFlags 3 >> >>"); + let sig = "<< /Type /Annot /Subtype /Widget /FT /Sig /T (Sig1) /Rect [0 0 0 0] /P 4 0 R >>"; + let now = |lang: &str, da: &str| { + signed.clone().revision(&[(1, Row::Obj(0, 0, &catalog(lang, da))), (4, Row::Obj(0, 0, &page("/Annots [6 0 R]"))), (6, Row::Obj(0, 0, sig))]) + }; + let respelled = now("<656E2D5553>", "<2F48656C76203020546620302067>"); + assert_eq!(verdict(&respelled.open(), signed.open(), Some(2)), Modification::Allowed(vec![super::kind::SIGNATURE])); + // A /DA that now says something else is a form change, however it is spelled. + let restyled = now("<656E2D5553>", "<2F48656C7620313220546620302067>"); + assert_eq!(verdict(&restyled.open(), signed.open(), None), Modification::Allowed(vec![super::kind::FORM_FILL, super::kind::SIGNATURE])); +} diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index 10b040e09f..0bb7b7cf5d 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -6535,8 +6535,8 @@ milestone = "M9" status = "shipped" commands = ["sign.panel"] tools = ["sign_list"] -evidence = ["crates/sign/tests/it/pdf.rs::later_changes_are_classified_under_the_signature_permissions", "crates/sign/tests/it/pdf.rs::signing_an_existing_field_and_counter_signing", "crates/engine/src/tests.rs::signing_saving_trusting_and_commenting_afterwards", "crates/sign/tests/it/pdf.rs::validation_data_and_xmp_added_after_signing_are_not_tampering", "crates/sign/tests/it/pdf.rs::a_dss_entry_does_not_make_the_object_it_names_part_of_the_store", "crates/sign/tests/it/pdf.rs::an_existing_dictionary_is_not_a_store_part_because_the_store_names_it", "crates/sign/tests/it/pdf.rs::an_existing_array_stays_what_it_is_when_the_store_points_at_it", "crates/sign/tests/it/pdf.rs::a_signed_store_may_grow_but_not_lose_or_swap_entries", "crates/sign/tests/it/pdf.rs::a_dictionary_labelled_metadata_is_not_a_metadata_change", "crates/sign/tests/it/pdf.rs::a_stream_already_labelled_metadata_when_signed_is_not_metadata_unless_the_catalog_names_it", "crates/sign/tests/it/pdf.rs::the_catalogs_own_xmp_stream_may_still_be_rewritten_after_signing", "crates/sign/tests/it/pdf.rs::rewritten_page_resources_are_page_content_whatever_keys_they_carry", "crates/sign/tests/it/pdf.rs::rewriting_a_certified_forms_xfa_data_is_a_form_fill", "crates/sign/tests/it/pdf.rs::the_catalogs_perms_do_not_change_under_no_changes"] -notes = "revision diff classified as signing, form fill, comments, metadata, page content or structure. Later revisions that add DSS data (new /Certs, /CRLs, /OCSPs, /VRI objects, or ones of the signed store that only grew) or rewrite the catalog's XMP are permitted changes; a /DSS entry that names other objects, or a /Type /DSS or /Type /Metadata label on an existing dictionary or stream, does not excuse changing them (negative tests at every DocMDP level). A signed page's resources (fonts, XObjects, graphics states and what they reach) are page content by the role they held when signed, whatever keys a later revision writes into them (#1787). Rewriting the XFA `datasets` or `form` packet the signed revision named is form fill, as a Reader Extensions (UR3) save of a certified form does (#2759); the packets' roles come from the signed revision. Under P=1 a change to the catalog's /Perms breaks the certification." +evidence = ["crates/sign/tests/it/pdf.rs::later_changes_are_classified_under_the_signature_permissions", "crates/sign/tests/it/pdf.rs::signing_an_existing_field_and_counter_signing", "crates/engine/src/tests.rs::signing_saving_trusting_and_commenting_afterwards", "crates/sign/tests/it/pdf.rs::validation_data_and_xmp_added_after_signing_are_not_tampering", "crates/sign/tests/it/pdf.rs::a_dss_entry_does_not_make_the_object_it_names_part_of_the_store", "crates/sign/tests/it/pdf.rs::an_existing_dictionary_is_not_a_store_part_because_the_store_names_it", "crates/sign/tests/it/pdf.rs::an_existing_array_stays_what_it_is_when_the_store_points_at_it", "crates/sign/tests/it/pdf.rs::a_signed_store_may_grow_but_not_lose_or_swap_entries", "crates/sign/tests/it/pdf.rs::a_dictionary_labelled_metadata_is_not_a_metadata_change", "crates/sign/tests/it/pdf.rs::a_stream_already_labelled_metadata_when_signed_is_not_metadata_unless_the_catalog_names_it", "crates/sign/tests/it/pdf.rs::the_catalogs_own_xmp_stream_may_still_be_rewritten_after_signing", "crates/sign/tests/it/pdf.rs::rewritten_page_resources_are_page_content_whatever_keys_they_carry", "crates/sign/tests/it/pdf.rs::rewriting_a_certified_forms_xfa_data_is_a_form_fill", "crates/sign/tests/it/pdf.rs::the_catalogs_perms_do_not_change_under_no_changes", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_respelled_in_hex_is_no_change", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_whose_bytes_change_is_still_a_change"] +notes = "revision diff classified as signing, form fill, comments, metadata, page content or structure. Later revisions that add DSS data (new /Certs, /CRLs, /OCSPs, /VRI objects, or ones of the signed store that only grew) or rewrite the catalog's XMP are permitted changes; a /DSS entry that names other objects, or a /Type /DSS or /Type /Metadata label on an existing dictionary or stream, does not excuse changing them (negative tests at every DocMDP level). A signed page's resources (fonts, XObjects, graphics states and what they reach) are page content by the role they held when signed, whatever keys a later revision writes into them (#1787). Rewriting the XFA `datasets` or `form` packet the signed revision named is form fill, as a Reader Extensions (UR3) save of a certified form does (#2759); the packets' roles come from the signed revision. Under P=1 a change to the catalog's /Perms breaks the certification. Signed objects are compared by value: a string is its bytes, so one a later save rewrites from literal to hexadecimal (or back) is unchanged (ISO 32000-2 §7.3.4.1)." [[feature]] id = "sign.view-signed-version" From 44428b36d318cf910ab25dd29ac9f5ed8961ff11 Mon Sep 17 00:00:00 2001 From: mtlcl410 Date: Sat, 7 Nov 2026 12:46:10 +0800 Subject: [PATCH 03/75] M9: a signed number saved back as an integer is no change after signing (sign.changes-after-signing) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ISO 32000-2 §7.3.3: wherever a real number is expected, an integer may be used instead. PdfCraft's own writer writes a whole real as an integer, so a form fill that rewrites a signed widget turned /Rect [20.0 20.0 120.0 40.0] into [20 20 120 40] and was refused as a page content change. same_object now compares an integer and a real by value (exactly, in i128), for the widget /Rect and every other signed comparison, such as a page /MediaBox rewritten when a comment is added. Also states the reordered page's verdict as None explicitly, in a test of its own next to the one from #5962. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn --- crates/sign/src/pdf/changes.rs | 61 ++++++++++++++++++- .../src/pdf/revision_classification_tests.rs | 38 ++++++++++++ parity/acrobat-features.toml | 4 +- 3 files changed, 98 insertions(+), 5 deletions(-) diff --git a/crates/sign/src/pdf/changes.rs b/crates/sign/src/pdf/changes.rs index b0d082a366..ce7349e2ad 100644 --- a/crates/sign/src/pdf/changes.rs +++ b/crates/sign/src/pdf/changes.rs @@ -680,12 +680,18 @@ fn same_entries(a: &Dict, b: &Dict, skip: &[&[u8]]) -> bool { /// Object equality where dictionaries (and stream dictionaries) ignore key order; arrays keep it. /// A string is its bytes (ISO 32000-2 §7.3.4.1): written as a literal or as hexadecimal, it is the -/// same string, so a later save that respells one changes nothing it says. References compare by -/// object and generation, and stream data byte for byte as stored. The recursion follows the -/// parsed nesting, which the parser bounds. +/// same string, so a later save that respells one changes nothing it says. Likewise a number is +/// its value (§7.3.3: "wherever a real number is expected, an integer may be used instead"), so a +/// signed `/Rect [20.0 20.0 120.0 40.0]` saved again as `[20 20 120 40]` is the same rectangle. +/// References compare by object and generation, and stream data byte for byte as stored. The +/// recursion follows the parsed nesting, which the parser bounds. fn same_object(a: &Object, b: &Object) -> bool { match (a, b) { (Object::String(a), Object::String(b)) => a.bytes == b.bytes, + // Compared in i128, where every i64 and every whole real near its range is exact (the cast + // saturates far beyond it): no fractional real, NaN or infinity (whose fract is NaN) + // equals an integer, nor does the real 2^63 pass for i64::MAX. + (Object::Int(i), Object::Real(r)) | (Object::Real(r), Object::Int(i)) => r.fract() == 0.0 && *r as i128 == i128::from(*i), (Object::Dict(a), Object::Dict(b)) => same_entries(a, b, &[]), (Object::Stream(a), Object::Stream(b)) => a.raw == b.raw && same_entries(&a.dict, &b.dict, &[]), (Object::Array(a), Object::Array(b)) => a.len() == b.len() && a.iter().zip(b).all(|(v, w)| same_object(v, w)), @@ -1023,6 +1029,55 @@ mod tests { assert_eq!(change_kind(&doc, &doc, &changed, Some(&before)), kind::PAGE_CONTENT); } + /// The page as signed and the same page with its inline resources reordered are each no + /// page change at all, stated as `None` rather than only equal to each other, and so is a page + /// whose numbers and strings are respelled as well (§7.3.3, §7.3.4.1). + #[test] + fn reordered_and_respelled_pages_are_no_page_change() { + let doc = Document::new_empty(); + let order = ["Type", "Subtype", "BaseFont", "Encoding"]; + let mut before = page(vec![("F1", font(&order, "Times-Roman")), ("F2", font(&order, "Helvetica"))], vec![]); + before.set(b"MediaBox".to_vec(), Object::Array(vec![Object::Int(0), Object::Int(0), Object::Real(612.0), Object::Real(792.0)])); + before.set( + b"PieceInfo".to_vec(), + Object::Dict(Dict::from_iter([(b"Note".to_vec(), Object::String(pdfcraft_cos::PdfString::literal("signed")))])), + ); + let mut reordered = page( + vec![ + ("F2", font(&["Encoding", "BaseFont", "Type", "Subtype"], "Helvetica")), + ("F1", font(&["BaseFont", "Subtype", "Encoding", "Type"], "Times-Roman")), + ], + vec![], + ); + reordered.set(b"MediaBox".to_vec(), Object::Array(vec![Object::Real(0.0), Object::Int(0), Object::Int(612), Object::Int(792)])); + reordered.set( + b"PieceInfo".to_vec(), + Object::Dict(Dict::from_iter([(b"Note".to_vec(), Object::String(pdfcraft_cos::PdfString { bytes: b"signed".to_vec(), hex: true }))])), + ); + assert_eq!(page_change_kind(&doc, &doc, &before, Some(&before)), None); + assert_eq!(page_change_kind(&doc, &doc, &reordered, Some(&before)), None); + assert_eq!(page_change_kind(&doc, &doc, &before, Some(&reordered)), None); + + // A real that is not the integer it replaced is a moved box. + let mut moved = reordered.clone(); + moved.set(b"MediaBox".to_vec(), Object::Array(vec![Object::Int(0), Object::Int(0), Object::Real(612.5), Object::Int(792)])); + assert_eq!(page_change_kind(&doc, &doc, &moved, Some(&before)), Some(kind::PAGE_CONTENT)); + } + + /// Numbers compare by value across integer and real, and only when the values are equal. + #[test] + fn integers_and_reals_compare_by_value() { + let same = |a: Object, b: Object| same_object(&a, &b) && same_object(&b, &a); + assert!(same(Object::Int(20), Object::Real(20.0))); + assert!(same(Object::Int(0), Object::Real(-0.0))); + assert!(same(Object::Int(-7), Object::Real(-7.0))); + assert!(!same(Object::Int(20), Object::Real(20.5))); + assert!(!same(Object::Int(20), Object::Real(f64::NAN))); + assert!(!same(Object::Int(i64::MAX), Object::Real(f64::INFINITY))); + assert!(!same(Object::Int(i64::MAX), Object::Real(9_223_372_036_854_775_808.0))); + assert!(!same(Object::Int(1), Object::Bool(true))); + } + #[test] fn a_swapped_nested_key_is_page_content_both_ways() { let doc = Document::new_empty(); diff --git a/crates/sign/src/pdf/revision_classification_tests.rs b/crates/sign/src/pdf/revision_classification_tests.rs index 3b63953a0d..ec95d1323b 100644 --- a/crates/sign/src/pdf/revision_classification_tests.rs +++ b/crates/sign/src/pdf/revision_classification_tests.rs @@ -359,3 +359,41 @@ fn a_signature_added_with_respelled_strings_is_signing() { let restyled = now("<656E2D5553>", "<2F48656C7620313220546620302067>"); assert_eq!(verdict(&restyled.open(), signed.open(), None), Modification::Allowed(vec![super::kind::FORM_FILL, super::kind::SIGNATURE])); } + +/// A signed widget whose /Rect was written with reals, filled and saved by PdfCraft's own writer, +/// which writes a whole real as an integer: the rectangle is the same numbers (ISO 32000-2 +/// §7.3.3, "wherever a real number is expected, an integer may be used instead"), so the save is a +/// form fill. A rectangle whose numbers did change is still a moved widget. +#[test] +fn a_widget_rect_saved_back_as_integers_is_a_form_fill() { + let field = "<< /Type /Annot /Subtype /Widget /FT /Tx /T (Name) /V (Ann) /Rect [20.0 20.0 120.0 40.0] /P 3 0 R >>"; + let signed = File::new().revision(&[ + (1, Row::Obj(0, 0, "<< /Type /Catalog /Pages 2 0 R /AcroForm << /Fields [5 0 R] >> >>")), + (2, Row::Obj(0, 0, PAGES)), + (3, Row::Obj(0, 0, &page("/Annots [5 0 R]"))), + (4, Row::Obj(0, 0, &page("/Annots []"))), + (5, Row::Obj(0, 0, field)), + ]); + let save = |rect: Option| { + let mut doc = signed.open(); + doc.update_dict(r(5, 0), |w| { + w.set(b"V".to_vec(), Object::String(pdfcraft_cos::PdfString::literal("Bob"))); + if let Some(x) = rect { + w.set(b"Rect".to_vec(), Object::Array(vec![Object::Real(x), Object::Real(20.0), Object::Real(120.0), Object::Real(40.0)])); + } + }) + .unwrap(); + let bytes = pdfcraft_cos::write_incremental(&doc, &pdfcraft_cos::SaveOptions { object_streams: false, ..Default::default() }).unwrap(); + Document::open(Arc::new(bytes)).unwrap() + }; + let filled = save(None); + let rect = filled.resolve(&Object::Ref(r(5, 0))).as_dict().and_then(|w| w.get(b"Rect").cloned()); + assert_eq!( + rect, + Some(Object::Array(vec![Object::Int(20), Object::Int(20), Object::Int(120), Object::Int(40)])), + "the writer respells whole reals" + ); + assert_eq!(verdict(&filled, signed.open(), Some(2)), Modification::Allowed(vec![super::kind::FORM_FILL])); + let moved = save(Some(20.5)); + assert_eq!(verdict(&moved, signed.open(), Some(2)), Modification::Disallowed(vec![super::kind::PAGE_CONTENT])); +} diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index 0bb7b7cf5d..58f49d3449 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -6535,8 +6535,8 @@ milestone = "M9" status = "shipped" commands = ["sign.panel"] tools = ["sign_list"] -evidence = ["crates/sign/tests/it/pdf.rs::later_changes_are_classified_under_the_signature_permissions", "crates/sign/tests/it/pdf.rs::signing_an_existing_field_and_counter_signing", "crates/engine/src/tests.rs::signing_saving_trusting_and_commenting_afterwards", "crates/sign/tests/it/pdf.rs::validation_data_and_xmp_added_after_signing_are_not_tampering", "crates/sign/tests/it/pdf.rs::a_dss_entry_does_not_make_the_object_it_names_part_of_the_store", "crates/sign/tests/it/pdf.rs::an_existing_dictionary_is_not_a_store_part_because_the_store_names_it", "crates/sign/tests/it/pdf.rs::an_existing_array_stays_what_it_is_when_the_store_points_at_it", "crates/sign/tests/it/pdf.rs::a_signed_store_may_grow_but_not_lose_or_swap_entries", "crates/sign/tests/it/pdf.rs::a_dictionary_labelled_metadata_is_not_a_metadata_change", "crates/sign/tests/it/pdf.rs::a_stream_already_labelled_metadata_when_signed_is_not_metadata_unless_the_catalog_names_it", "crates/sign/tests/it/pdf.rs::the_catalogs_own_xmp_stream_may_still_be_rewritten_after_signing", "crates/sign/tests/it/pdf.rs::rewritten_page_resources_are_page_content_whatever_keys_they_carry", "crates/sign/tests/it/pdf.rs::rewriting_a_certified_forms_xfa_data_is_a_form_fill", "crates/sign/tests/it/pdf.rs::the_catalogs_perms_do_not_change_under_no_changes", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_respelled_in_hex_is_no_change", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_whose_bytes_change_is_still_a_change"] -notes = "revision diff classified as signing, form fill, comments, metadata, page content or structure. Later revisions that add DSS data (new /Certs, /CRLs, /OCSPs, /VRI objects, or ones of the signed store that only grew) or rewrite the catalog's XMP are permitted changes; a /DSS entry that names other objects, or a /Type /DSS or /Type /Metadata label on an existing dictionary or stream, does not excuse changing them (negative tests at every DocMDP level). A signed page's resources (fonts, XObjects, graphics states and what they reach) are page content by the role they held when signed, whatever keys a later revision writes into them (#1787). Rewriting the XFA `datasets` or `form` packet the signed revision named is form fill, as a Reader Extensions (UR3) save of a certified form does (#2759); the packets' roles come from the signed revision. Under P=1 a change to the catalog's /Perms breaks the certification. Signed objects are compared by value: a string is its bytes, so one a later save rewrites from literal to hexadecimal (or back) is unchanged (ISO 32000-2 §7.3.4.1)." +evidence = ["crates/sign/tests/it/pdf.rs::later_changes_are_classified_under_the_signature_permissions", "crates/sign/tests/it/pdf.rs::signing_an_existing_field_and_counter_signing", "crates/engine/src/tests.rs::signing_saving_trusting_and_commenting_afterwards", "crates/sign/tests/it/pdf.rs::validation_data_and_xmp_added_after_signing_are_not_tampering", "crates/sign/tests/it/pdf.rs::a_dss_entry_does_not_make_the_object_it_names_part_of_the_store", "crates/sign/tests/it/pdf.rs::an_existing_dictionary_is_not_a_store_part_because_the_store_names_it", "crates/sign/tests/it/pdf.rs::an_existing_array_stays_what_it_is_when_the_store_points_at_it", "crates/sign/tests/it/pdf.rs::a_signed_store_may_grow_but_not_lose_or_swap_entries", "crates/sign/tests/it/pdf.rs::a_dictionary_labelled_metadata_is_not_a_metadata_change", "crates/sign/tests/it/pdf.rs::a_stream_already_labelled_metadata_when_signed_is_not_metadata_unless_the_catalog_names_it", "crates/sign/tests/it/pdf.rs::the_catalogs_own_xmp_stream_may_still_be_rewritten_after_signing", "crates/sign/tests/it/pdf.rs::rewritten_page_resources_are_page_content_whatever_keys_they_carry", "crates/sign/tests/it/pdf.rs::rewriting_a_certified_forms_xfa_data_is_a_form_fill", "crates/sign/tests/it/pdf.rs::the_catalogs_perms_do_not_change_under_no_changes", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_respelled_in_hex_is_no_change", "crates/sign/src/pdf/revision_classification_tests.rs::a_string_whose_bytes_change_is_still_a_change", "crates/sign/src/pdf/revision_classification_tests.rs::a_widget_rect_saved_back_as_integers_is_a_form_fill"] +notes = "revision diff classified as signing, form fill, comments, metadata, page content or structure. Later revisions that add DSS data (new /Certs, /CRLs, /OCSPs, /VRI objects, or ones of the signed store that only grew) or rewrite the catalog's XMP are permitted changes; a /DSS entry that names other objects, or a /Type /DSS or /Type /Metadata label on an existing dictionary or stream, does not excuse changing them (negative tests at every DocMDP level). A signed page's resources (fonts, XObjects, graphics states and what they reach) are page content by the role they held when signed, whatever keys a later revision writes into them (#1787). Rewriting the XFA `datasets` or `form` packet the signed revision named is form fill, as a Reader Extensions (UR3) save of a certified form does (#2759); the packets' roles come from the signed revision. Under P=1 a change to the catalog's /Perms breaks the certification. Signed objects are compared by value: a string is its bytes, so one a later save rewrites from literal to hexadecimal (or back) is unchanged (ISO 32000-2 §7.3.4.1), and a number is its value, so a signed widget /Rect or page /MediaBox of whole reals saved back as integers is unchanged (§7.3.3)." [[feature]] id = "sign.view-signed-version" From b07a84f00316949cee4f2320d557e899d4ad4ee3 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:29:31 +0800 Subject: [PATCH 04/75] M5: a comment whose colour is written [r g b null] is drawn, as CMYK with the null read as 0 (view.render-annotation-appearances) (#6503) * M5: a comment whose colour is written [r g b null] is drawn, as CMYK with the null read as 0 (view.render-annotation-appearances) ISO 32000-2 Table 166 /C (and /IC, Tables 178, 180, 181; /MK /BC and /BG, Table 192) is an array of 0, 1, 3 or 4 numbers whose length picks the colour space. A four-entry array with a non-number in it, such as [1 0 0 null] (PDFBOX-2699), made the /C parser give up, so a shape without an appearance stream was not drawn at all. PDFium, MuPDF and Poppler all read such an array as CMYK with the non-number as 0. - annot: a four-entry /C or /IC reads non-numbers as 0; other lengths still need every entry to be a number. - annot: the properties' fill uses the same parser (it read [r g b null] as RGB before), and /C no longer drops a non-number to make three numbers. - forms: /MK /BC and /BG read a four-entry array the same way instead of dropping the non-number and drawing RGB. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M5: comment properties report the stroke colour as drawn; a reference among four colour entries is not read as 0 (view.render-annotation-appearances) props() reads a non-FreeText /C through the appearance parser, so a [1 0 0 null] stroke reports CMYK cyan as drawn, and grey and CMYK strokes report their converted colour. A reference among the four entries of a /C array (ISO 32000-2 7.3.10) is left unreadable instead of standing in as 0, so /C [1 0 0 6 0 R] is never drawn cyan. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M5: re-run CI (fill_sign Escape test depends on the clock's second, passes locally) Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M5: the comment list reports a comment's colour as drawn (view.render-annotation-appearances) summaries() accepted only a three-entry /C, so a Square with /C [1 0 0 null], drawn cyan and reported cyan by props(), had no colour in the comment list; grey and CMYK colours were dropped the same way. Non-FreeText comments now read /C through drawn_color (the array itself followed when it is given by reference), keeping the three-number fallback and the FreeText /DA handling. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M5: the render inspector lists a comment's colour as drawn (view.render-annotation-appearances) The inspector accepted only a three-entry /C, with a referenced entry read as 0, so a /C [1 0 0 null], a grey or a CMYK colour listed none, unlike the comment list and the drawing. Non-FreeText comments now read /C by the appearance builder's rule (one entry grey, four CMYK with a non-number as 0 and a reference unreadable, finite and clamped), with the three-number fallback reading referenced entries. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn --------- Co-authored-by: mtlcl410 Co-authored-by: Claude Opus 5.5 --- crates/annot/src/appearance.rs | 19 +++++-- crates/annot/src/info.rs | 40 +++++++++++---- crates/annot/src/tests.rs | 65 ++++++++++++++++++++++++ crates/forms/src/appearance.rs | 10 +++- crates/forms/src/tests.rs | 32 ++++++++++++ crates/render/src/inspect/annotations.rs | 40 +++++++++++++-- crates/render/src/inspect/tests.rs | 34 +++++++++++++ parity/acrobat-features.toml | 4 +- 8 files changed, 222 insertions(+), 22 deletions(-) diff --git a/crates/annot/src/appearance.rs b/crates/annot/src/appearance.rs index dafc4b01b1..7996567f04 100644 --- a/crates/annot/src/appearance.rs +++ b/crates/annot/src/appearance.rs @@ -205,11 +205,24 @@ fn quads(context: &AppearanceContext<'_>) -> Option> { (!q.is_empty()).then_some(q) } -/// `/C`-style colour arrays: none (transparent), gray, RGB or CMYK. -fn color(d: &Dict, key: &[u8]) -> Option> { +/// `/C`-style colour arrays (ISO 32000-2 Table 166): none (transparent), gray, RGB or CMYK. +/// The entries should all be numbers; a four-entry array with a non-number in it, such as +/// `[1 0 0 null]`, is still read as CMYK with that entry 0, as PDFium, MuPDF and Poppler read it. +/// A reference is not such an entry: it stands for a value this builder can't see (§7.3.10), so +/// the colour is unreadable, as the other colour and number readers here treat one. +pub(crate) fn color(d: &Dict, key: &[u8]) -> Option> { let c = match d.get(key) { None => return Some(None), - Some(o) => o.as_array()?.iter().map(|o| o.as_f64()).collect::>>()?, + Some(o) => match o.as_array()? { + a if a.len() == 4 => a + .iter() + .map(|o| match o { + Object::Ref(_) => None, + o => Some(o.as_f64().unwrap_or(0.0)), + }) + .collect::>>()?, + a => a.iter().map(|o| o.as_f64()).collect::>>()?, + }, }; Some(match c.as_slice() { [] => None, diff --git a/crates/annot/src/info.rs b/crates/annot/src/info.rs index 8af76cb2ed..c4913e7ab8 100644 --- a/crates/annot/src/info.rs +++ b/crates/annot/src/info.rs @@ -105,7 +105,10 @@ pub fn summaries(doc: &Document) -> Vec { let r = nums(b"Rect"); let rect = if r.len() == 4 { [r[0].min(r[2]), r[1].min(r[3]), r[0].max(r[2]), r[1].max(r[3])] } else { [0.0; 4] }; let c = nums(b"C"); - let mut color = (c.len() == 3).then(|| [c[0], c[1], c[2]]); + let rgb = (c.len() == 3).then(|| [c[0], c[1], c[2]]); + // Other comments report the stroke colour as it is drawn (grey, CMYK, a null entry + // read as 0), like `props`; three numbers, some given by reference, are still read. + let mut color = if subtype == "FreeText" { rgb } else { summary_color(doc, d).or(rgb) }; if subtype == "FreeText" && let Some(da) = text_value(doc, d, b"DA") { @@ -244,8 +247,7 @@ pub fn props(doc: &Document, page: usize, index: usize) -> Option { let obj = doc.resolve(&entry); let d = obj.as_dict()?; let subtype = String::from_utf8_lossy(d.name(b"Subtype")?).into_owned(); - let c: Vec = d.get(b"C").and_then(|o| o.as_array().map(|a| a.iter().filter_map(|x| x.as_f64()).collect())).unwrap_or_default(); - let color = if subtype == "FreeText" { Some(appearance::parse_da(d).0) } else { (c.len() == 3).then(|| [c[0], c[1], c[2]]) }; + let color = if subtype == "FreeText" { Some(appearance::parse_da(d).0) } else { drawn_color(d, b"C") }; let width = matches!(subtype.as_str(), "Square" | "Circle" | "Line" | "Ink" | "Polygon" | "PolyLine" | "FreeText").then(|| { if subtype == "FreeText" { if !d.contains(b"BS") && !d.contains(b"Border") { 0.0 } else { appearance::border_width(d) } @@ -275,16 +277,32 @@ pub fn props(doc: &Document, page: usize, index: usize) -> Option { }) } -/// A shape’s `/IC` or FreeText `/C` as RGB. Grey and CMYK fills are converted. +/// A shape’s `/IC` or FreeText `/C` as RGB, read as it is drawn. Grey and CMYK fills are converted. fn fill_of(d: &Dict) -> Option { let key: &[u8] = if d.name(b"Subtype") == Some(b"FreeText") { b"C" } else { b"IC" }; - let c: Vec = d.get(key)?.as_array()?.iter().filter_map(|x| x.as_f64()).collect(); - let rgb = match c.as_slice() { - [g] => [*g; 3], - [r, g, b] => [*r, *g, *b], - [c, m, y, k] => [(1.0 - c) * (1.0 - k), (1.0 - m) * (1.0 - k), (1.0 - y) * (1.0 - k)], - _ => return None, - }; + drawn_color(d, key) +} + +/// `/C` as [`drawn_color`] reads it, the array itself followed when it is given by reference. +/// Only an array of at most four entries is copied out of the reference. +fn summary_color(doc: &Document, d: &Dict) -> Option<[f32; 3]> { + let rgb = match d.get(b"C") { + Some(o @ Object::Ref(_)) => { + let held = doc.resolve(o); + let entries = held.as_array().filter(|a| a.len() <= 4)?; + let mut direct = Dict::new(); + direct.set(b"C".to_vec(), Object::Array(entries.clone())); + drawn_color(&direct, b"C") + } + _ => drawn_color(d, b"C"), + }?; + Some(rgb.map(|x| x as f32)) +} + +/// The colour array at `key` as RGB, read as the appearance builder draws it: grey and CMYK are +/// converted, and a four-entry array with a non-number in it is CMYK with that entry 0. +fn drawn_color(d: &Dict, key: &[u8]) -> Option { + let rgb = appearance::color(d, key).flatten()?; finite(&rgb).then(|| rgb.map(|x| x.clamp(0.0, 1.0))) } diff --git a/crates/annot/src/tests.rs b/crates/annot/src/tests.rs index 0c9ac7ed5a..9e414078e9 100644 --- a/crates/annot/src/tests.rs +++ b/crates/annot/src/tests.rs @@ -2786,6 +2786,71 @@ fn indirect_typewriter_names_keep_outer_whitespace_in_summaries() { check(&reopen(&doc)); } +/// PDFBOX-2699: a `/C` written `[1 0 0 null]` (Table 166 asks for numbers) is still a four-entry +/// colour, CMYK with the null as 0 as PDFium, MuPDF and Poppler read it, so a shape without an +/// appearance is drawn instead of dropped. `/IC` is read the same way, and three numbers stay RGB. +/// The properties and the comment list report the colours as drawn. +#[test] +fn a_four_entry_colour_with_a_null_is_cmyk_with_zero() { + let doc = document_of(&[ + b"<< /Type /Catalog /Pages 2 0 R >>", + b"<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 100 40] >>", + b"<< /Type /Page /Parent 2 0 R /Annots [4 0 R 5 0 R] >>", + b"<< /Type /Annot /Subtype /Square /Rect [5 5 45 35] /C [1 0 0 null] /IC [0 0 1 null] >>", + b"<< /Type /Annot /Subtype /Square /Rect [55 5 95 35] /C [1 0 0] >>", + ]); + let display = with_missing_appearances(&doc).expect("both squares get a fallback appearance"); + let squares = list(&display, 0); + let cmyk = ap_content(&display, &squares[0]); + assert!(cmyk.contains("0 1 1 RG") && cmyk.contains("1 1 0 rg"), "CMYK (1,0,0,0) stroke and (0,0,1,0) fill: {cmyk}"); + let rgb = ap_content(&display, &squares[1]); + assert!(rgb.contains("1 0 0 RG"), "three numbers stay RGB: {rgb}"); + // The properties read the stroke and the fill as they are drawn. + let p = props(&doc, 0, 0).expect("the square has properties"); + assert_eq!((p.color, p.fill), (Some([0.0, 1.0, 1.0]), Some([1.0, 1.0, 0.0]))); + assert_eq!(props(&doc, 0, 1).and_then(|p| p.color), Some([1.0, 0.0, 0.0])); + // The comment list reports the same colours. + let colors: Vec> = summaries(&doc).iter().map(|s| s.color).collect(); + assert_eq!(colors, [Some([0.0, 1.0, 1.0]), Some([1.0, 0.0, 0.0])]); +} + +/// A `/C` given by reference is followed for the comment list, as three numbers were, and read +/// like a direct one: `[0 0 1 null]` is CMYK yellow, and a referenced grey level is grey. +#[test] +fn a_referenced_colour_is_listed_like_a_direct_one() { + let doc = document_of(&[ + b"<< /Type /Catalog /Pages 2 0 R >>", + b"<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 100 40] >>", + b"<< /Type /Page /Parent 2 0 R /Annots [4 0 R 5 0 R] >>", + b"<< /Type /Annot /Subtype /Square /Rect [5 5 45 35] /C 6 0 R >>", + b"<< /Type /Annot /Subtype /Circle /Rect [55 5 95 35] /C 7 0 R >>", + b"[0 0 1 null]", + b"[0.5]", + ]); + let colors: Vec> = summaries(&doc).iter().map(|s| s.color).collect(); + assert_eq!(colors, [Some([1.0, 1.0, 0.0]), Some([0.5, 0.5, 0.5])]); +} + +/// A reference among the four entries is not a non-number read as 0: `/C [1 0 0 6 0 R]` with +/// object 6 the number 1 is CMYK black (§7.3.10), so drawing it cyan would be wrong. The builder +/// doesn't follow it and draws no appearance, as for any colour it can't read. +#[test] +fn a_four_entry_colour_with_a_reference_is_not_read_as_zero() { + let doc = document_of(&[ + b"<< /Type /Catalog /Pages 2 0 R >>", + b"<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 100 40] >>", + b"<< /Type /Page /Parent 2 0 R /Annots [4 0 R 5 0 R] >>", + b"<< /Type /Annot /Subtype /Square /Rect [5 5 45 35] /C [1 0 0 6 0 R] >>", + b"<< /Type /Annot /Subtype /Square /Rect [55 5 95 35] /C [1 0 0] >>", + b"1", + ]); + let display = with_missing_appearances(&doc).expect("the RGB square gets a fallback appearance"); + let squares = list(&display, 0); + assert!(!squares[0].contains(b"AP"), "no cyan stand-in for CMYK black"); + assert_eq!(props(&doc, 0, 0).and_then(|p| p.color), None); + assert_eq!(summaries(&doc)[0].color, None, "nor in the comment list"); +} + /// The `y` of every `1 0 0 1 x y Tm` in `content`: where each piece of text sits. fn baselines(content: &str) -> Vec { content.split(" Tm").filter_map(|s| s.rsplit(' ').next()?.parse().ok()).collect() diff --git a/crates/forms/src/appearance.rs b/crates/forms/src/appearance.rs index 91cc352701..891a533eaa 100644 --- a/crates/forms/src/appearance.rs +++ b/crates/forms/src/appearance.rs @@ -399,9 +399,17 @@ fn helvetica() -> Object { Object::Dict(f) } +/// An `/MK /BC` or `/BG` colour (ISO 32000-2 Table 192) as a fill operator. A four-entry array +/// with a non-number in it, such as `[1 0 0 null]`, is CMYK with that entry 0, as annotation `/C` +/// colours are read. fn color_array(doc: &Document, mk: Option<&Dict>, key: &[u8]) -> Option { let a = doc.resolve(mk?.get(key)?); - let v: Vec = a.as_array()?.iter().filter_map(|x| doc.resolve(x).as_f64()).collect(); + let a = a.as_array()?; + let v: Vec = if a.len() == 4 { + a.iter().map(|x| doc.resolve(x).as_f64().unwrap_or(0.0)).collect() + } else { + a.iter().filter_map(|x| doc.resolve(x).as_f64()).collect() + }; match v.as_slice() { [g] => Some(format!("{} g", n(*g))), [r, g, b] => Some(format!("{} {} {} rg", n(*r), n(*g), n(*b))), diff --git a/crates/forms/src/tests.rs b/crates/forms/src/tests.rs index 353545f2b3..6cf207d0de 100644 --- a/crates/forms/src/tests.rs +++ b/crates/forms/src/tests.rs @@ -5706,3 +5706,35 @@ fn locked_button_export_edits_leave_the_document_unchanged() { assert_eq!(write_incremental(&doc, &SaveOptions::default()).unwrap(), before); } } + +/// A `/MK /BC` or `/BG` written `[1 0 0 null]` is CMYK with the null as 0, as annotation `/C` +/// colours are read (PDFBOX-2699), not RGB from the three numbers left; three numbers stay RGB. +#[test] +fn a_four_entry_mk_colour_with_a_null_is_cmyk_with_zero() { + let widget = |name: &str, mk: &str| { + format!("<< /Type /Annot /Subtype /Widget /FT /Btn /T ({name}) /Rect [50 700 70 720] /P 3 0 R /MK << {mk} >> /BS << /W 1 /S /S >> >>") + }; + let objs: Vec = vec![ + "<< /Type /Catalog /Pages 2 0 R /AcroForm 4 0 R >>".into(), + "<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 600 800] >>".into(), + "<< /Type /Page /Parent 2 0 R /Annots [5 0 R 6 0 R] >>".into(), + "<< /Fields [5 0 R 6 0 R] >>".into(), + widget("cmyk", "/BG [0 0 1 null] /BC [1 0 0 null]"), + widget("rgb", "/BG [0 0 1] /BC [1 0 0]"), + ]; + let mut doc = document(&objs); + let off = |doc: &mut Document, name: &str| { + redraw_field(doc, name).unwrap(); + let all = fields(doc); + let w = field(&all, name).widgets[0].obj; + let wd = doc.get(w).as_dict().cloned().unwrap(); + let n = doc.resolve(wd.get(b"AP").unwrap().as_dict().unwrap().get(b"N").unwrap()); + let r = n.as_dict().unwrap().reference(b"Off").expect("an Off appearance"); + let Object::Stream(s) = &*doc.get(r) else { panic!("Off is a stream") }; + String::from_utf8_lossy(&s.decoded().unwrap()).into_owned() + }; + let cmyk = off(&mut doc, "cmyk"); + assert!(cmyk.contains("0 0 1 0 k") && cmyk.contains("1 0 0 0 K"), "CMYK background and border: {cmyk}"); + let rgb = off(&mut doc, "rgb"); + assert!(rgb.contains("0 0 1 rg") && rgb.contains("1 0 0 RG"), "three numbers stay RGB: {rgb}"); +} diff --git a/crates/render/src/inspect/annotations.rs b/crates/render/src/inspect/annotations.rs index 7c4543b733..5ed635fdcb 100644 --- a/crates/render/src/inspect/annotations.rs +++ b/crates/render/src/inspect/annotations.rs @@ -22,6 +22,31 @@ pub(super) const MAX_ANNOTATION_BYTES: usize = 64 << 20; /// key does not fit is not listed. pub(super) const MAX_LINK_ATTACHMENT_TEXT_BYTES: usize = 16 << 20; +/// A colour array as RGB, read as pdfcraft-annot's appearance builder reads it (`appearance::color` +/// and `drawn_color`): one entry is grey, three RGB, four CMYK; among four entries a non-number is +/// 0 but a reference is unreadable, otherwise every entry must be a number. `None` for anything +/// else, an empty array or a value out of range; each component is clamped to 0..=1. +fn drawn_color(c: &[Object]) -> Option<[f32; 3]> { + let v: Vec = if c.len() == 4 { + c.iter() + .map(|o| match o { + Object::Reference(_) => None, + o => Some(o.as_float().unwrap_or(0.0)), + }) + .collect::>()? + } else { + c.iter().map(|o| o.as_float().ok()).collect::>()? + }; + let rgb = match v.as_slice() { + [g] => [*g; 3], + [r, g, b] => [*r, *g, *b], + [c, m, y, k] => [(1.0 - c) * (1.0 - k), (1.0 - m) * (1.0 - k), (1.0 - y) * (1.0 - k)], + _ => return None, + }; + // pdfcraft-annot's `finite`: finite and below 1e7 in magnitude. + rgb.iter().all(|x| x.is_finite() && x.abs() < 1e7).then(|| rgb.map(|x| x.clamp(0.0, 1.0))) +} + /// The bytes a listed annotation keeps. pub(super) fn retained(a: &Annotation) -> usize { let text = [&a.author, &a.contents, &a.modified, &a.name, &a.in_reply_to, &a.state, &a.intent]; @@ -119,13 +144,18 @@ impl<'a> Inspector<'a> { continue; } let rect = rect4(self.resolve(d.get(b"Rect").unwrap_or(&Object::Null))); - let mut color = match d.get(b"C").map(|o| self.resolve(o)) { - Ok(Object::Array(c)) if c.len() == 3 => { - let f = |i: usize| c[i].as_float().unwrap_or(0.0); - Some([f(0), f(1), f(2)]) - } + let c = match d.get(b"C").map(|o| self.resolve(o)) { + Ok(Object::Array(c)) => Some(c), _ => None, }; + // Three numbers, some given by reference, with a non-number read as 0. + let rgb = c.filter(|c| c.len() == 3).map(|c| { + let f = |i: usize| c.get(i).map_or(0.0, |o| self.resolve(o).as_float().unwrap_or(0.0)); + [f(0), f(1), f(2)] + }); + // Other comments list the stroke colour as pdfcraft-annot draws it and lists it + // (`summaries`): grey, CMYK, a non-number among four entries read as 0. + let mut color = if subtype == "FreeText" { rgb } else { c.and_then(|c| drawn_color(c)).or(rgb) }; // A text box's /C is its background; its text colour is in /DA. if subtype == "FreeText" && let Some(da) = self.text(d, b"DA") diff --git a/crates/render/src/inspect/tests.rs b/crates/render/src/inspect/tests.rs index 907401f8bb..012e40f79a 100644 --- a/crates/render/src/inspect/tests.rs +++ b/crates/render/src/inspect/tests.rs @@ -877,6 +877,40 @@ fn an_annotation_reached_through_a_reference_chain_is_listed_once() { } } +/// Comments list their `/C` as pdfcraft-annot draws and lists it: four entries with a null are +/// CMYK with the null as 0, one entry is grey, a reference among four entries is unreadable, +/// three numbers (some by reference) are RGB, and a text box keeps its three-number `/C`. +#[test] +fn comment_colours_are_listed_as_drawn() { + let pdf = "%PDF-1.7 +1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj + 2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj + 3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /Annots [4 0 R 5 0 R 6 0 R 7 0 R 8 0 R 9 0 R] >> endobj + 4 0 obj << /Type /Annot /Subtype /Square /Rect [0 0 10 10] /C [1 0 0 null] >> endobj + 5 0 obj << /Type /Annot /Subtype /Square /Rect [0 20 10 30] /C 10 0 R >> endobj + 6 0 obj << /Type /Annot /Subtype /Square /Rect [0 40 10 50] /C [1 0 0 11 0 R] >> endobj + 7 0 obj << /Type /Annot /Subtype /Square /Rect [0 60 10 70] /C [11 0 R 0 0] >> endobj + 8 0 obj << /Type /Annot /Subtype /Circle /Rect [0 80 10 90] /C [0 0 0 1] >> endobj + 9 0 obj << /Type /Annot /Subtype /FreeText /Rect [0 100 10 110] /C [0 1 0] >> endobj + 10 0 obj [0.5] endobj + 11 0 obj 1 endobj + trailer << /Root 1 0 R >> +%%EOF"; + let mut doc = Document::load_mem(pdf.as_bytes()).expect("loads"); + let mut saved = Vec::new(); + doc.save_to(&mut saved).expect("saves"); + let lazy = LazyStructure::new(Arc::new(saved), None).expect("opens"); + let page = PageInfo { width: 200.0, height: 200.0, crop: [0.0, 0.0, 200.0, 200.0], rotation: 0, label: String::new() }; + for reader in [&doc as &dyn Structure, &lazy] { + let mut info = DocInfo { pages: vec![page.clone()], ..Default::default() }; + Inspector::new(reader).fill(&mut info); + let mut listed: Vec<(f32, Option<[f32; 3]>)> = info.annotations.iter().map(|a| (a.rect[1], a.color)).collect(); + listed.sort_by(|a, b| a.0.total_cmp(&b.0)); + let colors: Vec> = listed.into_iter().map(|(_, c)| c).collect(); + assert_eq!(colors, [Some([0.0, 1.0, 1.0]), Some([0.5, 0.5, 0.5]), None, Some([1.0, 0.0, 0.0]), Some([0.0, 0.0, 0.0]), Some([0.0, 1.0, 0.0])]); + } +} + /// A reply finds its comment where the listing keeps it: on the first page of a page-shared /// `/Annots` array, whether its `/IRT` names the comment or an object holding a reference to it. #[test] diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index 67ee899f85..af026b3a85 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -818,8 +818,8 @@ milestone = "M2" status = "shipped" commands = [] tools = [] -evidence = ["crates/render/src/raster/tests.rs::widget_appearance_states", "crates/render/src/raster/tests.rs::highlights_without_an_appearance_are_drawn", "crates/engine/src/tests.rs::comments_without_appearances_are_drawn_but_not_saved", "crates/engine/src/tests.rs::cli_display_adds_missing_appearances_without_changing_the_source", "crates/engine/src/tests.rs::fields_left_to_the_viewer_are_displayed_from_their_values", "crates/forms/src/tests.rs::need_appearances_draws_every_text_and_choice_field_in_the_display_copy", "crates/forms/src/tests.rs::the_field_appearance_cache_sees_every_input", "crates/engine/src/tests.rs::buttons_and_combs_left_to_the_viewer_are_displayed", "crates/forms/src/tests.rs::check_boxes_without_an_appearance_show_their_caption_symbol", "crates/forms/src/tests.rs::push_buttons_without_an_appearance_show_their_caption_and_frame", "crates/forms/src/tests.rs::need_appearances_draws_check_boxes_and_radio_buttons_from_their_captions", "crates/engine/src/tests.rs::need_appearances_buttons_are_displayed_from_their_captions", "crates/render/src/raster/tests.rs::invisible_unknown_annotations_and_popups_paint_nothing", "crates/print/src/tests.rs::invisible_annotations_of_unknown_types_do_not_print", "crates/engine/src/tests.rs::stamps_without_appearances_draw_the_stamp_they_name", "crates/automation/tests/it/automation.rs::page_render_draws_stamps_without_an_appearance"] -notes = "comments without /AP (FreeText, Ink, notes, stamps, shapes, markup) are drawn from their dictionaries in the page view, page_render and page images, including the CLI render command, never saved; a stamp from another program draws the stamp its /Name names (the fourteen standard names of ISO 32000-2 Table 184, Draft when there is none) or, for any other name, a grey frame with both diagonals; text and choice fields are drawn from their values when the form sets /NeedAppearances (every one) or a widget has no /N (#3470), also display-only and cached per widget (a field edit on a 1000-field /NeedAppearances form takes ~13 ms against ~7 ms on a form that needs none) (a read-only document still gets the renderer's Highlight fallback only); check boxes and radio buttons with no appearance for their /AS are drawn display-only from /MK (background, border, and when on the /MK /CA ZapfDingbats symbol in the /DA colour, a check or a dot by default), and when the form sets /NeedAppearances so is every one that has an /MK, over the appearance the file carries (ISO 32000-2 Table 224; one without /MK keeps its own), and push buttons with none from /MK and /BS with their caption; Link borders without /AP and printing are not covered yet; an annotation of a non-standard type with the Invisible flag (ISO 32000-2 Table 167) is neither drawn nor printed, and a popup's own /AP is never painted on the page (its parent's text shows in the popup window, §12.5.6.14)" +evidence = ["crates/render/src/raster/tests.rs::widget_appearance_states", "crates/render/src/raster/tests.rs::highlights_without_an_appearance_are_drawn", "crates/engine/src/tests.rs::comments_without_appearances_are_drawn_but_not_saved", "crates/engine/src/tests.rs::cli_display_adds_missing_appearances_without_changing_the_source", "crates/engine/src/tests.rs::fields_left_to_the_viewer_are_displayed_from_their_values", "crates/forms/src/tests.rs::need_appearances_draws_every_text_and_choice_field_in_the_display_copy", "crates/forms/src/tests.rs::the_field_appearance_cache_sees_every_input", "crates/engine/src/tests.rs::buttons_and_combs_left_to_the_viewer_are_displayed", "crates/forms/src/tests.rs::check_boxes_without_an_appearance_show_their_caption_symbol", "crates/forms/src/tests.rs::push_buttons_without_an_appearance_show_their_caption_and_frame", "crates/forms/src/tests.rs::need_appearances_draws_check_boxes_and_radio_buttons_from_their_captions", "crates/engine/src/tests.rs::need_appearances_buttons_are_displayed_from_their_captions", "crates/render/src/raster/tests.rs::invisible_unknown_annotations_and_popups_paint_nothing", "crates/print/src/tests.rs::invisible_annotations_of_unknown_types_do_not_print", "crates/engine/src/tests.rs::stamps_without_appearances_draw_the_stamp_they_name", "crates/automation/tests/it/automation.rs::page_render_draws_stamps_without_an_appearance", "crates/annot/src/tests.rs::a_four_entry_colour_with_a_null_is_cmyk_with_zero", "crates/forms/src/tests.rs::a_four_entry_mk_colour_with_a_null_is_cmyk_with_zero"] +notes = "comments without /AP (FreeText, Ink, notes, stamps, shapes, markup) are drawn from their dictionaries in the page view, page_render and page images, including the CLI render command, never saved; a four-entry colour with a non-number in it (/C, /IC, /MK /BC and /BG written [1 0 0 null]) is read as CMYK with that entry 0; a stamp from another program draws the stamp its /Name names (the fourteen standard names of ISO 32000-2 Table 184, Draft when there is none) or, for any other name, a grey frame with both diagonals; text and choice fields are drawn from their values when the form sets /NeedAppearances (every one) or a widget has no /N (#3470), also display-only and cached per widget (a field edit on a 1000-field /NeedAppearances form takes ~13 ms against ~7 ms on a form that needs none) (a read-only document still gets the renderer's Highlight fallback only); check boxes and radio buttons with no appearance for their /AS are drawn display-only from /MK (background, border, and when on the /MK /CA ZapfDingbats symbol in the /DA colour, a check or a dot by default), and when the form sets /NeedAppearances so is every one that has an /MK, over the appearance the file carries (ISO 32000-2 Table 224; one without /MK keeps its own), and push buttons with none from /MK and /BS with their caption; Link borders without /AP and printing are not covered yet; an annotation of a non-standard type with the Invisible flag (ISO 32000-2 Table 167) is neither drawn nor printed, and a popup's own /AP is never painted on the page (its parent's text shows in the popup window, §12.5.6.14)" [[feature]] id = "view.render-error-isolation" From 9b231f241d58ee221015870c7533b2a611d5d9a5 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:29:44 +0800 Subject: [PATCH 05/75] perf(render): undo TIFF predictor 2 a byte row at a time at 8 and 16 bits (#6505) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit TIFF predictor 2 (ISO 32000-2 §7.4.4.4, Table 10) went through the bit-level BitChunks/BitWriter path for every bit depth, re-reading the previous output pixel through a new BitReader per pixel. Whole-byte samples (8 and 16 bits per component) are now undone on byte slices, each sample added to the same component of the pixel to its left, wrapping at its width; 1-, 2- and 4-bit samples keep the bit-level path. The output is byte-identical, including the dropped partial last row. A page of 45 full-page 8-bit grey images so predicted (PDFBOX-1350) renders in about 4 s instead of 23 s with the same pixels. Vendored hayro-syntax patch 42. Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn Co-authored-by: mtlcl410 Co-authored-by: Claude Opus 5.5 --- crates/render/src/raster/tests.rs | 4 + .../src/raster/tests/tiff_predicted_images.rs | 94 +++++++++ parity/acrobat-features.toml | 4 +- vendor/README.md | 2 +- vendor/hayro-syntax/src/filter/lzw_flate.rs | 187 ++++++++++++++++++ 5 files changed, 288 insertions(+), 3 deletions(-) create mode 100644 crates/render/src/raster/tests/tiff_predicted_images.rs diff --git a/crates/render/src/raster/tests.rs b/crates/render/src/raster/tests.rs index be232d80eb..f6569e51fa 100644 --- a/crates/render/src/raster/tests.rs +++ b/crates/render/src/raster/tests.rs @@ -8983,6 +8983,10 @@ mod devicen_inks; // #5430: PNG-predicted Flate images are undone byte by byte (vendored hayro-syntax patch 35). mod png_predicted_images; +// TIFF-predicted Flate images at 8 and 16 bits are undone byte by byte (vendored hayro-syntax +// patch 42). +mod tiff_predicted_images; + // #3776: isolated DeviceCMYK groups blend in CMYK (vendored hayro-interpret patch 70, hayro // patch 27). mod isolated_cmyk_groups; diff --git a/crates/render/src/raster/tests/tiff_predicted_images.rs b/crates/render/src/raster/tests/tiff_predicted_images.rs new file mode 100644 index 0000000000..30a163daeb --- /dev/null +++ b/crates/render/src/raster/tests/tiff_predicted_images.rs @@ -0,0 +1,94 @@ +//! Flate images with TIFF predictor 2 (ISO 32000-2 §7.4.4.4, Table 10) at 8 and 16 bits per +//! component are undone a byte row at a time (vendored hayro-syntax patch 42), with the samples +//! the bit-level decoder gave. + +use super::rgb_at; +use crate::{PageRenderer, RenderConfig, RenderRequest, RequestKind}; +use flate2::{Compression, write::ZlibEncoder}; +use std::io::Write; +use std::sync::Arc; + +/// The page is 150 × 200 pt, drawn at 72 dpi; a 3 × 4 image fills it. +const WIDTH: usize = 150; +const HEIGHT: usize = 200; + +/// Renders a 3 × 4 image whose samples are `rows` (each row's bytes, before prediction), +/// differenced by TIFF predictor 2 with `pixel_len`-byte pixels of `sample_len`-byte samples. +fn render(rows: &[Vec], colour_space: &str, colors: usize, bpc: usize) -> Vec { + let (sample_len, pixel_len) = (bpc / 8, colors * bpc / 8); + let mut predicted = Vec::new(); + for row in rows { + let mut out = row.clone(); + for i in (pixel_len..row.len()).step_by(sample_len).rev() { + if sample_len == 2 { + let cur = u16::from_be_bytes([row[i], row[i + 1]]); + let left = u16::from_be_bytes([row[i - pixel_len], row[i - pixel_len + 1]]); + out[i..i + 2].copy_from_slice(&cur.wrapping_sub(left).to_be_bytes()); + } else { + out[i] = row[i].wrapping_sub(row[i - pixel_len]); + } + } + predicted.extend(out); + } + let mut encoder = ZlibEncoder::new(Vec::new(), Compression::default()); + encoder.write_all(&predicted).unwrap(); + let data = encoder.finish().unwrap(); + + let content = "q 150 0 0 200 0 0 cm /Im0 Do Q"; + let mut pdf = format!( + "%PDF-1.7\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\ + 2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n\ + 3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 150 200] /Resources << /XObject << /Im0 5 0 R >> >> /Contents 4 0 R >> endobj\n\ + 4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\ + 5 0 obj << /Type /XObject /Subtype /Image /Width 3 /Height 4 /ColorSpace /{colour_space} /BitsPerComponent {bpc} \ + /Filter /FlateDecode /DecodeParms << /Predictor 2 /Colors {colors} /BitsPerComponent {bpc} /Columns 3 >> /Length {} >> stream\n", + content.len(), + data.len() + ) + .into_bytes(); + pdf.extend_from_slice(&data); + pdf.extend_from_slice(b"\nendstream endobj\ntrailer << /Root 1 0 R >>\n%%EOF"); + + let mut renderer = PageRenderer::new(Arc::new(pdf), RenderConfig::default()); + let page = renderer.render(RenderRequest { page: 0, kind: RequestKind::Pixels, tile: None, scale: 1.0, tag: 0, preview: false }); + assert!(page.error.is_none(), "{:?}", page.error); + assert_eq!((page.width, page.height), (WIDTH as u32, HEIGHT as u32)); + page.rgba.to_vec() +} + +#[test] +fn an_eight_bit_rgb_image_with_tiff_predicted_rows_draws_its_colours() { + // Each pixel differs from the one to its left, and some differences wrap past 255. + let pixels: [[[u8; 3]; 3]; 4] = [ + [[255, 0, 0], [0, 255, 0], [0, 0, 255]], + [[10, 200, 30], [250, 5, 128], [0, 0, 0]], + [[255, 255, 255], [1, 2, 3], [128, 64, 32]], + [[0, 128, 255], [255, 128, 0], [64, 64, 64]], + ]; + let rows: Vec> = pixels.iter().map(|row| row.iter().flatten().copied().collect()).collect(); + let rgba = render(&rows, "DeviceRGB", 3, 8); + for (row, colours) in pixels.iter().enumerate() { + for (column, rgb) in colours.iter().enumerate() { + let (x, y) = (column * 50 + 25, row * 50 + 25); + assert_eq!(rgb_at(&rgba, WIDTH, x, y), *rgb, "image row {row}, column {column}"); + } + } +} + +#[test] +fn a_sixteen_bit_grey_image_with_tiff_predicted_rows_draws_its_levels() { + // The high byte decides the level drawn; the differences carry across the byte boundary. + let levels: [[u16; 3]; 4] = [[0xffff, 0x00ff, 0x8000], [0x0100, 0xff00, 0x4080], [0, 0x7fff, 0xc000], [0x2000, 0x2000, 0xe0ff]]; + let rows: Vec> = levels.iter().map(|row| row.iter().flat_map(|v| v.to_be_bytes()).collect()).collect(); + let rgba = render(&rows, "DeviceGray", 1, 16); + for (row, values) in levels.iter().enumerate() { + for (column, value) in values.iter().enumerate() { + let (x, y) = (column * 50 + 25, row * 50 + 25); + let [r, g, b] = rgb_at(&rgba, WIDTH, x, y); + let expected = i32::from(value >> 8); + for got in [r, g, b] { + assert!((i32::from(got) - expected).abs() <= 1, "image row {row}, column {column}: {got} for {value:#06x}"); + } + } + } +} diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index af026b3a85..abc7e8ceeb 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -118,8 +118,8 @@ milestone = "M1" status = "shipped" commands = [] tools = [] -evidence = ["crates/filters/src/lib.rs::chain_three_filters_with_predictor", "crates/filters/src/lzw.rs::spec_example", "crates/filters/src/predictor.rs::png_hand_vectors", "crates/filters/src/lib.rs::raw_flate_fallback_preserves_output_limits", "crates/cos/src/object.rs::raw_flate_fallback_streams_preserve_the_caller_limit"] -notes = "Raw-deflate fallback preserves output limits in strict and tolerant decoding, including misleading zlib headers. Image codecs remain outside this general-purpose filter implementation." +evidence = ["crates/filters/src/lib.rs::chain_three_filters_with_predictor", "crates/filters/src/lzw.rs::spec_example", "crates/filters/src/predictor.rs::png_hand_vectors", "crates/filters/src/lib.rs::raw_flate_fallback_preserves_output_limits", "crates/cos/src/object.rs::raw_flate_fallback_streams_preserve_the_caller_limit", "crates/render/src/raster/tests/tiff_predicted_images.rs::an_eight_bit_rgb_image_with_tiff_predicted_rows_draws_its_colours", "crates/render/src/raster/tests/tiff_predicted_images.rs::a_sixteen_bit_grey_image_with_tiff_predicted_rows_draws_its_levels"] +notes = "Raw-deflate fallback preserves output limits in strict and tolerant decoding, including misleading zlib headers. The renderer undoes PNG predictors, and TIFF predictor 2 at 8 and 16 bits per component, a byte row at a time. Image codecs remain outside this general-purpose filter implementation." [[feature]] id = "core.image-filters" diff --git a/vendor/README.md b/vendor/README.md index 7ae867cb11..aa8b43ba4a 100644 --- a/vendor/README.md +++ b/vendor/README.md @@ -43,7 +43,7 @@ Other resource categories retain their existing parent-chain lookup. | hayro | 0.7.1 | Apache-2.0 OR MIT | (1) `RenderSettings::{x_offset, y_offset}` to render one tile of a large page; (2) images over `MAX_IMAGE_PIXELS` (2^28) are skipped (a fuzzed `/W 4294967295` hung in resampling); (3) `tiling_cell_scale` keeps a tiling pattern's cell pixmap within 3000 px a side whatever its `/XStep`/`/YStep` (a fuzzed step far beyond `/BBox` allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed `/LW 9223372036854775807` allocated 10 GB in stroke expansion); (5) `usable_dash_pattern`: a stroke whose dash pattern could cut it into more than `MAX_DASHES_PER_STROKE` (250,000) pieces, counting kurbo's restart of the pattern at every subpath and `ClosePath`, is drawn solid (a fuzzed `/D [[11] 0]` on a line from x = 92234775807 allocated over 5 GB, `[0.000001] 0 d` on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid (`[-1 -1] 0 d` looped for ever in kurbo); (7) `render_size` and `render_into`: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); `render` is `render_into` on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) `may_paint_canvas`: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); `set_skip_offscreen_paths` turns this off for the regression test; (10) `/Interpolate` governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour (since (21), only inside Type 3 glyphs), planned without the Catmull-Rom filter limits (`image_resampling_size` takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) `may_paint_rect`: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (`InterpreterSettings::preview`: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under `PREVIEW_MIN_IMAGE_AREA` device pixels, while final renders always draw them; (13) reading colours (#897): `RenderSettings::color_replacement` takes a `ColorReplacement` (text colour, background, `only_black`, `line_art`) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; `render_pdf` gains a `color_replacement: Option` argument (pass `None` for true colours); the upstream-only `[dev-dependencies.log]` entry is removed from `Cargo.toml` (log is a normal dependency since the logging patch), so `cargo test -p hayro` runs this crate's unit tests; (14) a shrunk image that is interpolated (or, since (21), area-averaged) and whose axes lie along the device's (upright, mirrored or a quarter turn, as on a page under `/Rotate`) is snapped to whole device pixels: shrunk once to its rounded device rectangle (halves round up, so a tile snaps an edge where the whole page does) and drawn 1:1 from the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (`snap_minified_image`; scans looked soft next to Acrobat, and until #5530 a turned one still did, up to 200 grey levels off the same page drawn upright). An image counts as axis-aligned when the terms the snap drops move its far edges by at most a hundredth of a device pixel (`SNAP_SKEW`), not by a tolerance on the coefficients, which let a slight skew of a very wide image through; and the snapped 1:1 transform is used only when the shrink produced the snapped size, so a failed resize keeps the full transform instead of drawing the source raster at its source size; (15) a soft mask of another size than its image is filtered when shrunk, whatever its `/Interpolate` (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroke thinner than the shrink factor; (16) under a soft mask or a blend mode, axial and radial shadings are sampled instead of drawn as native gradients (#2128): vello_cpu 0.0.8 copies an opaque gradient over fully covered tiles without the mask or blend mode, so PowerPoint's faded gradient fills came out solid; (17) the renderer keeps annotation paints out of the reading-colours mapping between `begin_annotations` and `end_annotations` (#1548); (18) the page renders as one transparency group; under reading colours its background is painted inside it as the initial backdrop, so a Multiply highlight blends with the dark background instead of staying an opaque block, and otherwise the background stays outside, so a blend mode over an empty part of the page keeps its colour instead of blending with the white paper (#2970); (19) `set_paint` samples a mesh shading's texture over its triangles alone (`EncodedShadingPattern::extent`, padded rather than repeated) and skips a mesh with none on the canvas: a BricsCAD plan's 310 one-triangle `sh` each covered the whole tile, 2 s per 1024 px tile at any zoom (the watchdog gave the page up on a busy machine), now about 70 ms; (20) a tiling pattern whose step is too large for a cell pixmap, where what is painted (a fill, or a stroke with its width) can only show the first cell (PowerPoint's `/XStep 2147480100` for a picture fill drawn once), has its `/BBox` drawn at full scale with a transparent border and padded instead of tiled (`shows_one_cell`; a cell whose scale isn't finite and positive, from a `/BBox` too thin to divide by, keeps the old path: `usable_cell_scale`); shrinking the cell to fit the 3000 px limit of (3) left the picture under a pixel, so network diagrams lost their pictures (#3349); (21) an image that does not request interpolation and is drawn below its resolution, outside a Type 3 glyph, is shrunk by area averaging (`area_shrink`: each device pixel the coverage-weighted mean of the samples it covers, in 16.16 fixed point whose weights sum to exactly one, premultiplied when the image has a mask; a soft mask no larger than the image is first sampled onto the image's grid by `mask_on_image_grid` (nearest-neighbour, or bilinearly when the mask asks for `/Interpolate`, so its edge stays as smooth as an enlarged mask's), so its colour and mask are averaged together instead of the averaged colour of transparent samples showing under a mask layer at the mask's edges) to its snapped device rectangle, instead of nearest-neighbour: `/Interpolate` asks for smoothing only when an image is enlarged (ISO 32000-2 §8.9.5.3), and nearest-neighbour kept one sample in N per axis, so the text of a 300 dpi scan at 100% lost most of its strokes (sampling them again at a fractional offset dropped more). PicScale's `Box` and `Area` filters were not used: shrinking, both average a window four target pixels wide. On a 13-page 300 dpi scan every rendered sample at 96, 120 and 192 dpi is within 1 of the exact area average (it was up to 255 off, 17-23% of samples by more than 1). A snapped image whose shrink failed keeps the full transform, so it is not drawn at its source size. The shrink never enlarges: a side drawn at or above its resolution (an image shrunk across and stretched down, as a 204 × 98 dpi fax is) keeps its samples for the sampler. Each target row averages the whole source rows it covers (contiguous, cheap) and is then shrunk across with taps worked out once per image, which made the shrink about three times faster (a 300 dpi page at 96 dpi: 25 ms, nearest-neighbour 11 ms); working a row at a time (`area_shrink_rows`) keeps its work memory to a few rows. An image with a soft mask, a stencil's ink included, is premultiplied a source row at a time as it is averaged (`area_shrink_premultiplied`), never as a full-resolution RGBA copy: a 10718 × 13701 stencil shown at page size peaked at 1.16 GB with that copy and peaks at 1.03 GB without it, as with nearest-neighbour. A source with a side over `MAX_RESAMPLING_SOURCE_SIDE` (2^20) keeps (10)'s nearest-neighbour shrink, and `area_shrink` refuses it, so its accumulator row and tap table stay a few times one source row (a 2^27 × 2 image of zeros is a tiny file); (22) knockout groups: while one is innermost, fills, strokes, glyph outlines and shadings are drawn with `Compose::Copy` (weighted by coverage), so each element replaces what earlier elements of the group drew instead of compositing over it; an image or nested group, drawn into a layer of its own, first has what is under its shape cleared (`knockout_image` for an unmasked image's rectangle, only when `draw_image` will draw it, decoded at the size it is drawn; a knockout erase pass that draws a nested group's content with every paint clearing instead of painting, a shading fill's clear clipped to the shading's `/BBox` as its paint is) and then composites over it (#2971); (23) backport of hayro 0.8: shadings that `as_svg_gradient` can represent are painted as a `vello_cpu` linear or two-point radial gradient instead of a texture sampled pixel by pixel through the shading function (InDesign fades are luminosity soft masks of one full-page axial `sh`: 40 on a report page took 15 s at 144 dpi, now under 1 s); without upstream's half-pixel shift, which suits vello_cpu 0.3 but moves 0.0.8's gradients off pixel centres; (25) a tiling cell is drawn at the scale that makes one `/XStep` (`/YStep`) exactly the cell pixmap's whole number of pixels (`tiling_cell_pixels`, at least 1 and at most the cap), so the repeated pixmap puts cells at whole multiples of the step at any device scale; drawing at the unrounded scale repeated the cell up to half a pixel off at every step, and the error grew with distance from the pattern's origin (ISO 32000-2 8.7.3.1); (28) an image of page-scan size (`SMOOTHED_IMAGE_SIDE`: 512 samples or more on its shorter side) that does not request interpolation is drawn with `ImageQuality::Medium` (bilinear, as interpolated images are) when enlarged by more than 1% on either side (`SMOOTHED_IMAGE_SCALE`: a scan drawn at its resolution can come out a few ulps above 1, and blurred by half a pixel at a half-pixel offset), also when the other side is shrunk (the area shrink leaves that side at 1:1), outside a Type 3 glyph: scans rarely set `/Interpolate`, and drawn nearest-neighbour every sample of a zoomed scan became a block, so its text had staircase edges; icons, QR codes and pixel art keep their hard pixels. Bicubic (`High`) looked the same on a zoomed 300 dpi scan and drew four 1024 px tiles at 400% in 267 ms against 50 ms | (1) `raster::tests::tiles_match_full_render`; (2) `raster::tests::absurd_image_dimensions_are_skipped`; (3) `raster::tests::tiling_cells_stay_small_whatever_the_step`; (4) `raster::tests::absurd_predictor_columns_and_line_widths_render`; (5) `raster::tests::dash_patterns_that_cut_a_stroke_into_billions_of_pieces_render`; (6) `raster::tests::invalid_dash_patterns_terminate_and_are_drawn_solid`; (7) `raster::tests::pixels_are_a_plain_renders_bytes_in_a_buffer_a_gui_can_take_over`; (8) `raster::tests::image_resampling_rejects_anisotropic_target_growth`, `raster::tests::image_resampling_bounds_intermediate_before_planning`, `raster::tests::image_resampling_rejects_invalid_sources_and_scales`, `raster::tests::image_resampling_keeps_valid_sizes_and_exact_limits`, `raster::tests::resampling_fallback_preserves_original_geometry_and_pixels`, `raster::tests::image_resampling_wide_sources_shrink_before_backend_side_limits`, `raster::tests::image_resampling_wide_transparent_sources_still_shrink`, `raster::tests::image_resampling_padded_backend_dimensions_stay_checked`, `raster::tests::image_resampling_mismatched_alpha_mask_keeps_pixels_and_placement`; (9) `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel`; (10) `raster::tests::non_interpolated_images_are_not_smoothed_when_minified`, `raster::tests::non_interpolated_gray_images_are_not_smoothed_when_minified`, `raster::tests::non_integer_minification_averages_the_columns_each_pixel_covers`, `raster::tests::non_integer_minification_averages_colour_and_mask_together`, `raster::tests::non_interpolated_alpha_images_keep_their_colours_and_mask_when_minified`, `raster::tests::non_interpolated_type3_glyph_images_are_not_smoothed_when_minified`, `raster::tests::image_resampling_nearest_accepts_wide_sources_the_filter_refuses`, `raster::tests::wide_non_interpolated_sources_draw_after_shrinking`; (11) `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel`; (12) `raster::tests::preview_renders_skip_tiny_images_but_finals_draw_everything`; (13) `color_replacement_tests::{black_maps_to_text_colour, white_maps_to_background, midgrey_interpolates, alpha_copied_verbatim, only_black_threshold, line_art_false_skips_paths_but_not_text}`, `color_replacement_render_tests::{color_replacement_maps_text_and_line_art, color_replacement_maps_stencil_mask_ink}`; (14) `raster::tests::minified_image_edges_snap_to_whole_pixels`, `raster::tests::a_tile_snaps_half_pixel_image_edges_as_the_whole_page_does` (`cargo test -p hayro`), `raster::tests::a_turned_or_mirrored_minified_image_snaps_to_whole_pixels`, `rotated_image_snap::a_shrunk_image_on_a_turned_page_is_the_upright_render_turned` (`cargo test -p pdfcraft-render --test it`), `renderer::snap_tests::a_skew_that_moves_a_far_edge_is_not_snapped` (`cargo test -p hayro`); (15) `raster::tests::shrunk_soft_mask_larger_than_its_image_keeps_thin_strokes`; (16) `raster::tests::soft_mask_applies_to_a_shading_pattern_fill`; (17), (18) `raster::tests::reading_colors_keep_highlight_annotations_readable`; (19) `mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing}` (`cargo test -p hayro`); (20) `raster::tests::a_tiling_pattern_drawn_once_shows_its_cell`, `raster::tests::a_large_step_still_tiles_where_the_next_cell_shows`, `raster::tests::a_tiling_cell_too_thin_to_scale_keeps_the_old_path`; (21) `raster::tests::non_interpolated_minification_keeps_thin_strokes`, `raster::tests::non_integer_minification_averages_the_columns_each_pixel_covers`, `raster::tests::non_integer_minification_averages_colour_and_mask_together`, `raster::tests::non_interpolated_minification_does_not_bleed_transparent_colour`, `raster::tests::a_smaller_soft_mask_does_not_bleed_transparent_colour_when_minified`, `raster::tests::an_interpolated_smaller_soft_mask_stays_smooth_when_its_image_is_minified`, `raster::tests::a_shrunk_stencil_mask_keeps_thin_strokes`, `raster::tests::minified_image_edges_snap_to_whole_pixels`, `raster::tests::an_image_shrunk_one_way_and_stretched_the_other_is_still_averaged`, `raster::tests::wide_non_interpolated_sources_draw_after_shrinking`, `renderer::area_shrink_tests` (`cargo test -p hayro`); (22) `raster::tests::knockout_groups_replace_what_their_elements_cover`; (23) `raster::tests::axial_and_radial_shadings_draw_as_native_gradients`; (25) `tiling_cell_placement::{far_cells_of_a_fill_sit_at_whole_steps, far_cells_sit_at_whole_steps_under_a_turned_scaled_matrix, far_cells_of_a_stroke_sit_at_whole_steps, a_cell_pixmap_spans_exactly_one_step}` (`cargo test -p pdfcraft-render --test it`); (28) `raster::tests::large_non_interpolated_images_are_smoothed_when_enlarged`, `raster::tests::small_non_interpolated_images_keep_hard_pixels_when_enlarged`, `raster::tests::a_scan_shrunk_one_way_is_smoothed_the_way_it_is_stretched`, `raster::tests::a_scan_drawn_a_rounding_error_above_its_resolution_stays_sharp` | | hayro-jbig2 | 0.3.0 | Apache-2.0 OR MIT | (1) at most 2^28 pixels per bitmap and page bitmap, and 2^25 cells per halftone grid and grayscale image (each side could be 65535: 4.3 billion pixels decoded one by one, and a 4-byte value per cell); (2) at most 2^26 symbol pixels per symbol dictionary (a fuzzed dictionary decoded for minutes); (3) a symbol dictionary's height classes are capped at one per new symbol and its export runs at `2 × symbols + 1`, each plus `EXTRA_EMPTY_STEPS` (1024): an empty height class or a zero-length export run doesn't advance, and past the end of its data the arithmetic decoder can decode them for ever (one changed byte hung the renderer); (4) a segment header's long-form referred-to segment count is capped at `u16::MAX` and a larger count is refused before its retention or referred-to bytes are read: the 29-bit count went straight to `Vec::with_capacity`, a reservation of about 2 GiB whose failure aborts the process (from storytold/craft-libs#176); (5) what a page's segments hold for later ones (intermediate regions, pattern and symbol dictionaries with their headers and retained contexts, and Huffman tables) is at most 64 MiB, two square bitmaps of 16,384 a side (`MAX_RETAINED_WORDS`), and a dictionary's new and exported symbols are counted against it as they are made: each was bounded on its own, but a few dozen bytes an intermediate region held 33 MB each, and 8 MiB symbols exported again and again killed the process, so a few kilobytes could exhaust memory; (6) a segment may refer to at most 2^20 symbols, checked before any are collected: references may repeat, and 65,535 of them to one dictionary, in a 78 KB stream, asked for 17 GB; (7) a pattern dictionary keeps at most 16 MiB of shifted copies, which only speed up drawing, and none that wouldn't fit what the page has left (`MAX_SHIFTED_WORDS`); (8) a custom Huffman table has at most 2^16 lines, and a stream with its globals at most 2^20 segments and 2^22 references, checked as they are read: both grew before anything counted them, and PDFs of a kilobyte or so reached 5 GB and more behind nested Flate; (9) a page past the bounds of (1) is refused when the stream is opened: a caller sizes its output from the header before the page bitmap is made, and an Indexed image in a 743-byte PDF asked for 137 GB, which aborts; (10) a stream may ask for at most 2^30 units of decoding work, four of the largest pages, about 6 s to 17 s (`MAX_WORK`): decoding a bitmap costs a pixel each and 4 a row, copying one a word each and 1 a row, and each integer an item reads costs 38, what one takes at worst, so one with no pixels pays too, before it is decoded: each bitmap was bounded, but nothing bounded how many there were, and a 795-byte PDF took 265 s; (11) `Image::decode_within` spends from what a run of images may still do, starting with its page's pixels, and `charge_reading` charges reading a stream and its globals before anything is parsed, so an operation over a document can bound what `MAX_WORK` bounds for one image, and images that fail or decode nothing add up too: a file of thousands of images made to reach that limit took as many times as long; (12) adding to a location, a reference offset, a symbol's corner or a Huffman range line near the end of an `i32`, and negating a decoded integer of 2^31, overflowed: a panic wherever overflow is checked. The refinement context, a text region's corners, the integer decoder and the Huffman table wrap, as a release build did, and `Bitmap::combine` saturates, which comes to the same, as a location that overflows is past the end of any page. What is drawn doesn't change; (13) a bitmap filled with a default pixel of 1 (a page, or an intermediate text or halftone region) had 1s past its width in the last word of each row, and a refinement, which reads its reference a word at a time, read them as pixels where the standard has 0 (6.3.5.2): refining such a reference not a multiple of 32 wide decoded to other pixels than were coded, release builds too. `Bitmap::clear_padding` clears those bits when a bitmap is filled with 1s, as cutting symbols and patterns from a collective bitmap already did. Ghostscript reads the test's page and text streams as they decode now; (14) a region's location is unsigned (5.4.4), so one past `i32::MAX` is off the page, where the decoder took it for a signed one: a region there was drawn back onto the page from the left or the top (at 0xFFFFFFFC, four pixels left of it), and a refinement there, of the page or of a region at the same place (7.4.7.5), was refused as an overflow, and the page with it. `Bitmap::combine_region` keeps such a location past the page. A refinement of another region refines it where it is, as Table 35 has it (GRREFERENCEDX and GRREFERENCEDY 0), whatever the two locations say; a refinement of the page, see (16); (15) a refinement whose reference is offset right of it (GRREFERENCEDX above 0, which a refined text instance has from a wider bitmap, Table 12, and a refined symbol from an RDX above 0, Table 18) read the reference's first 32 columns as 0: `RefinementContextGatherer::maybe_reload_buffers` emptied its buffers left of the reference and didn't load them again at its first column. Such a refinement decoded to other pixels than were coded, or failed and lost the page. Left of the reference the buffers now keep the words of its first column each row starts with; with an offset of 0 or less nothing changes; (16) a refinement of the page refines "the current contents of the page buffer, restricted to the area of the page buffer specified by this segment's region segment information field" (7.4.7.4), at an offset of 0 (Table 35): `Bitmap::area` copies that area a word at a time, 0 off the page and past its width whatever the padding of its rows holds, and every refinement region reads its reference at offset 0. The page around the area was read too, so a refinement of the page decoded to other pixels than were coded, and one wider or taller than the page was refused. And `RefinementContextGatherer::start_row` loads the reference's words a column left of where a row starts in it, as the reload would, not from column 0: with the reference offset 32 or more left of the refinement (GRREFERENCEDX below −31, as a refined text instance has from a narrower bitmap or an RDX below −31, and a refined symbol from such an RDX), each row's first context read 0 for the reference column it starts at, and from 33 left for the one before it too, and the refinement decoded to other pixels or failed | (1) `raster::tests::jbig2_regions_of_billions_of_pixels_are_refused`, `raster::tests::jbig2_halftone_grids_past_their_bound_are_refused`; (2) the nightly `cargo xtask fuzz` finding, which now finishes in about a second; (3) `raster::tests::jbig2_symbol_dictionaries_that_never_advance_terminate`; (4) `segment::tests::test_referred_to_count_is_capped` (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`); (5) `raster::tests::jbig2_regions_held_past_a_pages_bound_are_refused`, `jbig2_symbols_without_pixels_count_towards_a_pages_bound`, `jbig2_retained_contexts_count_towards_a_pages_bound`, `jbig2_huffman_tables_count_towards_a_pages_bound`, `jbig2_new_symbols_count_towards_a_pages_bound`, `jbig2_new_symbols_are_held_to_what_a_page_has_left`, `jbig2_pattern_dictionaries_count_towards_a_pages_bound`, and in the crate (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`) `decode::symbol::tests::exported_symbols_are_counted_before_they_are_copied`, `tests::the_room_left_shrinks_as_segments_are_stored`; (6) `raster::tests::jbig2_segments_referring_to_too_many_symbols_are_refused`; (7) `raster::tests::jbig2_pattern_dictionaries_too_large_to_shift_still_draw`, `jbig2_shifted_patterns_never_refuse_an_image`; (8) `raster::tests::jbig2_huffman_tables_of_too_many_lines_are_refused`, `jbig2_streams_of_too_many_segments_are_refused`, `jbig2_streams_of_too_many_references_are_refused`, and in the crate `tests::the_segments_of_a_stream_and_its_globals_count_together`; (9) `raster::tests::jbig2_pages_past_the_bounds_are_refused_before_anything_is_sized_from_them`, `tests::a_page_past_the_bounds_is_refused_when_opened`; (10) `raster::tests::jbig2_decoding_work_is_counted_item_by_item`; (11) `jbig2::decode::tests::a_run_of_images_spends_its_allowance_down_to_the_step` (pdfcraft-filters); (12) `raster::tests::jbig2_placements_at_the_end_of_an_i32_draw_nothing`; (13) `jbig2::decode::tests::a_refinement_reads_nothing_past_its_references_width` (pdfcraft-filters), and in the crate `tests::a_bitmap_of_1s_has_nothing_past_its_width`; (14) `jbig2::decode::tests::a_region_located_past_i32_max_is_off_the_page` (pdfcraft-filters), and in the crate `tests::a_region_located_past_i32_max_is_off_the_page` and `tests::a_refinement_of_a_region_decodes_the_same_wherever_they_are`; (15) `jbig2::decode::tests::a_refinement_reads_a_reference_right_of_it_from_its_first_column` (pdfcraft-filters), and in the crate `tests::a_refinement_reads_a_reference_right_of_it_from_its_first_column`; (16) `jbig2::decode::tests::a_refinement_refines_the_page_in_its_area_and_any_reference_where_it_is` (pdfcraft-filters), and in the crate `tests::a_refinement_refines_the_page_in_its_area_and_any_reference_where_it_is` and `tests::a_page_area_is_the_page_inside_it_and_0_outside` | | lopdf | 0.45.0 | MIT | (1) PNG predictor rows longer than the data are refused before their buffers are allocated (a fuzzed `/Columns 4294967295` allocated two 4 GiB rows), and empty frames return without row buffers; (2) predictor row widths use saturating arithmetic or reject overflow; (3) sub-byte TIFF Predictor 2 scratch is bounded by the samples present in the input instead of the declared `/Colors`, aligned with [lopdf#589](https://github.com/J-F-Liu/lopdf/pull/589); (4) the object lookup keeps its cycle history through compressed objects, requires an ordinary in-file container, and caps the chain of indirect lookups at 16 (a chain of new objects, each a `/Length` compressed in a container whose own `/Length` names the next, or a stream whose `/Length` names the next stream, overflowed the stack: a dev build's 2 MiB thread within 800 hops); (5) object-stream members are parsed within their own extent (up to the next distinct offset), an entry whose offset repeats an earlier entry's is skipped (so a repeated object number can resolve to its earlier entry), and offsets are processed in index order; (6) `find_latest_trailer`: a rebuild takes a `/Type /XRef` stream dictionary as the trailer when the file has no `trailer` keyword (the latest in the file wins, as a later `trailer` would), so a concatenated or truncated file keeps its /Info and /ID instead of failing to load; a candidate whose /Root is the catalog beats one whose /Root is another dictionary, and an in-file /Root must match its generation (#2365) | (1) `inspect::tests::png_predictor_rows_longer_than_the_data_are_refused_before_allocating`, `inspect::tests::empty_png_predictor_frames_do_not_allocate_rows`; (2) `inspect::tests::tiff_subbyte_predictor_row_width_overflow_is_refused`; (3) `inspect::tests::tiff_subbyte_predictor_scratch_is_bounded_by_available_samples`, `inspect::tests::tiff_subbyte_predictors_keep_components_rows_and_padding`; (4) `inspect::tests::compressed_object_self_container_does_not_recurse`, `inspect::tests::compressed_object_mutual_containers_do_not_recurse`, `inspect::tests::object_stream_length_cycle_does_not_recurse`, `inspect::tests::compressed_object_container_chain_does_not_recurse`, `inspect::tests::compressed_length_chain_does_not_recurse`, `inspect::tests::plain_stream_length_chain_does_not_recurse`, `inspect::tests::compressed_objects_and_indirect_stream_length_still_load`, `inspect::tests::compressed_length_in_a_container_with_an_indirect_length_still_loads`; (5) `inspect::tests::object_stream_member_extents_bound_owned_strings`, `inspect::tests::object_stream_member_extents_preserve_index_order_and_whitespace`; (6) `inspect::tests::rebuilt_file_keeps_the_info_of_its_cross_reference_stream`, `inspect::tests::rebuilt_file_prefers_a_trailer_whose_root_is_the_catalog` | -| hayro-syntax | 0.7.2 | Apache-2.0 OR MIT | (1) Page-tree cycle guard in `resolve_pages` (visited set + depth cap): a `/Kids` loop among compressed objects overflowed the stack and aborted the process; (2) `ccitt_size_ok`: CCITT images wider than 2^20 columns or over 2^28 pixels are not decoded (a fuzzed `/Columns 4294967295` allocated 4 GiB); (3) Flate/LZW predictor row widths are saturating `usize`, a zero width is refused and a row longer than the data is not allocated (a fuzzed `/Columns 9223372036854775807` wrapped to a 2^61-byte allocation and aborted); (4) the inline-image end-of-data check (`EI`) doesn't nest inside its own look-ahead and examines at most 256 candidates and steps (50 images with "EI" in their data ran for over ten minutes); an EI followed by a delimiter (EI%..., EI[...) ends the image like one followed by white space; only a regular character after EI skips the candidate; (5) a stream whose `/Length` is wrong is recovered only when `stream` is the next token — scanning ahead to a later object's `stream` made an indirect appearance-state dictionary (Quartz checkboxes: `/AP /N` → `<< /Yes … /Off … >>`) render as a non-form, so the widget was skipped; (6) JPX rescaling uses an exact integer sample range for 1–32-bit components (32 shifted past u32, and floating-point maxima at 25–31 bits wrapped to zero), and refuses zero-sized rows; (7) `decode_limit`: Flate (through flate2 and the fallback inflater), LZW, RunLength and ASCII85 never write past `MAX_DECODED_STREAM` (256 MiB), or, for the last filter of an image of known size, past twice its pixel bytes (a predictor tag byte per byte), at most 1 GiB; `Stream::decoded_within` for budgeted callers; a page's `/Contents` array shares one `MAX_DECODED_STREAM` (a 2 MB page whose content stream or 8 × 8 image inflated to 2 GB took the renderer past 3 GB in under a second; naming the stream twenty times in `/Contents`, 5 GB); (8) high-precision numbers (jsPDF MediaBox coordinates with over 19 digits) parse without overflowing the u64 mantissa: fractional digits past it are dropped, integer digits past it scale the value (a real, kept finite); (9) large dictionary indexes are built on first use and shared across repeated reads within each PDF (#307); the validated-byte cache is bounded, reader contexts remain separate, repairs and inline/content dictionaries bypass sharing, and small dictionaries stay eager; (10) decoded object-stream slots publish once under contention, cold loading rejects recursive dependencies, and xref repair serializes without holding recursive read guards; this permits one parser per native render pool with private worker caches; (11) `sync.rs`: a poisoned lock is recovered instead of unwrapped (a worker panic no longer cascades into every reader of the same PDF), and xref repair takes a blocking write lock instead of asserting a non-blocking one; (12) `SegmentList` (`util.rs`) returns `None` for an out-of-range index instead of panicking; (13) a page's `/UserUnit` scales `render_dimensions` and `initial_transform`, so the page and everything on it are sized in points, as Acrobat shows them; it is read as Acrobat reads it (the page's own entry, not inherited; a number from 1 up, at most 75,000; anything else counts as 1), and a page without it keeps exactly the transform it had; (14) revision 2–4 passwords are tried as given and then in PDFDocEncoding (`legacy_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Convert to PDFDocEncoding`): passwords arrive as UTF-8 text, so a password with characters outside ASCII (`Çok güzel`, `ılık€`) written in PDFDocEncoding, as ISO 32000-2 §7.6.4.3.2 asks and qpdf and Acrobat write it, was always wrong; (15) page content is read a piece at a time (`page::ContentReader`, `Page::content_reader`): plain streams and Flate streams without a predictor are decoded as they are read, with a budget of 64 times their compressed size and never less than `MAX_DECODED_STREAM` (a stream that expands past it is cut and reported); other streams are decoded whole, as before, and share `MAX_DECODED_STREAM`. The content tokenizer reads a window at a time (`UntypedIter::new_window`, `TypedIter::needs_more`, `TypedIter::committed`): an instruction that runs past the window is not decided until more data follows, which `Reader` records by flagging reads and `at_end` past the end of the data, including through the clones the parser makes. The Flate decoder keeps its own 32 KiB history, so it pauses at any output size. `ContentReader::new` reuses raw/decrypted stream bytes for budget accounting and piece creation, and `decoded_raw_within` applies the existing Whole-stream filters and limits to those bytes without decrypting twice; (16) revision 5–6 (AES-256) passwords are tried in their SASLprep form (RFC 4013, then UTF-8, at most 127 bytes, as ISO 32000-2 §7.6.4.3.3 Algorithm 2.A asks and pdfcraft-crypt writes them) and then, if that differs, as given (`aes256_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Do the above.`; `stringprep`, already in the tree through pdfcraft-crypt, is an optional dependency enabled by `std`, and without `std` passwords are used as given): the renderer used the password as typed, so one that SASLprep changes (a decomposed `café`, a no-break space, the fullwidth `ab` a Japanese IME types in full-width mode) was wrong, although pdfcraft-cos accepted it: protecting a document with such an open password was rolled back ("the password is incorrect"), and a file with such a password did not open with it.; (17) CCITT `/Rows 0` means the height is not predetermined (ISO 32000-2 Table 11), not zero rows: scanned Group 4 pages written that way rendered blank; (18) The streaming Flate decoder (`lzw_flate::flate::fallback`) copies a back-reference in runs of up to its distance (`extend_from_within`), not byte by byte. The copy reads only bytes already in the output, and a copy whose source does not exist still writes nothing, so the bytes are those a byte-by-byte copy writes. Inflating the A1 sample's content stream (278.6 MB) takes about 14% less time (0.66 s to 0.56 s); the output is identical; (19) CCITT `/Rows` below the image's `/Height` decodes to the height: a plotter driver wrote `/Rows` equal to `/Columns`, so only the first rows were decoded and the rest drawn black (#3308) (20) `stream` followed by spaces or tabs and then the line break (`stream \r\n`, as Microsoft Reporting Services writes it) starts the data after the line break; the content stream was taken to start at the space and the page rendered blank.; (21) a JP2 colour specification naming an enumerated colour space the decoder refuses (CIEJab in a veraPDF fixture, YCbCr, PhotoYCC, YPbPr, bilevel…) no longer fails the image: it is decoded again from a copy whose `colr` names the space the decoder assumes without one, greyscale, sRGB or CMYK by component count, as MuPDF and Poppler draw it (`without_refused_enumerated_colour`; the page drew blank, #3306); (22) an object stream's index of `objnum offset` pairs is parsed once per stream and kept beside its decoded data (`ObjectStreamIndexes`, `XRef::with_object_stream_indexes`), and an offset past `usize` is malformed instead of wrapping: `XRef::get_with` parsed the whole index again for every object it fetched from the stream, so reading all of a stream of N objects took N² steps (a Firefox page whose 65,542 objects share one stream took 37 s to draw and 40 s to extract its text); (23) the inline-image end-of-data look-ahead no longer rejects a candidate `EI` because a byte above 0x7F lies between it and the next one: when the candidate is the end, those bytes are the content stream that follows the image, and its strings carry such bytes (WinAnsi `(Grüße) Tj`, the two-byte codes of any Type 0 font), so the operator check that follows decides alone, bounded as in (4) (a scanned page whose OCR text held an accented word and a later capital "EI" — EINGANG, BEI, KLEIN — drew and extracted nothing after the image: 40 of 275 pages across five real documents); (24) `xref::prev_offset`: a trailer `/Prev` written as a reference `N G R` is read as the offset that object holds when the section's own entries place it (iText's `indirectPrev.pdf`; each section's entries are read before its `Prev` is followed), and otherwise as offset N (tiff2pdf; JHOVE's PDF-HUL-157 files, #2366), as `pdfcraft-cos` and other readers take it, so the earlier sections' pages are seen instead of one; (25) direct arrays and dictionaries share a reader-local nesting cap of 128, counted while skipping and while parsing a dictionary (its lenient repair reads, rather than skips, an object found where a key belongs, so `<< << …` recursed through that read); deeper objects return `None` instead of overflowing the stack (including unused trailer values, trailer entries, content operands and marked-content properties); a chain of 65 direct stitching functions now stops at this cap, before the 64-function guard; (26) an xref stream subsection must be backed by its decoded data: all-zero `/W` widths (an entry reads no bytes) and a declared count longer than the data are refused before any entry is inserted, where a tiny stream with a large `/Size` or `/Index` count inserted that many entries (from storytold/craft-libs#239); a subsection running past object number `i32::MAX` is refused instead of overflowing or turning negative; (27) an xref stream's type field is read big-endian at its full `/W[0]` width, where only its first byte was read and wider fields shifted every later entry, and `/W[0]` over 8 bytes is refused like `/W[1]` (from storytold/craft-libs#241); (28) a content instruction keeps up to 256 operands (`MAX_OPERANDS`) instead of 10, and past that drops the oldest instead of ending the content stream: the eleventh operand of a twelve-ink DeviceN `scn` (ISO 32000-2 §8.6.6.5 allows any number of components) blanked the rest of the page (#4260); (29) a `/JBIG2Globals` stream that is itself JBIG2 isn't decoded: globals are raw segments, and a chain of 16,000 such streams overflowed the stack, which aborts the process; (30) a stream's own crypt filter: `/Crypt` in its chain read as the filter it names, `/Identity` without one, instead of the document's filter and then a filter that can't be decoded, which left such a stream unreadable in any spelling (ISO 32000-2 §7.4.10); (31) `set_page_jbig2_work`: the JBIG2 images decoded on a thread spend one allowance, set per page, 2^30 steps (`MAX_PAGE_JBIG2_WORK`), so a page of many images made to reach their own limit stops at one, and what they cost to read counts, whether they decode or not; unbounded until a page sets it; (32) a group of array items read together (a tuple: an image's `/Decode`, `/W`, `/W2`, and a function's `/Domain`, `/Range`, `/Encode` and `/Decode`) can start with a reference: it was taken for the whole group and lost, which ended the array, so an image with `/Decode [7 0 R 0]` didn't draw and CID widths fell back to `/DW`; (33) an object that holds only a reference stands for what that names, stored on its own or in an object stream, up to 32 references: read as a number it was the reference's own object number, so a `/Decode` or `/W` entry naming such an object took a wrong value, and as anything else it was lost. Past 32, as any cycle goes, it reads as a missing object. An object stream isn't reached that way, as its checks are of the object named; reading its index counts against the 64 nested loads and reads its dictionary alone, as a chain of them, each with a member of the next for its `/N`, `/First` or `/Length`, ran out of stack; and a comment inside a reference counts as white space; (34) a table rebuilt by scanning the file (`fallback_xref_map`) finds an object by its number when its header names another generation, including one no generation may be (`3 18446744073709551616 obj` in `poppler-85140-0.pdf`, #5580): the header is read leniently only in that scan (`read_object_header`), the object is also filed under generation 0 unless a real `N 0 obj` holds that key (a later header for the number moves the stand-in), `XRef::get_with` matches by object number only in a repaired table, and `XRef::objects` yields each location once; (35) PNG predictors (`lzw_flate::unfilter_png`) are undone a byte row at a time on slices, as the PNG filters are defined, instead of through the bit-level readers TIFF predictors need: a 993 × 936 RGB image took about 63 ms to unfilter whatever its row filters, and now takes about 3 ms (Up) to 16 ms (Paeth) in a release build (#5430); a row ending part-way through a pixel (three 4-bit samples, an odd width) keeps its last byte, where the bit-level readers dropped it and each later row started a byte early, shearing the picture | (1) `cos/tests/writer.rs::page_tree_loops_in_object_streams_do_not_crash_readers`; (2) `raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding`; (3) `raster::tests::absurd_predictor_columns_and_line_widths_render`; (4) `raster::tests::inline_images_with_ei_in_their_data_parse_in_linear_time`, `content::tests::inline_image_ei_may_be_followed_by_a_delimiter`; (5) `raster::tests::indirect_appearance_state_dict_is_not_another_stream`; (6) `raster::tests::jpx_stream_bit_depth_boundaries_do_not_panic`, `raster::tests::jpx_32_bit_image_does_not_prevent_other_page_content_from_rendering`, `raster::tests::jpx_gray_rgb_and_alpha_keep_eight_bit_samples`, `raster::tests::jpx_streams_reject_out_of_range_bit_depths`; (7) `raster::tests::decompression_bombs_stop_at_the_stream_limit`, `raster::tests::page_contents_and_nested_forms_share_one_budget`; (8) `inspect::tests::jspdf_high_precision_mediabox_does_not_overflow_dimensions`, `object::number::tests::high_precision_float_does_not_overflow_mantissa` (hayro-syntax); (9) `object::dict::lazy_index_tests`, `raster::tests::large_resource_indexes_preserve_pixels_and_text`; (10) `data::tests`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`, `raster::tests::shared_parser_is_lazy_and_initializes_once_under_contention`; (11) `data::tests::panicking_initializer_wakes_waiters_and_prevents_retries`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`; (12) `util::tests::segment_list_rejects_out_of_range_indices`; (13) `raster::tests::user_unit_scales_the_page_and_everything_on_it`; (14) `crypto::pdfcraft_password_tests`, `pdfcraft-engine` `tests::legacy_passwords_with_turkish_characters_open`; (15) `content::tests::windows_keep_every_instruction_at_every_split`, `content::tests::windows_keep_every_instruction_in_pieces_of_any_size`, `content::tests::windows_match_the_slice_on_random_content`, `content::tests::windows_drop_blanks_and_keep_a_comment_that_may_go_on`, `page::content_reader_tests::content_past_the_cap_is_read_to_the_end`, `page::content_reader_tests::a_stream_that_expands_past_the_budget_is_cut_there`, `page::content_reader_tests::a_small_expansion_under_the_cap_is_not_cut`, `raster::tests::page_content_past_the_old_cap_renders`, `raster::tests::content_reader_returns_the_same_bytes_in_any_pieces`, `raster::tests::content_bombs_stop_at_the_budget_and_say_so`, `page::content_reader_tests::plain_and_whole_pieces_keep_their_bytes`, `page::content_reader_tests::repeated_content_reference_counts_compressed_bytes_once`, `page::content_reader_tests::whole_cap_still_counts_later_unique_streams`; (16) `crypto::pdfcraft_password_tests::aes256_passwords_are_tried_in_their_saslprep_form_then_as_given`, `crypto::pdfcraft_password_tests::aes256_passwords_open_as_typed_and_in_their_saslprep_form`, `pdfcraft-engine` `tests::aes256_open_passwords_that_saslprep_changes_protect_and_reopen_as_typed`, `tests::aes256_files_whose_passwords_saslprep_changes_open_as_typed`; (17) `raster::tests::group4_fax_with_rows_zero_decodes_to_the_image_height`; (18) `filter::lzw_flate::tests::back_reference_longer_than_its_distance_repeats_its_source`, `back_reference_cut_by_the_limit_resumes_in_the_next_call` and `back_reference_reads_bytes_written_by_an_earlier_call`; each fails when the copy returns the wrong remaining count; (19) `raster::tests::group4_fax_with_rows_below_its_height_decodes_to_the_image_height`; (20) `raster::tests::space_before_the_line_break_after_stream_still_renders`; (21) `raster::tests::a_jp2_image_in_a_refused_enumerated_colour_space_is_drawn`, `filter::jpx::tests::refused_enumerated_colour_spaces_become_the_component_counts_default`, `filter::jpx::tests::what_the_decoder_handles_or_cant_be_read_is_left_alone` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (22) `raster::tests::objects_in_one_large_object_stream_are_read_in_linear_time`; (23) `text::tests::text_after_an_inline_image_survives_a_non_ascii_byte_before_a_later_ei`, `content::tests::an_inline_image_ends_at_its_own_ei_when_later_text_holds_a_non_ascii_byte` (hayro-syntax); (24) `xref::tests::prev_as_reference_reaches_the_earlier_section`, `xref::tests::prev_as_reference_to_an_object_holding_the_offset_reads_that_offset`, `pdfcraft-cos` `document::tests::a_prev_written_as_a_reference_still_reaches_the_earlier_sections`, `document::tests::a_prev_reference_to_an_object_holding_the_offset_reads_that_offset`; (25) `raster::tests::deeply_nested_direct_objects_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_content_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_the_trailer_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_as_marked_content_properties_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_are_read_up_to_the_nesting_cap`, `raster::tests::nesting_at_the_cap_fits_a_512_kib_stack` (native only), `raster::tests::function_limits_depth_guard_refuses_a_chain_of_references`; (26) `xref::tests::xref_stream_entries_need_backing_bytes`; (27) `xref::tests::xref_stream_type_field_uses_the_full_w0_width`, `xref::tests::xref_stream_loader_rejects_oversized_w0` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (28) `content::ops::tests::scn_with_twelve_components_keeps_them_all`, `content::ops::tests::operands_past_the_cap_keep_the_latest_in_order` (hayro-syntax), `raster::tests::devicen_inks::a_twelve_ink_devicen_colour_paints`, `raster::tests::devicen_inks::a_four_ink_devicen_colour_paints`, `raster::tests::devicen_inks::an_operator_with_hundreds_of_operands_does_not_end_the_content`; (29) `raster::tests::jbig2_globals_that_are_jbig2_themselves_are_not_decoded`, `raster::tests::jbig2_globals_are_read`, `raster::tests::the_jbig2_images_of_a_page_share_one_allowance`, `raster::tests::an_image_that_fails_spends_the_pages_allowance_too`; (30) `raster::tests::a_stream_with_a_crypt_filter_of_its_own_is_drawn`; (32) `raster::tests::a_decode_array_with_references_in_it_is_read`, `raster::tests::a_functions_arrays_can_start_with_a_reference` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret); (33) `raster::tests::a_reference_to_a_reference_is_what_it_names`, `raster::tests::a_decode_array_with_references_in_it_is_read` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret), `xref::tests::an_object_stream_isnt_reached_through_a_reference`, `xref::tests::a_chain_of_object_streams_ends_at_the_limit_of_nested_loads`, `xref::tests::a_chain_of_object_streams_takes_no_more_work_for_two_ways_on`, `xref::tests::a_reference_held_in_either_storage_is_followed`, `xref::tests::a_page_a_reference_is_held_to_is_found_once` (hayro-syntax); (34) `raster::tests::a_rebuilt_table_finds_an_object_whose_header_names_another_generation`, `raster::tests::a_rebuilt_tables_generation_0_stand_in_is_not_a_second_object` (pdfcraft-render), `xref::tests::an_object_stream_and_its_first_member_are_both_enumerated` (hayro-syntax); (35) `filter::lzw_flate::pdfcraft_png_predictor_tests` (hayro-syntax: every filter type at 1–16 bits and 1–4 colours against the PNG specification's byte-by-byte definition, a row ending part-way through a pixel, Average and Paeth over 6-byte pixels), `raster::tests::png_predicted_images::a_four_bit_rgb_image_of_odd_width_with_up_predicted_rows_is_not_sheared` | +| hayro-syntax | 0.7.2 | Apache-2.0 OR MIT | (1) Page-tree cycle guard in `resolve_pages` (visited set + depth cap): a `/Kids` loop among compressed objects overflowed the stack and aborted the process; (2) `ccitt_size_ok`: CCITT images wider than 2^20 columns or over 2^28 pixels are not decoded (a fuzzed `/Columns 4294967295` allocated 4 GiB); (3) Flate/LZW predictor row widths are saturating `usize`, a zero width is refused and a row longer than the data is not allocated (a fuzzed `/Columns 9223372036854775807` wrapped to a 2^61-byte allocation and aborted); (4) the inline-image end-of-data check (`EI`) doesn't nest inside its own look-ahead and examines at most 256 candidates and steps (50 images with "EI" in their data ran for over ten minutes); an EI followed by a delimiter (EI%..., EI[...) ends the image like one followed by white space; only a regular character after EI skips the candidate; (5) a stream whose `/Length` is wrong is recovered only when `stream` is the next token — scanning ahead to a later object's `stream` made an indirect appearance-state dictionary (Quartz checkboxes: `/AP /N` → `<< /Yes … /Off … >>`) render as a non-form, so the widget was skipped; (6) JPX rescaling uses an exact integer sample range for 1–32-bit components (32 shifted past u32, and floating-point maxima at 25–31 bits wrapped to zero), and refuses zero-sized rows; (7) `decode_limit`: Flate (through flate2 and the fallback inflater), LZW, RunLength and ASCII85 never write past `MAX_DECODED_STREAM` (256 MiB), or, for the last filter of an image of known size, past twice its pixel bytes (a predictor tag byte per byte), at most 1 GiB; `Stream::decoded_within` for budgeted callers; a page's `/Contents` array shares one `MAX_DECODED_STREAM` (a 2 MB page whose content stream or 8 × 8 image inflated to 2 GB took the renderer past 3 GB in under a second; naming the stream twenty times in `/Contents`, 5 GB); (8) high-precision numbers (jsPDF MediaBox coordinates with over 19 digits) parse without overflowing the u64 mantissa: fractional digits past it are dropped, integer digits past it scale the value (a real, kept finite); (9) large dictionary indexes are built on first use and shared across repeated reads within each PDF (#307); the validated-byte cache is bounded, reader contexts remain separate, repairs and inline/content dictionaries bypass sharing, and small dictionaries stay eager; (10) decoded object-stream slots publish once under contention, cold loading rejects recursive dependencies, and xref repair serializes without holding recursive read guards; this permits one parser per native render pool with private worker caches; (11) `sync.rs`: a poisoned lock is recovered instead of unwrapped (a worker panic no longer cascades into every reader of the same PDF), and xref repair takes a blocking write lock instead of asserting a non-blocking one; (12) `SegmentList` (`util.rs`) returns `None` for an out-of-range index instead of panicking; (13) a page's `/UserUnit` scales `render_dimensions` and `initial_transform`, so the page and everything on it are sized in points, as Acrobat shows them; it is read as Acrobat reads it (the page's own entry, not inherited; a number from 1 up, at most 75,000; anything else counts as 1), and a page without it keeps exactly the transform it had; (14) revision 2–4 passwords are tried as given and then in PDFDocEncoding (`legacy_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Convert to PDFDocEncoding`): passwords arrive as UTF-8 text, so a password with characters outside ASCII (`Çok güzel`, `ılık€`) written in PDFDocEncoding, as ISO 32000-2 §7.6.4.3.2 asks and qpdf and Acrobat write it, was always wrong; (15) page content is read a piece at a time (`page::ContentReader`, `Page::content_reader`): plain streams and Flate streams without a predictor are decoded as they are read, with a budget of 64 times their compressed size and never less than `MAX_DECODED_STREAM` (a stream that expands past it is cut and reported); other streams are decoded whole, as before, and share `MAX_DECODED_STREAM`. The content tokenizer reads a window at a time (`UntypedIter::new_window`, `TypedIter::needs_more`, `TypedIter::committed`): an instruction that runs past the window is not decided until more data follows, which `Reader` records by flagging reads and `at_end` past the end of the data, including through the clones the parser makes. The Flate decoder keeps its own 32 KiB history, so it pauses at any output size. `ContentReader::new` reuses raw/decrypted stream bytes for budget accounting and piece creation, and `decoded_raw_within` applies the existing Whole-stream filters and limits to those bytes without decrypting twice; (16) revision 5–6 (AES-256) passwords are tried in their SASLprep form (RFC 4013, then UTF-8, at most 127 bytes, as ISO 32000-2 §7.6.4.3.3 Algorithm 2.A asks and pdfcraft-crypt writes them) and then, if that differs, as given (`aes256_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Do the above.`; `stringprep`, already in the tree through pdfcraft-crypt, is an optional dependency enabled by `std`, and without `std` passwords are used as given): the renderer used the password as typed, so one that SASLprep changes (a decomposed `café`, a no-break space, the fullwidth `ab` a Japanese IME types in full-width mode) was wrong, although pdfcraft-cos accepted it: protecting a document with such an open password was rolled back ("the password is incorrect"), and a file with such a password did not open with it.; (17) CCITT `/Rows 0` means the height is not predetermined (ISO 32000-2 Table 11), not zero rows: scanned Group 4 pages written that way rendered blank; (18) The streaming Flate decoder (`lzw_flate::flate::fallback`) copies a back-reference in runs of up to its distance (`extend_from_within`), not byte by byte. The copy reads only bytes already in the output, and a copy whose source does not exist still writes nothing, so the bytes are those a byte-by-byte copy writes. Inflating the A1 sample's content stream (278.6 MB) takes about 14% less time (0.66 s to 0.56 s); the output is identical; (19) CCITT `/Rows` below the image's `/Height` decodes to the height: a plotter driver wrote `/Rows` equal to `/Columns`, so only the first rows were decoded and the rest drawn black (#3308) (20) `stream` followed by spaces or tabs and then the line break (`stream \r\n`, as Microsoft Reporting Services writes it) starts the data after the line break; the content stream was taken to start at the space and the page rendered blank.; (21) a JP2 colour specification naming an enumerated colour space the decoder refuses (CIEJab in a veraPDF fixture, YCbCr, PhotoYCC, YPbPr, bilevel…) no longer fails the image: it is decoded again from a copy whose `colr` names the space the decoder assumes without one, greyscale, sRGB or CMYK by component count, as MuPDF and Poppler draw it (`without_refused_enumerated_colour`; the page drew blank, #3306); (22) an object stream's index of `objnum offset` pairs is parsed once per stream and kept beside its decoded data (`ObjectStreamIndexes`, `XRef::with_object_stream_indexes`), and an offset past `usize` is malformed instead of wrapping: `XRef::get_with` parsed the whole index again for every object it fetched from the stream, so reading all of a stream of N objects took N² steps (a Firefox page whose 65,542 objects share one stream took 37 s to draw and 40 s to extract its text); (23) the inline-image end-of-data look-ahead no longer rejects a candidate `EI` because a byte above 0x7F lies between it and the next one: when the candidate is the end, those bytes are the content stream that follows the image, and its strings carry such bytes (WinAnsi `(Grüße) Tj`, the two-byte codes of any Type 0 font), so the operator check that follows decides alone, bounded as in (4) (a scanned page whose OCR text held an accented word and a later capital "EI" — EINGANG, BEI, KLEIN — drew and extracted nothing after the image: 40 of 275 pages across five real documents); (24) `xref::prev_offset`: a trailer `/Prev` written as a reference `N G R` is read as the offset that object holds when the section's own entries place it (iText's `indirectPrev.pdf`; each section's entries are read before its `Prev` is followed), and otherwise as offset N (tiff2pdf; JHOVE's PDF-HUL-157 files, #2366), as `pdfcraft-cos` and other readers take it, so the earlier sections' pages are seen instead of one; (25) direct arrays and dictionaries share a reader-local nesting cap of 128, counted while skipping and while parsing a dictionary (its lenient repair reads, rather than skips, an object found where a key belongs, so `<< << …` recursed through that read); deeper objects return `None` instead of overflowing the stack (including unused trailer values, trailer entries, content operands and marked-content properties); a chain of 65 direct stitching functions now stops at this cap, before the 64-function guard; (26) an xref stream subsection must be backed by its decoded data: all-zero `/W` widths (an entry reads no bytes) and a declared count longer than the data are refused before any entry is inserted, where a tiny stream with a large `/Size` or `/Index` count inserted that many entries (from storytold/craft-libs#239); a subsection running past object number `i32::MAX` is refused instead of overflowing or turning negative; (27) an xref stream's type field is read big-endian at its full `/W[0]` width, where only its first byte was read and wider fields shifted every later entry, and `/W[0]` over 8 bytes is refused like `/W[1]` (from storytold/craft-libs#241); (28) a content instruction keeps up to 256 operands (`MAX_OPERANDS`) instead of 10, and past that drops the oldest instead of ending the content stream: the eleventh operand of a twelve-ink DeviceN `scn` (ISO 32000-2 §8.6.6.5 allows any number of components) blanked the rest of the page (#4260); (29) a `/JBIG2Globals` stream that is itself JBIG2 isn't decoded: globals are raw segments, and a chain of 16,000 such streams overflowed the stack, which aborts the process; (30) a stream's own crypt filter: `/Crypt` in its chain read as the filter it names, `/Identity` without one, instead of the document's filter and then a filter that can't be decoded, which left such a stream unreadable in any spelling (ISO 32000-2 §7.4.10); (31) `set_page_jbig2_work`: the JBIG2 images decoded on a thread spend one allowance, set per page, 2^30 steps (`MAX_PAGE_JBIG2_WORK`), so a page of many images made to reach their own limit stops at one, and what they cost to read counts, whether they decode or not; unbounded until a page sets it; (32) a group of array items read together (a tuple: an image's `/Decode`, `/W`, `/W2`, and a function's `/Domain`, `/Range`, `/Encode` and `/Decode`) can start with a reference: it was taken for the whole group and lost, which ended the array, so an image with `/Decode [7 0 R 0]` didn't draw and CID widths fell back to `/DW`; (33) an object that holds only a reference stands for what that names, stored on its own or in an object stream, up to 32 references: read as a number it was the reference's own object number, so a `/Decode` or `/W` entry naming such an object took a wrong value, and as anything else it was lost. Past 32, as any cycle goes, it reads as a missing object. An object stream isn't reached that way, as its checks are of the object named; reading its index counts against the 64 nested loads and reads its dictionary alone, as a chain of them, each with a member of the next for its `/N`, `/First` or `/Length`, ran out of stack; and a comment inside a reference counts as white space; (34) a table rebuilt by scanning the file (`fallback_xref_map`) finds an object by its number when its header names another generation, including one no generation may be (`3 18446744073709551616 obj` in `poppler-85140-0.pdf`, #5580): the header is read leniently only in that scan (`read_object_header`), the object is also filed under generation 0 unless a real `N 0 obj` holds that key (a later header for the number moves the stand-in), `XRef::get_with` matches by object number only in a repaired table, and `XRef::objects` yields each location once; (35) PNG predictors (`lzw_flate::unfilter_png`) are undone a byte row at a time on slices, as the PNG filters are defined, instead of through the bit-level readers TIFF predictors need: a 993 × 936 RGB image took about 63 ms to unfilter whatever its row filters, and now takes about 3 ms (Up) to 16 ms (Paeth) in a release build (#5430); a row ending part-way through a pixel (three 4-bit samples, an odd width) keeps its last byte, where the bit-level readers dropped it and each later row started a byte early, shearing the picture; (42) TIFF predictor 2 on 8- and 16-bit samples (`lzw_flate::undo_tiff_bytes`) is undone a byte row at a time on slices, each sample added to the same component of the pixel to its left, wrapping at the sample's width (16-bit samples big-endian), instead of through the bit-level readers 1-, 2- and 4-bit samples still need; the bytes are those the bit-level path wrote, and a partial last row is still dropped. A page of 45 full-page 2445 × 4724 8-bit grey images so predicted (PDFBOX-1350) took about 23 s to render at 96 dpi and now takes about 4 s, with the same pixels | (1) `cos/tests/writer.rs::page_tree_loops_in_object_streams_do_not_crash_readers`; (2) `raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding`; (3) `raster::tests::absurd_predictor_columns_and_line_widths_render`; (4) `raster::tests::inline_images_with_ei_in_their_data_parse_in_linear_time`, `content::tests::inline_image_ei_may_be_followed_by_a_delimiter`; (5) `raster::tests::indirect_appearance_state_dict_is_not_another_stream`; (6) `raster::tests::jpx_stream_bit_depth_boundaries_do_not_panic`, `raster::tests::jpx_32_bit_image_does_not_prevent_other_page_content_from_rendering`, `raster::tests::jpx_gray_rgb_and_alpha_keep_eight_bit_samples`, `raster::tests::jpx_streams_reject_out_of_range_bit_depths`; (7) `raster::tests::decompression_bombs_stop_at_the_stream_limit`, `raster::tests::page_contents_and_nested_forms_share_one_budget`; (8) `inspect::tests::jspdf_high_precision_mediabox_does_not_overflow_dimensions`, `object::number::tests::high_precision_float_does_not_overflow_mantissa` (hayro-syntax); (9) `object::dict::lazy_index_tests`, `raster::tests::large_resource_indexes_preserve_pixels_and_text`; (10) `data::tests`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`, `raster::tests::shared_parser_is_lazy_and_initializes_once_under_contention`; (11) `data::tests::panicking_initializer_wakes_waiters_and_prevents_retries`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`; (12) `util::tests::segment_list_rejects_out_of_range_indices`; (13) `raster::tests::user_unit_scales_the_page_and_everything_on_it`; (14) `crypto::pdfcraft_password_tests`, `pdfcraft-engine` `tests::legacy_passwords_with_turkish_characters_open`; (15) `content::tests::windows_keep_every_instruction_at_every_split`, `content::tests::windows_keep_every_instruction_in_pieces_of_any_size`, `content::tests::windows_match_the_slice_on_random_content`, `content::tests::windows_drop_blanks_and_keep_a_comment_that_may_go_on`, `page::content_reader_tests::content_past_the_cap_is_read_to_the_end`, `page::content_reader_tests::a_stream_that_expands_past_the_budget_is_cut_there`, `page::content_reader_tests::a_small_expansion_under_the_cap_is_not_cut`, `raster::tests::page_content_past_the_old_cap_renders`, `raster::tests::content_reader_returns_the_same_bytes_in_any_pieces`, `raster::tests::content_bombs_stop_at_the_budget_and_say_so`, `page::content_reader_tests::plain_and_whole_pieces_keep_their_bytes`, `page::content_reader_tests::repeated_content_reference_counts_compressed_bytes_once`, `page::content_reader_tests::whole_cap_still_counts_later_unique_streams`; (16) `crypto::pdfcraft_password_tests::aes256_passwords_are_tried_in_their_saslprep_form_then_as_given`, `crypto::pdfcraft_password_tests::aes256_passwords_open_as_typed_and_in_their_saslprep_form`, `pdfcraft-engine` `tests::aes256_open_passwords_that_saslprep_changes_protect_and_reopen_as_typed`, `tests::aes256_files_whose_passwords_saslprep_changes_open_as_typed`; (17) `raster::tests::group4_fax_with_rows_zero_decodes_to_the_image_height`; (18) `filter::lzw_flate::tests::back_reference_longer_than_its_distance_repeats_its_source`, `back_reference_cut_by_the_limit_resumes_in_the_next_call` and `back_reference_reads_bytes_written_by_an_earlier_call`; each fails when the copy returns the wrong remaining count; (19) `raster::tests::group4_fax_with_rows_below_its_height_decodes_to_the_image_height`; (20) `raster::tests::space_before_the_line_break_after_stream_still_renders`; (21) `raster::tests::a_jp2_image_in_a_refused_enumerated_colour_space_is_drawn`, `filter::jpx::tests::refused_enumerated_colour_spaces_become_the_component_counts_default`, `filter::jpx::tests::what_the_decoder_handles_or_cant_be_read_is_left_alone` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (22) `raster::tests::objects_in_one_large_object_stream_are_read_in_linear_time`; (23) `text::tests::text_after_an_inline_image_survives_a_non_ascii_byte_before_a_later_ei`, `content::tests::an_inline_image_ends_at_its_own_ei_when_later_text_holds_a_non_ascii_byte` (hayro-syntax); (24) `xref::tests::prev_as_reference_reaches_the_earlier_section`, `xref::tests::prev_as_reference_to_an_object_holding_the_offset_reads_that_offset`, `pdfcraft-cos` `document::tests::a_prev_written_as_a_reference_still_reaches_the_earlier_sections`, `document::tests::a_prev_reference_to_an_object_holding_the_offset_reads_that_offset`; (25) `raster::tests::deeply_nested_direct_objects_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_content_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_the_trailer_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_as_marked_content_properties_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_are_read_up_to_the_nesting_cap`, `raster::tests::nesting_at_the_cap_fits_a_512_kib_stack` (native only), `raster::tests::function_limits_depth_guard_refuses_a_chain_of_references`; (26) `xref::tests::xref_stream_entries_need_backing_bytes`; (27) `xref::tests::xref_stream_type_field_uses_the_full_w0_width`, `xref::tests::xref_stream_loader_rejects_oversized_w0` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (28) `content::ops::tests::scn_with_twelve_components_keeps_them_all`, `content::ops::tests::operands_past_the_cap_keep_the_latest_in_order` (hayro-syntax), `raster::tests::devicen_inks::a_twelve_ink_devicen_colour_paints`, `raster::tests::devicen_inks::a_four_ink_devicen_colour_paints`, `raster::tests::devicen_inks::an_operator_with_hundreds_of_operands_does_not_end_the_content`; (29) `raster::tests::jbig2_globals_that_are_jbig2_themselves_are_not_decoded`, `raster::tests::jbig2_globals_are_read`, `raster::tests::the_jbig2_images_of_a_page_share_one_allowance`, `raster::tests::an_image_that_fails_spends_the_pages_allowance_too`; (30) `raster::tests::a_stream_with_a_crypt_filter_of_its_own_is_drawn`; (32) `raster::tests::a_decode_array_with_references_in_it_is_read`, `raster::tests::a_functions_arrays_can_start_with_a_reference` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret); (33) `raster::tests::a_reference_to_a_reference_is_what_it_names`, `raster::tests::a_decode_array_with_references_in_it_is_read` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret), `xref::tests::an_object_stream_isnt_reached_through_a_reference`, `xref::tests::a_chain_of_object_streams_ends_at_the_limit_of_nested_loads`, `xref::tests::a_chain_of_object_streams_takes_no_more_work_for_two_ways_on`, `xref::tests::a_reference_held_in_either_storage_is_followed`, `xref::tests::a_page_a_reference_is_held_to_is_found_once` (hayro-syntax); (34) `raster::tests::a_rebuilt_table_finds_an_object_whose_header_names_another_generation`, `raster::tests::a_rebuilt_tables_generation_0_stand_in_is_not_a_second_object` (pdfcraft-render), `xref::tests::an_object_stream_and_its_first_member_are_both_enumerated` (hayro-syntax); (35) `filter::lzw_flate::pdfcraft_png_predictor_tests` (hayro-syntax: every filter type at 1–16 bits and 1–4 colours against the PNG specification's byte-by-byte definition, a row ending part-way through a pixel, Average and Paeth over 6-byte pixels), `raster::tests::png_predicted_images::a_four_bit_rgb_image_of_odd_width_with_up_predicted_rows_is_not_sheared`; (42) `filter::lzw_flate::pdfcraft_tiff_predictor_tests` (hayro-syntax: random 8- and 16-bit rows at 1–7 colours and 1–50 columns against the bit-level path, partial and empty rows, and a row counter showing whole-byte samples skip the bit-level path), `raster::tests::tiff_predicted_images` | | winit | 0.30.13 | Apache-2.0 | (1) on Windows 11 (build 22000 and later), `WM_DPICHANGED` applies the rect Windows suggests. It used to nudge the window onto the monitor `MonitorFromWindow` names, which during a drag is still the monitor being left; the nudge sent another `WM_DPICHANGED`, so a window dragged to a monitor with another scale factor grew and stayed on the wrong monitor at the wrong scale (#324). Backport of winit master 488c036a ([winit#4341](https://github.com/rust-windowing/winit/pull/4341)), which no 0.30 release has; Windows 10 keeps the old adjustment, as on winit master. (2) on Wayland, a seat's text input is destroyed when the seat goes away, not whenever it loses a capability: losing any capability (a touch screen or tablet unplugged, a remote-desktop session ending) destroyed it, and only an added capability recreated it, so IME stopped working until restart. Port of [winit#4747](https://github.com/rust-windowing/winit/pull/4747); (3) on Windows, an IME composition reports its cursor from `GCS_CURSORPOS` only when `WM_IME_COMPOSITION` says it is set, and at the end of the composition otherwise: the Microsoft Korean IME never sets it, so the cursor read as 0 (egui then drew no caret while composing; it now draws it after the syllable). Port of [winit#4746](https://github.com/rust-windowing/winit/pull/4746); (4) on Windows, a hidden window is not maximized until it is shown, and a hidden window shown while maximized gets one `ShowWindow(SW_MAXIMIZE)`: `SW_MAXIMIZE` shows a hidden window, so PdfCraft, which eframe keeps hidden until its first frame is drawn, appeared unpainted (white) when it restored a maximized window, and could lose its focus. Port of [winit#4587](https://github.com/rust-windowing/winit/pull/4587) (the later, focus-safe form of [winit#4575](https://github.com/rust-windowing/winit/pull/4575)); (5) on X11, when allowing or disallowing IME replaces the focused window's input context (egui does it each time a text field gains or loses focus), the new context is focused: no focus event follows, so it stayed unfocused until the window lost and regained focus. Port of [winit#4727](https://github.com/rust-windowing/winit/pull/4727); (6) on macOS, `insertText` commits through IME when a composition was started (`pending_commit`, set by `setMarkedText`) rather than when marked text is left: input methods that end a composition with `unmarkText` before `insertText` lost the text, and Apple Korean dropped the key typed after a syllable (`한` then `5` or `;`); other `insertText` calls stay `KeyboardInput`; a commit outside a key press (a candidate clicked with the mouse, dictation) clears the marked text and leaves no key to swallow, as in [winit#4748](https://github.com/rust-windowing/winit/pull/4748) (without it the next key, such as Backspace, was lost). `validAttributesForMarkedText` lists the underline and marked-clause attributes, without which some input methods don't compose. Port of [winit#4650](https://github.com/rust-windowing/winit/pull/4650), the 0.30 backport of [winit#4651](https://github.com/rust-windowing/winit/pull/4651) (merged on winit master), with the underline attribute spelled as AppKit's `NSUnderlineStyleAttributeName` (`NSUnderline`); (7) on macOS, disallowing IME with a composition in progress (focus leaving a text field) tells the input method to discard its marked text (`discardMarkedText`): it kept composing it once IME was allowed again, so the abandoned syllable came back (`하`, then `가` typed in the next field, became `학`, then `하` + `가`; Japanese `か` then `さ` became `傘`). Port of [winit#4745](https://github.com/rust-windowing/winit/pull/4745); (8) on macOS, text inserted outside a key press without a composition (the emoji picker) is reported as `Ime::Commit`; it was dropped. Port of the emoji-picker part of [winit#4749](https://github.com/rust-windowing/winit/pull/4749) (0.30 backport of [winit#4748](https://github.com/rust-windowing/winit/pull/4748)); its `selectedRange` change for dictation is left out because it switches on press-and-hold for every held key (an accent menu instead of key repeat, and the chosen accent added after the repeated letters); (9) on macOS, a key whose text was committed through IME isn't also sent as `KeyboardInput` unless AppKit forwards it as a command (Enter): Apple Korean answers Space with no syllable in progress with `setMarkedText(" ")`, `insertText(" ")`, `setMarkedText("")`, and the last call reset winit's state, so the space arrived twice. Same goal as [winit#4478](https://github.com/rust-windowing/winit/pull/4478), whose changes don't reach this call order (macOS 14); to be proposed upstream there; (10) on macOS, when the first key in a fresh app reaches Apple Korean's `insertText` as a raw compatibility jamo instead of starting a composition, that key is interpreted once more, synchronously, before it is sent as `KeyboardInput` (g k s typed ㅎㅏㄴ instead of 한); only an unmodified, non-repeated first key whose one callback inserts that same jamo, with IME enabled and nothing composed, focus and input source unchanged, is retried (`ime_startup.rs`). Port of [winit#4744](https://github.com/rust-windowing/winit/pull/4744), with its policy tests.; (11) on X11 without `Xft.dpi`, the scale factor comes from the X screen's configured DPI (its size in pixels and millimetres) before the output's physical density from RandR: Xorg can deliberately run a dense panel at 96 DPI, and only winit applications then looked zoomed; `Xft.dpi` still wins, `WINIT_X11_SCALE_FACTOR` still overrides, and a screen reporting no size keeps the RandR fallback. PdfCraft patch (#1150, replacing the launcher pin proposed in #908). (12) on macOS, `WinitView` answers `mouseDownCanMoveWindow` with the window's `movable_by_window_background` attribute (NO unless asked): NSView answers YES for a view that isn't opaque, so with the transparent title bar and full-size content view PdfCraft uses, AppKit moved the window for any drag that began in the 28 pt title-bar band before the app saw it, and document tabs (which live in that band) could not be dragged to reorder them. The empty strip still moves the window through `ViewportCommand::StartDrag`. PdfCraft patch. The published crate's asset files (`docs/res/*.svg`, `examples/data/*.png`) are left out, so its `window` example doesn't build; (13) on Wayland, pixel-only wheel, wheel-tilt and continuous-source scroll events report `Moved` rather than starting an unterminated finger gesture, so egui follows modifier release instead of keeping Ctrl zoom latched (#991). Finger gestures and unknown sources retain their existing phase policy | (1) `pdfcraft` `tests::winit_carries_the_windows_11_monitor_scale_fix` keeps the patch in the build; the drag itself needs two monitors with different scale factors on Windows 11; (2) `pdfcraft` `tests/winit_wayland_text_input.rs`; the capability change itself needs a Wayland session and an input method; (3) `pdfcraft` `tests/winit_windows_ime_cursor.rs`; composing itself needs the Microsoft Korean IME; (4) `pdfcraft` `tests/winit_windows_hidden_maximized.rs`; the start-up itself needs Windows and a window that was closed maximized; (5) `pdfcraft` `tests/winit_x11_ime_focus.rs`; switching fields itself needs an X11 session and an input method (IBus, fcitx5); (6) `pdfcraft` `tests/winit_macos_ime_commit.rs`; composing itself needs macOS and an input method; (7) `pdfcraft` `tests/winit_macos_ime_discard.rs`; composing itself needs macOS and an input method; (8) `pdfcraft` `tests/winit_macos_emoji_picker.rs`; the picker itself needs macOS; (9) `pdfcraft` `tests/winit_macos_korean_space.rs`; composing itself needs macOS and Apple Korean; (10) `pdfcraft` `tests/winit_macos_korean_first_key.rs` and `ime_startup.rs`'s own tests (run them with `rustc --test`); typing itself needs macOS and Apple Korean; (11) `pdfcraft` `tests/winit_x11_dpi.rs` keeps the patch in the build, and `util::randr` tests the policy (run them in the vendored crate); the scale itself needs an X11 session; (12) `pdfcraft` `tests/winit_macos_title_bar_drag.rs` keeps the patch in the build; the drag itself needs macOS and a pointer; (13) `pdfcraft` `tests/winit_wayland_scroll.rs` compiles the production phase policy and exercises modifier release through egui; live COSMIC confirmation remains open | | egui-wgpu | 0.36.2 | MIT OR Apache-2.0 | (1) `winit::surface_fit`: a window's surface is configured within the device's `max_texture_dimension_2d`, both sides shrunk by one factor that egui then draws at (`pixels_per_point × render_scale`), so the whole window is drawn into the surface (on DX12 it is stretched over the window and lines up with the pointer; wgpu's GL backend copies it unscaled into a corner); the longer side becomes exactly the limit; the depth and MSAA textures follow, and a recreated surface starts again from the window's own size. egui-wgpu configured the window's size as is, and `Surface::configure` panics beyond the limit: a window restored at another monitor's scale (opening an Outlook attachment on a mixed-DPI laptop asked an 8192-pixel device for 8600 × 3423, #577) or stretched across monitors closed PdfCraft, before its window appeared when it happened at start-up ([emilk/egui#8361](https://github.com/emilk/egui/issues/8361)). The published crate leaves out egui's licence files; `LICENSE-APACHE` and `LICENSE-MIT` are from the 0.36.2 tag; (2) `winit::catch_errors`: a new window's surface is configured inside wgpu error scopes (validation, out of memory, internal), and a failure is returned as `WgpuError::ConfigureSurface` (for the root window at start-up eframe returns it as `Error::Wgpu`, which PdfCraft retries with OpenGL); nothing is kept for the window. wgpu's default handler panicked instead, and eframe configures the root surface before the app is created, so a driver that can't configure one closed PdfCraft at launch (a 2015 Intel GPU through wgpu's GL backend timed out waiting for the GPU to go idle, `GpuWaitTimeout`, #519). Unchanged, so still a panic: an error in a later reconfigure (resize, a lost surface), a device lost during the first configure (wgpu reports it only to a callback), and a window shown at zero size (first configured on resize) | (1) `pdfcraft` `tests::a_window_larger_than_the_gpu_limit_gets_a_surface_within_it`, `tests::surface_fit_never_empties_or_overflows_a_surface`, and `tests::vendor_provenance_egui_wgpu_surface_fit_identity`, which keeps the patch in the build and in the paths that size a surface; (2) `pdfcraft` `tests::a_gpu_error_while_a_window_is_set_up_is_returned_not_a_panic` (a real device: an error inside `catch_errors` is returned, not a panic) and `tests::vendor_provenance_egui_wgpu_catch_errors_identity`, which keeps it around `add_surface`'s configure | | vello_common | 0.0.8 | Apache-2.0 OR MIT | (1) `clip::intersect` (the clip step every clipped path passes through) starts its row walk at the first row both paths have strips in, not the earlier of the two first rows. The walk visited every row from the top of the clip for each path (on the A1 sample's full-page canvas, about half its strip rows per stroke), and rows where one path has no strips produce no output, so the strips are unchanged (vello_common 0.1 starts at the later row too). Its cost grew with the page, and a full-page render of the A1 sample was about 30% faster; the output is byte-identical | (1) `clip::tests::intersect_walks_only_rows_both_paths_have_strips` counts the rows the walk visits (`cargo test --manifest-path vendor/vello_common/Cargo.toml`) and fails without the patch; the render crate's pixel tests run through the patched crate | diff --git a/vendor/hayro-syntax/src/filter/lzw_flate.rs b/vendor/hayro-syntax/src/filter/lzw_flate.rs index 5a0027597b..2ebfa4b7f9 100644 --- a/vendor/hayro-syntax/src/filter/lzw_flate.rs +++ b/vendor/hayro-syntax/src/filter/lzw_flate.rs @@ -1017,6 +1017,17 @@ fn apply_predictor(data: Vec, params: &PredictorParams) -> Option> { return Some(unfilter_png(&data, row_len, bpp)); } + if i == 2 && matches!(params.bits_per_component, 8 | 16) && !bit_level_only() { + // PdfCraft patch: whole-byte samples are undone a row at a time on byte slices + // instead of through the bit-level readers, which packed samples need. + let sample_len = usize::from(params.bits_per_component / 8); + let pixel_len = usize::from(params.colors).saturating_mul(sample_len); + return Some(undo_tiff_bytes(data, row_len, pixel_len, sample_len)); + } + + #[cfg(all(test, feature = "std"))] + BIT_LEVEL_TIFF_ROWS.with(|rows| rows.set(rows.get().saturating_add(num_rows))); + let (bit_size, chunk_len) = (params.bits_per_component, params.colors as usize); let zero_row = vec![0; row_len]; let mut prev_row = BitChunks::new(&zero_row, bit_size, chunk_len)?; @@ -1054,6 +1065,69 @@ fn apply_predictor(data: Vec, params: &PredictorParams) -> Option> { } } +/// PdfCraft patch: undoes TIFF predictor 2 (ISO 32000-2 §7.4.4.4, Table 10; TIFF 6.0 §14) +/// for 8- and 16-bit samples, in rows of `row_len` bytes holding pixels of `pixel_len` bytes, +/// each sample `sample_len` bytes (1 or 2, big-endian): every sample is added to the same +/// component of the pixel to its left, wrapping at the sample's width. Bytes after the last +/// complete row are dropped, as the bit-level path drops them. +fn undo_tiff_bytes( + mut data: Vec, + row_len: usize, + pixel_len: usize, + sample_len: usize, +) -> Vec { + let num_rows = data.len() / row_len.max(1); + data.truncate(num_rows.saturating_mul(row_len)); + if pixel_len == 0 { + return data; + } + + for row in data.chunks_exact_mut(row_len) { + let mut pixels = row.chunks_exact_mut(pixel_len); + let Some(mut left) = pixels.next() else { + continue; + }; + for px in pixels { + if sample_len == 2 { + for (c, l) in px.chunks_exact_mut(2).zip(left.chunks_exact(2)) { + if let ([c0, c1], [l0, l1]) = (c, l) { + let sum = u16::from_be_bytes([*c0, *c1]) + .wrapping_add(u16::from_be_bytes([*l0, *l1])); + [*c0, *c1] = sum.to_be_bytes(); + } + } + } else { + for (c, l) in px.iter_mut().zip(left.iter()) { + *c = c.wrapping_add(*l); + } + } + left = px; + } + } + + data +} + +#[cfg(all(test, feature = "std"))] +std::thread_local! { + /// PdfCraft patch tests: rows the bit-level TIFF predictor path has undone on this thread. + static BIT_LEVEL_TIFF_ROWS: core::cell::Cell = const { core::cell::Cell::new(0) }; + /// PdfCraft patch tests: send whole-byte TIFF samples through the bit-level path too. + static BIT_LEVEL_ONLY: core::cell::Cell = const { core::cell::Cell::new(false) }; +} + +/// PdfCraft patch: whether whole-byte TIFF samples must take the bit-level path (only when a +/// test compares the two paths). +#[cfg(all(test, feature = "std"))] +fn bit_level_only() -> bool { + BIT_LEVEL_ONLY.with(core::cell::Cell::get) +} + +#[cfg(not(all(test, feature = "std")))] +fn bit_level_only() -> bool { + false +} + /// PdfCraft patch: undoes PNG row filters (ISO 32000-2 §7.4.4.4, PNG specification §9) in /// rows of `row_len` bytes, each after its filter-type byte, with `bpp` bytes per complete pixel /// (at least 1). An unknown filter type copies the row as it is. Bytes after the last complete @@ -1491,3 +1565,116 @@ mod pdfcraft_png_predictor_tests { ); } } + +/// PdfCraft patch tests: TIFF predictor 2 on 8- and 16-bit samples is undone on byte rows +/// (`undo_tiff_bytes`), with the same output as the bit-level path. +#[cfg(all(test, feature = "std"))] +mod pdfcraft_tiff_predictor_tests { + use super::{BIT_LEVEL_ONLY, BIT_LEVEL_TIFF_ROWS}; + use crate::filter::lzw_flate::{PredictorParams, apply_predictor}; + + fn tiff(colors: u8, bits_per_component: u8, columns: usize) -> PredictorParams { + PredictorParams { + predictor: 2, + colors, + bits_per_component, + columns, + early_change: false, + } + } + + /// Both paths' output, the byte path's first; `None` from either is kept. + fn both(data: &[u8], params: &PredictorParams) -> (Option>, Option>) { + let fast = apply_predictor(data.to_vec(), params); + BIT_LEVEL_ONLY.with(|only| only.set(true)); + let bits = apply_predictor(data.to_vec(), params); + BIT_LEVEL_ONLY.with(|only| only.set(false)); + (fast, bits) + } + + #[test] + fn whole_byte_samples_match_the_bit_level_path_on_random_rows() { + let mut seed: u32 = 0x9e37_79b9; + let mut next = move || { + seed ^= seed << 13; + seed ^= seed >> 17; + seed ^= seed << 5; + seed + }; + for case in 0..600 { + let colors = [1u8, 2, 3, 4, 5, 7][(next() % 6) as usize]; + let bits_per_component = [8u8, 16][(next() % 2) as usize]; + let columns = 1 + (next() % 50) as usize; + let row_len = columns * usize::from(colors) * usize::from(bits_per_component) / 8; + let rows = (next() % 7) as usize; + // A partial row at the end, or none. + let tail = (next() % 4) as usize % row_len; + let data: Vec = (0..rows * row_len + tail).map(|_| next() as u8).collect(); + let params = tiff(colors, bits_per_component, columns); + let (fast, bits) = both(&data, ¶ms); + assert_eq!( + fast, bits, + "case {case}: {colors} colours at {bits_per_component} bits, {columns} columns, {rows} rows and {tail} bytes" + ); + } + } + + #[test] + fn whole_byte_samples_add_to_the_same_component_of_the_pixel_to_their_left() { + // Two 8-bit colours, three columns, two rows; sums wrap at 256. + let data = [10, 20, 1, 2, 3, 4, /* row 2 */ 5, 5, 250, 10, 1, 1]; + assert_eq!( + apply_predictor(data.to_vec(), &tiff(2, 8, 3)).unwrap(), + [10, 20, 11, 22, 14, 26, 5, 5, 255, 15, 0, 16] + ); + // One 16-bit colour: a carry crosses the byte boundary and the sum wraps at 65,536. + let data = [ + 0x01, 0x00, 0xff, 0xff, 0x00, 0x02, /* row 2 */ 0x00, 0xff, 0x00, 0x01, 0, 0, + ]; + assert_eq!( + apply_predictor(data.to_vec(), &tiff(1, 16, 3)).unwrap(), + [ + 0x01, 0x00, 0x00, 0xff, 0x01, 0x01, 0x00, 0xff, 0x01, 0x00, 0x01, 0x00 + ] + ); + } + + #[test] + fn whole_byte_samples_do_not_take_the_bit_level_path() { + let start = BIT_LEVEL_TIFF_ROWS.with(core::cell::Cell::get); + for bits_per_component in [8, 16] { + let data = vec![7; 3 * 4 * usize::from(bits_per_component) / 8 * 5]; + apply_predictor(data, &tiff(3, bits_per_component, 4)).unwrap(); + } + assert_eq!(BIT_LEVEL_TIFF_ROWS.with(core::cell::Cell::get), start); + + // Packed samples still do. + apply_predictor(vec![0x5f; 4], &tiff(1, 4, 2)).unwrap(); + assert_eq!(BIT_LEVEL_TIFF_ROWS.with(core::cell::Cell::get), start + 4); + } + + #[test] + fn rows_without_a_complete_row_of_data_are_dropped_on_both_paths() { + for (data, params, want) in [ + // Shorter than one row: nothing. + (vec![1, 2, 3], tiff(1, 8, 4), Some(vec![])), + (vec![1, 2, 3], tiff(1, 16, 2), Some(vec![])), + // A zero-width row is refused. + (vec![1, 2, 3], tiff(1, 8, 0), None), + (vec![], tiff(1, 16, 0), None), + // No data at all. + (vec![], tiff(3, 8, 2), Some(vec![])), + // The trailing partial row is dropped. + ( + vec![1, 1, 1, 1, 2, 2], + tiff(1, 8, 4), + Some(vec![1, 2, 3, 4]), + ), + (vec![0, 1, 0, 1, 9], tiff(1, 16, 2), Some(vec![0, 1, 0, 2])), + ] { + let (fast, bits) = both(&data, ¶ms); + assert_eq!(fast, want, "{data:?}"); + assert_eq!(bits, want, "{data:?}"); + } + } +} From 42b21e5362c99be04b3fd48c23ba58290a815987 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:30:14 +0800 Subject: [PATCH 06/75] fix(render): a form XObject that draws itself is drawn once (hayro-interpret patch 89) (#6515) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * M2: a form XObject that draws itself is drawn once (hayro-interpret patch 89) A form whose content names itself, directly or through another form, a pattern or a glyph, drew itself again at every level until the page's nested-paint budget (50,000) ran out. PDFBOX-5782's poppler-748 page 2, whose form names a /Pages dictionary as /X0 and so finds itself in the page's resources, took 38 s. hayro-interpret now keeps the forms being drawn on the context and skips a form already on it, reporting ContentTruncated (ISO 32000-2 §8.10.1: a form is a self-contained description). The lenient fallback from a form's own resources to the page's stays: MuPDF, Poppler and PDFium all draw a name from the page's resources when the form's /Resources has no XObject subdictionary (MuPDF and Poppler also when it has an empty one). Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * fix(render): a form that reaches itself through a pattern or a glyph is drawn once (hayro-interpret patch 89) The forms being drawn are now kept per page and thread rather than per context, so the tiling patterns, Type 3 glyphs, soft masks and groups painted inside a form, which hayro interprets in contexts of their own, see that the form is being drawn and skip it inside themselves. A form drawn again after it has finished draws as before. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA * chore: re-run CI after the Fill & Sign save comparison failed on the clock (#6066) `fill_sign_enter_adds_lines_outside_click_commits_and_escape_cancels` compares two saves whose /ModDate fell in different seconds; it passes locally on this commit. No change. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA --------- Co-authored-by: mtlcl410 Co-authored-by: Claude Opus 5.5 --- crates/render/src/raster/tests.rs | 86 ++++++++++++++++++++++++++ parity/acrobat-features.toml | 2 +- vendor/README.md | 2 +- vendor/hayro-interpret/src/context.rs | 58 +++++++++++++++++ vendor/hayro-interpret/src/x_object.rs | 18 ++++++ 5 files changed, 164 insertions(+), 2 deletions(-) diff --git a/crates/render/src/raster/tests.rs b/crates/render/src/raster/tests.rs index f6569e51fa..92ec72afee 100644 --- a/crates/render/src/raster/tests.rs +++ b/crates/render/src/raster/tests.rs @@ -5796,6 +5796,92 @@ fn page_contents_and_nested_forms_share_one_budget() { assert_eq!(&page.rgba[((38 * 40 + 1) * 4)..][..4], &[255, 0, 0, 255], "the rest of the page draws"); } +/// A form XObject that draws itself (PDFBOX-5782, a fuzzed file whose form names a `/Pages` +/// dictionary as `/X0`, found the form again in the page's resources and spent 38 s drawing +/// itself until the page's nested-paint budget ran out) is drawn once: the drawing inside +/// itself is skipped and reported, as other readers draw it. Vendored patch: hayro-interpret +/// `Context::enter_form` (89). +#[test] +fn a_form_that_draws_itself_is_drawn_once() { + // Each form paints a 10 x 10 square and draws `/X0` 20 units to its right. + let draw = "0 0 1 rg 0 0 10 10 re f q 1 0 0 1 20 0 cm /X0 Do Q"; + // Two forms that draw each other: blue, then red, then blue again would follow. + let other = "1 0 0 rg 0 0 10 10 re f q 1 0 0 1 20 0 cm /X0 Do Q"; + let other = format!( + "6 0 obj << /Type /XObject /Subtype /Form /BBox [0 0 200 40] /Resources << /XObject << /X0 5 0 R >> >> /Length {} >> stream\n{other}\nendstream endobj\n", + other.len() + ); + let cases = [ + ("its own resources name it", "/Resources << /XObject << /X0 5 0 R >> >>", ""), + ("it names a /Pages dictionary and the page's resources name it", "/Resources << /XObject << /X0 2 0 R >> >>", ""), + ("it has no resources and the page's name it", "", ""), + ("two forms draw each other", "/Resources << /XObject << /X0 6 0 R >> >>", other.as_str()), + ]; + for (what, resources, extra) in cases { + let mut pdf = b"%PDF-1.7\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 200 40] /Contents 4 0 R /Resources << /XObject << /X0 5 0 R >> >> >> endobj\n4 0 obj << /Length 7 >> stream\n/X0 Do\nendstream endobj\n".to_vec(); + pdf.extend_from_slice( + format!( + "5 0 obj << /Type /XObject /Subtype /Form /BBox [0 0 200 40] {resources} /Length {} >> stream\n{draw}\nendstream endobj\n", + draw.len() + ) + .as_bytes(), + ); + pdf.extend_from_slice(extra.as_bytes()); + pdf.extend_from_slice(b"trailer << /Root 1 0 R >>\n%%EOF\n"); + let mut r = PageRenderer::new(Arc::new(pdf), RenderConfig::default()); + let page = r.render(RenderRequest { page: 0, kind: RequestKind::Pixels, tile: None, scale: 1.0, tag: 0, ..Default::default() }); + assert!(page.error.is_none(), "{what}: {:?}", page.error); + assert_eq!(page.width, 200, "{what}"); + assert_eq!(rgb_at(&page.rgba, 200, 5, 35), [0, 0, 255], "{what}: the form draws"); + let second = if extra.is_empty() { [255, 255, 255] } else { [255, 0, 0] }; + assert_eq!(rgb_at(&page.rgba, 200, 25, 35), second, "{what}: the form one level in"); + assert_eq!(rgb_at(&page.rgba, 200, 45, 35), [255, 255, 255], "{what}: the form is not drawn inside itself"); + assert!(page.warnings.contains(&RenderWarning::ContentTruncated), "{what}: the skipped drawing is reported"); + } +} + +/// A form that reaches itself through a tiling pattern it fills with, or a Type 3 glyph it +/// shows, is not drawn inside itself either: the pattern cell or the glyph procedure is +/// interpreted in a context of its own, which still knows the form is being drawn. The form +/// paints a blue square, then fills a square 20 units to its right with the pattern, or shows +/// the glyph there, whose content draws the form. Vendored patch: hayro-interpret +/// `Context::enter_form` (89). +#[test] +fn a_form_that_reaches_itself_through_a_pattern_or_a_glyph_is_drawn_once() { + let cell = "q 1 0 0 1 20 0 cm /X0 Do Q"; + let pattern = format!( + "7 0 obj << /Type /Pattern /PatternType 1 /PaintType 1 /TilingType 1 /BBox [0 0 200 40] /XStep 200 /YStep 40 /Resources << /XObject << /X0 5 0 R >> >> /Length {} >> stream\n{cell}\nendstream endobj\n", + cell.len() + ); + let glyph = "0 0 d0 /X0 Do"; + let glyph = format!( + "7 0 obj << /Type /Font /Subtype /Type3 /FontBBox [0 0 200 40] /FontMatrix [1 0 0 1 0 0] /CharProcs << /a 8 0 R >> /Encoding << /Type /Encoding /Differences [97 /a] >> /FirstChar 97 /LastChar 97 /Widths [10] /Resources << /XObject << /X0 5 0 R >> >> >> endobj\n8 0 obj << /Length {} >> stream\n{glyph}\nendstream endobj\n", + glyph.len() + ); + let cases = [ + ("a tiling pattern", "/Resources << /Pattern << /P0 7 0 R >> >>", "0 0 1 rg 0 0 10 10 re f /Pattern cs /P0 scn 20 0 10 10 re f", pattern), + ("a Type 3 glyph", "/Resources << /Font << /T3 7 0 R >> >>", "0 0 1 rg 0 0 10 10 re f BT /T3 1 Tf 20 0 Td (a) Tj ET", glyph), + ]; + for (what, resources, draw, extra) in cases { + let mut pdf = b"%PDF-1.7\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 200 40] /Contents 4 0 R /Resources << /XObject << /X0 5 0 R >> >> >> endobj\n4 0 obj << /Length 7 >> stream\n/X0 Do\nendstream endobj\n".to_vec(); + pdf.extend_from_slice( + format!( + "5 0 obj << /Type /XObject /Subtype /Form /BBox [0 0 200 40] {resources} /Length {} >> stream\n{draw}\nendstream endobj\n", + draw.len() + ) + .as_bytes(), + ); + pdf.extend_from_slice(extra.as_bytes()); + pdf.extend_from_slice(b"trailer << /Root 1 0 R >>\n%%EOF\n"); + let mut r = PageRenderer::new(Arc::new(pdf), RenderConfig::default()); + let page = r.render(RenderRequest { page: 0, kind: RequestKind::Pixels, tile: None, scale: 1.0, tag: 0, ..Default::default() }); + assert!(page.error.is_none(), "{what}: {:?}", page.error); + assert_eq!(rgb_at(&page.rgba, 200, 5, 35), [0, 0, 255], "{what}: the form draws"); + assert_eq!(rgb_at(&page.rgba, 200, 25, 35), [255, 255, 255], "{what}: the form is not drawn inside itself"); + assert!(page.warnings.contains(&RenderWarning::ContentTruncated), "{what}: the skipped drawing is reported"); + } +} + /// The page content budget counts what is held at once, not everything decoded: a form /// with 2 MiB of content painted three hundred times (600 MiB in all) draws every time. #[test] diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index abc7e8ceeb..d8918e5bd6 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -58,7 +58,7 @@ milestone = "M1" status = "shipped" commands = [] tools = [] -evidence = ["crates/ui-egui/tests/it/ui.rs::garbage_input_is_rejected_without_panicking", "crates/cos/src/parser.rs::garbage_never_panics", "crates/render/src/raster/tests.rs::missing_page_and_garbage_input_fail_gracefully"] +evidence = ["crates/ui-egui/tests/it/ui.rs::garbage_input_is_rejected_without_panicking", "crates/cos/src/parser.rs::garbage_never_panics", "crates/render/src/raster/tests.rs::missing_page_and_garbage_input_fail_gracefully", "crates/render/src/raster/tests.rs::a_form_that_draws_itself_is_drawn_once"] notes = "fuzzed nightly (M1.10)" [[feature]] diff --git a/vendor/README.md b/vendor/README.md index aa8b43ba4a..4733eba49c 100644 --- a/vendor/README.md +++ b/vendor/README.md @@ -39,7 +39,7 @@ Other resource categories retain their existing parent-chain lookup. | Crate | Version | Licence | Patches | Test | |---|---|---|---|---| -| hayro-interpret | 0.7.0 | Apache-2.0 OR MIT | (1) select `/AP /N` appearance states via `/AS` (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to `/Encoding` glyph names, then printable ASCII; (3) `Type3Glyph::advance_width`; (4) `InterpreterSettings::ocg_overrides` for viewer layer toggles; (5) soft masks whose group lacks `/CS` were dropped (Chrome gradient text); (6) paint nesting cap `MAX_PAINT_NESTING` for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in `/W` and `/W2` clamped to `MAX_CID` (a `0 4294967295 w` range hung); (8) `InterpreterSettings::hide_comments` skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) `image_size_ok`: images and masks over 2^28 pixels are not decoded (a fuzzed `/W 4294967295` stencil mask allocated 4 GiB); (10) `MAX_NESTED_PAINTS`: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (`Charset::glyph_id` returns `Option`); (12) annotation appearances whose `/BBox` or `/Matrix` collapses to a line or a point are skipped (mapping them onto `/Rect` divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its `/QuadPoints`, `/C` and `/CA` with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font `/Widths` (`read_widths`): entries only for codes 0–255 (one entry per code below `/FirstChar` was allocated: `4294967295` asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a `/LastChar` below `/FirstChar` gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto `/Rect` corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from `/Rect` was drawn shifted); (16) a clip set by `W`/`W*` takes effect after whichever path-painting operator ends the path (`S`, `f`, `B`, … as well as `n`; ISO 32000-2 §8.5.4): it was applied only on `n`, so `re W* S` clipped nothing and the clip stayed pending for a later, unrelated `n`; (17) `InterpreterSettings::cancelled`: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) `MAX_PAGE_CONTENT`: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (`decode_content`; once it runs out, `InterpreterWarning::ContentTruncated` is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most `MAX_SAMPLE_ENTRIES` (2^21) table points and `MAX_SAMPLE_VALUES` (2^23) samples, read no further than `/Size` asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (`ShadingPattern::encode_within`, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (`TriangleSampler`), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most `MAX_MESH_PATCHES` (2^16) patches and `MAX_MESH_TRIANGLES` (2^20) triangles, a lattice needs two to 2^19 vertices per row (`/VerticesPerRow 0` pushed empty rows for ever; one huge row read the whole stream), and a page samples at most `MAX_PAGE_MESH_TRIANGLES` (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws `.notdef` for a CID of an Adobe character collection (Japan1, GB1, CNS1, Korea1) that it can't find by Unicode: hayro took the CID for a glyph index of the substitute, which numbers its glyphs its own way, so the character was drawn as an unrelated one (𠮷 as "Q" in BIZ UDPGothic); (22) a Type 0 font drawn with a substitute finds a character given with a variation selector (`/ToUnicode`, or Adobe-Japan1-UCS2, which maps the JIS X 0208 forms of the kanji JIS X 0213:2004 redrew that way: CID 1247 is 噂 U+E0100) through the substitute's variation sequences, else as the character alone: such a CID found no glyph and was drawn as an unrelated one; (23) backport of hayro 0.8's `gradient` module: `EncodedShadingPattern::as_svg_gradient` turns an axial or radial shading into stops of an sRGB gradient (within 0.01 per channel, a stop at every stitching-function bound via `Function::stitching_bounds`) when a native gradient draws it exactly (extended past the box drawn in; a radial one with no background colour); (24) `DecodedImageCache`: decoded image pixels shared across renders through the interpreter cache within a 64 MiB budget (an image-heavy page decoded its thousands of images again per tile and per scale), keyed by every decode input (`DecodedImageKey`: the stream's place and dictionary bytes, its object, mask/stencil/raster output, interpolation, decode target size, preview or final); images under a transfer function and inline images are not cached, an image over the budget is never kept, and a poisoned lock decodes uncached; (25) `Device::actual_text` hands the device a marked content sequence's `/ActualText` (inline or named in `/Properties`), so text extraction keeps the Thai, Indic and Arabic clusters Chrome marks this way (their shaped tone marks and vowels have no Unicode of their own); (26) non-embedded simple fonts drawn with a standard-font stand-in build the Latin Extended-A letters the stand-in lacks (`/Gbreve`, `/Idotaccent`, `/Scedilla`, `/aogonek`, `/ccaron`…) from the base letter and accent it has (`StandardKind::composite_glyph`, `ACCENTED_LETTERS`): PDFium's Foxit stand-ins hold only the Standard/WinAnsi/Mac Roman glyphs, so iText letters with Times New Roman unembedded showed blanks for Ğ, İ, Ş, ğ and ş; the comma-below letters (`/scommaaccent`, `/Tcommaaccent`, `/kcommaaccent`…, found by character, so `/uni0136` and the glyph list's `/Kcedilla` too) are the letter and the font's comma, smaller and under the baseline, since the stand-ins have no `commaaccent` (Romanian ș drew nothing and ț a cedilla in forms laid out from XFA; ģ, whose comma is turned and above, is not built); (27) an embedded TrueType font without a base encoding reads its codes as WinAnsi for text extraction (`EmbeddedKind::code_to_unicode_name`), matching pdftotext's output, instead of Standard: OCR text layers of scans write WinAnsi codes with no `/Encoding`, and Standard has no ü, ö or ß and reads é as Ø (glyph selection unchanged); (28) an image whose pixels hold at most 16 bits of samples (an 8-bit Separation or DeviceN image, a 1-, 2- or 4-bit gray or indexed one) is converted once per possible pixel value through a lookup table instead of once per pixel (`rgb_via_lookup`): a spot-colour tint transform is often a PostScript calculator program, run for every pixel, so a fillable character sheet's 3304 × 4105 DeviceN background took four seconds and 430 MB to draw a page; each entry is converted on its own (`convert_each`), so a tint transform giving the wrong number of components for a value the image never uses cannot shift the colours of the values it does; (29) `Device::begin_annotations`/`end_annotations` bracket the page's annotations, so a device can tell comment paints from document paints (reading colours recolour the document, never a comment, #1548); (30) the Annex D.2 alias codes read as the ink they draw: WinAnsiEncoding 0xA0 and 0xAD, and MacRomanEncoding 0xCA, carry the glyph names `space` and `hyphen` -- the names those tables also give 0x20 and 0x2D -- not `nbspace` and `sfthyphen`, whose glyph list values U+00A0 and U+00AD are invisible formatting characters. Extraction hid drawn ink behind them, so search and redaction matched neither (`XiaoBiaoSong.pdf` in the pdf.js corpus drew 104 soft hyphens; across the 989 files the soft hyphens we extract fall from 117 to 13 and the non-breaking spaces from 30 to 26, what remains coming from `/ToUnicode` maps, which say so explicitly). The base table only, in the text direction only: a `/Differences` name keeps its glyph list value and glyph lookup is untouched, both as pdftotext does; (31) a Type 0 sampled function keeps its samples in one flat array indexed by strides (`Type0::entry`) instead of a hash map keyed by each grid point: the map was built again, one hashed insert per sample, every time a shading using the function was painted (a 2,097,152-sample ramp painted 30 times took 38 s; `bug1721218` in the pdf.js corpus spent most of a 1.5 s page on it); the interpolation is unchanged, so results are bit for bit the same, a sample missing from a short stream still gives no value, and the maximum sample value is computed without overflowing at 32 bits per sample; (32) the decoded-image cache (24) keys a decode by the size the image is drawn at only when a JPX stream decodes it or its soft mask or mask (`decode_reads_target`): JPX is the only filter that decodes a reduced resolution for a size, so a JPEG or Flate scan was decoded again at every zoom step and kept once per zoom level; one decode now serves every size; `InterpreterCache::decoded_image_decodes` counts the cache's decodes; (33) Coons and tensor-product patches copy their vertices' colours less: each point of a patch's 20 × 20 triangle grid is interpolated once (`generate_patch_triangles`; each cell interpolated its four corners itself, 1,444 interpolations a patch for 400 points), a cell's second triangle takes the vertices its first copied, and the mesh sampler moves a patch's triangles into device space instead of copying each vertex (`TriangleSampler::sample_owned`); the same values, 10–16% faster on mesh-gradient pages; (34) `color::tint_transform_runs`: how many colours the calling thread has run through a Separation or DeviceN tint transform (counted once per batch in `convert_f32`), so a test counts patch (28)'s once-per-sample-value runs instead of timing them (its 10 s deadline failed on a busy FreeBSD VM in CI); (35) `EncodedShadingPattern::extent`: where a mesh shading (types 4–7) can paint at all, the pixels its triangles touched when it was encoded (`None` for the types that paint everywhere, and when the shading has a `/Background`), so the renderer samples a texture there alone and skips a mesh with nothing on the canvas; (36) a mesh shading's pixel whose corner no triangle holds strictly (on an edge two triangles share, or on the mesh's side) takes the triangle that covers its centre, edges included (`Triangle::covers`): testing only the corner left a white seam along every shared edge, and the colour is still taken at the corner; (38) the `/DefaultGray`, `/DefaultRGB` and `/DefaultCMYK` colour spaces of a resource dictionary (ISO 32000-2 §8.6.5.6) replace the device space content selects: with `g`/`G`, `rg`/`RG`, `k`/`K` and `cs`/`CS` (by name, or through a `/ColorSpace` resource that is a device space), and as the `/ColorSpace` of an image XObject, an inline image or a shading painted by `sh` or as a shading pattern (`DefaultColorSpaces`, looked up once per content window in the resources and those they inherit, so a form XObject without a default of its own draws in its page's, as Poppler and MuPDF draw it); a default that is a Lab, Indexed, Pattern or device space, has another number of components or is no colour space is ignored and device colour is drawn as before; device spaces inside other spaces (an Indexed base, a Separation or DeviceN alternate, a Pattern's underlying space), shading dictionaries, masks and transparency group spaces are unchanged, and the decoded-image cache (24) keys a decode by the default it was drawn in. hayro drew device colour as it was, so pages that calibrate it this way (common in PDF/A) came out in the wrong colours; (39) knockout groups (#2971): `Device::set_knockout` right after a `/K true` group is pushed; `Device::knockout_image` for an unmasked image before its group is pushed, when `Device::in_knockout_group`: the device clears its placement rectangle only if it will draw the image (on the canvas, not a preview's tiny image, and it decodes), decoded at its draw size so the draw reuses that decode; and, when `Device::knockout_erase_wanted`, a nested group's content drawn once between `begin_knockout_erase` and `end_knockout_erase` before the group itself, which then always gets a layer of its own; (40) an optional content membership dictionary's visibility expression (`/VE`: `/And`, `/Or`, `/Not`, nested up to 32 deep) decides its visibility ahead of `/OCGs` and `/P` (ISO 32000-2 §8.11.2.2); it was ignored, so content under it stayed visible with its layers hidden; a malformed expression, including one naming an object the document doesn't list in `/OCProperties /OCGs`, falls back to `/OCGs` and `/P` (#2986); (41) a transparency group that is neither isolated nor knockout, entered at full opacity with no soft mask and Normal blending, is drawn straight onto the page instead of into a layer of its own (§11.4.7: its elements composite with the backdrop; the layer made a `/BM /Multiply` highlight appearance an opaque block, #1968); (42) text in a Type 3 font is drawn filled whatever the text rendering mode, except mode 3 (invisible): only that mode has any effect on Type 3 fonts (ISO 32000-1 §9.3.6). Stroking ran the glyph procedure a second time, so `2 Tr` made bitmap fonts look bold, `1 Tr` painted an uncoloured glyph in the stroking colour, and `7 Tr` hid it (`7 Tr` since follows 64); (43) Lab colour spaces honour their own `/WhitePoint`: L\*a\*b\* goes to XYZ relative to that white, is Bradford-adapted to D65 and converted to sRGB (ISO 32000-2 §8.6.5.4) instead of through `LAB.icc` as if every white were D50 (a D65 blue came out 45 levels too red; the profile is removed); components normalised to [0, 1] (`manual_scale`, an `/Indexed` palette over Lab) map onto L\* 0..100 and the a\*/b\* `/Range`; (44) a Type 0 font whose `/ToUnicode` is the name `/Identity-H` or `/Identity-V` (not a CMap stream; WSCAD writes it) reads each code as its Unicode value, as Acrobat shows it (`Type0Font::identity_to_unicode`); the identity CMap the name resolved to maps codes to CIDs only, so such pages had no text to select or find; (46) an image XObject's group applies the fill opacity, so the fill alpha is reset to 1 before a stencil mask takes the fill colour as its paint, and a pattern paint drops the opacity it captured at `scn` (`Pattern::without_opacity`) (it was applied twice: a mask at `ca 0.5` was drawn at a quarter, #2975); (47) `path_as_rect`: a five-element path is a rectangle only when its four points are distinct, each step from one to the next runs along a single axis, and the last returns to the first. The length test admits one MoveTo and four LineTo as well as one MoveTo, three LineTo and a ClosePath, and asking only that every corner of the bounding box was visited says nothing about the order or the steps between, so a five-point polyline whose ends reached the corners was painted as that box -- a fourth side the path does not have -- and a clip of that shape was discarded; the coordinates are compared exactly, since the matrix is applied after this check and a `cm` of 10,000 turns `is_nearly_equal`'s 256th of a unit into tens of device units (#3327); (48) an `/ICCBased` colour space is cached by what its profile is read from (`IccProfileKey`): the stream object it names, looked up without reading the stream, and then the stream's dictionary and its own bytes (decrypted, not decoded), so a byte-for-byte copy of a profile in another object is read once, and a profile stream written directly in its array, which reports the object holding the array as every other stream written there does, is told apart by its bytes. hayro keyed it by the stream dictionary's bytes alone, so two different profiles with the same `/N` and `/Length` shared whichever was read first, and colours in the second were drawn through the first. `InterpreterCache::icc_profile_reads` counts the profiles read; (49) `OpenTypeFontBlob::new` also accepts a `post` table version 1.0, 2.0 or 3.0 written with its 16-bit halves swapped (`00 00 00 03`), as ScanSoft PDF Create! embeds every font; such fonts were refused and their text drawn with a substitute (Hebrew came out garbled or missing, #2946); `glyph_names` reads such a table from a copy with the version written correctly, so a 2.0 table's names serve `/Differences`. Any other unknown version is still refused (pdf.js 9462); (50) `CffFontBlob::glyph_index_by_name` looks a glyph name up in a name to glyph map built on first use (shared between clones) instead of searching the whole charset, with a string lookup per glyph, for every character code of a font; the first glyph with a name still wins; (51) at most 4096 graphics states saved at once, holding at most 64 MiB together (each `q` copies the whole state, dash array and clip list included, so short operators multiplied one large state); a `q` past either limit is counted but not copied, and its `Q` still removes the clips added since (from storytold/craft-libs#239); (52) a `/W2` run whose first CID is past `MAX_CID` stops there, as in `/W` (7), checked before the run's metrics are read: a first CID of 4294967295 inserted one entry and then overflowed, discarding the whole `/W2` table (from storytold/craft-libs#254); (55) a Separation or DeviceN tint transform that cannot be read keeps its colour space, and one that cannot be read or evaluated paints the alternate space's initial colour, as evaluation failures already did; that fallback, and a `cs`/`CS` colour space that cannot be read (drawn in DeviceGray as before), are now reported once per content stream as `InterpreterWarning::ColorFallback` (`RenderWarning::ColourFallback` in `pdfcraft-render`) instead of silently (#4260); (56) an annotation whose Invisible flag (bit 1) is set and whose `/Subtype` is none of the standard types of ISO 32000-2 Table 171 (`is_standard_annotation_type`; no `/Subtype` counts as unknown) is not drawn (Table 167), and a `/Popup` annotation's appearance is never painted on the page (§12.5.6.14: a popup has no appearance stream of its own; the viewer shows it as a window for its parent annotation). Both were drawn from their `/AP`; (57) a simple font drawn with a standard-font stand-in because the document names a font it doesn't embed (`StandardKind` with `fallback`) narrows a glyph to a narrower `/Widths` advance but no longer widens it to a wider one: the advance already places the next glyph (ISO 32000-2 §9.2.4; /Widths in Table 109), and widening drew a non-embedded `Arial,Bold` with 600-unit widths as a heavy fixed-pitch face whose `i` and `l` were solid blocks twice their width; an exact standard-14 font still matches its standard metrics both ways; (58) the fallback request for a Type 0 font that is not embedded (`FallbackFontQuery::with_descriptor`) takes its serif, italic and small-cap flags, `/FontWeight`, `/FontFamily` and `/FontStretch` from the descendant CIDFont's font descriptor (ISO 32000-2 §9.7.4.1, §9.7.6.1; §9.8): it read them from the Type 0 dictionary, which has none, so every such request asked for a regular sans-serif face and no resolver could tell a Song or Mincho font from a Hei or Gothic one, or see `/FontWeight 700`; (59) the NoRotate flag (`/F` bit 5) is honoured: the appearance, fitted to `/Rect`, is turned against the page's `/Rotate` about the rectangle's upper-left corner, which stays where it is, so it reads upright as the page is displayed (ISO 32000-2 §12.5.3). Sticky notes, which are saved NoRotate, were drawn turned with the page (#2583). Form fields and links stay in `/Rect`, where PdfCraft takes their click targets from it; (64) text in a Type 3 font follows ISO 32000-2 §9.3.6 in the clipping modes: under `7 Tr` it is not rendered (reported to the device as clipping text, `GlyphDrawMode::Clip`, so extraction keeps it), and in modes 4 to 7 a Type 3 glyph adds nothing to the clipping path (`clip_glyph`), so text shown only in a Type 3 font leaves no clip at `ET`. Since (42) `7 Tr` painted the glyphs in the fill colour (black by default), so a fill meant to follow showed them wherever it didn't cover them; 4 to 6 still render filled; (66) a Separation colour space whose colourant is the special name `All` paints every colourant at once (ISO 32000-2 §8.6.6.4): on screen and in an RGB print raster its tint t is the gray 1 − t, complemented because the device is additive, whatever its alternate space and tint transform say (hayro drew it through the tint transform, so a `/DeviceCMYK` transform giving cyan drew registration marks light cyan); `/None` still paints nothing, any other name still goes through its alternate, and a DeviceN space naming `All` keeps its alternate, since the clause gives the name that meaning in a Separation space; (63) a luminosity soft mask whose group colour space is `/DeviceCMYK` (by name, not an ICC-based or calibrated space) reads a `/DeviceCMYK` fill or stroke colour drawn in it, and its `/BC`, as their gray `1 - min(1, 0.3 c + 0.59 m + 0.11 y + k)` (ISO 32000-2 §10.4.2.3, §11.5.3: device colour converts to gray with no colour calibration; `Context::luminosity_in_device_cmyk`, `color::device_cmyk_to_gray`): it went through the CMYK display profile, where 100% K is a dark grey, so a 100% K mask let about an eighth of the masked paint through. CMYK images, shadings and tiling patterns inside such a mask, and CMYK on the page, still use the profile; (71) a lookup table in an indirect string reads: it was read as a stream first, which uses up a reference (ISO 32000-2 §8.6.6.3); (72) an uncoloured pattern's underlying space given as a name or a reference is used: it was skipped, so a CMYK cyan drew red (ISO 32000-2 §8.7.3.3); (73) `InterpreterSettings::jbig2_work`, set for each page in `interpret_page` and ended with it, and a JBIG2 image or an image with a JBIG2 mask past it left out with `ContentTruncated`; (74) colour spaces nest at most 32 levels, through every way one names another, so a chain of them ends instead of overflowing the stack; an ICC profile's `/Alternate` is read where it's met and isn't kept with the profile, or a page could build a chain past the limit on spaces already read; (75) a soft hyphen (U+00AD) that a CFF substitute for a CID font has no glyph for draws its hyphen, as Annex D.2 reads WinAnsi 0xAD in (30), instead of the CID taken as a glyph index (#5888) | (1) `raster::tests::widget_appearance_states`; (2, 3) `xtask text-oracle` on corpus/pdfjs issue918; (4) `raster::tests::layer_override_hides_content`; (5) `raster::tests::soft_mask_without_group_colour_space_is_applied`; (6) `raster::tests::self_referencing_tiling_pattern_terminates`, `raster::tests::self_referencing_type3_glyph_terminates`; (7) `raster::tests::huge_cid_width_ranges_terminate`; (8) `raster::tests::hiding_comments_keeps_fields`; (9) `raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding`; (10) `raster::tests::type3_glyphs_that_fan_out_into_themselves_finish`; (11) `raster::tests::type1_font_programs_with_long_runs_of_numbers_load_quickly`; (12) `raster::tests::degenerate_annotation_appearances_are_skipped`; (13) `raster::tests::highlights_without_an_appearance_are_drawn`; (14) `raster::tests::malformed_simple_font_widths_keep_later_widths_and_the_font`, `raster::tests::huge_simple_font_first_char_terminates`; (15) `raster::tests::annotation_appearance_box_maps_onto_rect`; (16) `raster::tests::clipping_path_applies_after_any_painting_operator`; (17) `raster::tests::dropping_the_pool_cancels_its_renders`, `raster::tests::the_watchdog_stops_the_render_it_gives_up_on`; (18) `raster::tests::page_contents_and_nested_forms_share_one_budget`, `raster::tests::content_painted_again_and_again_keeps_drawing`, `raster::tests::sampled_functions_with_billions_of_samples_are_refused`; (19) `raster::tests::mesh_shadings_far_off_the_page_render`, `raster::tests::many_overlapping_mesh_triangles_finish`, `raster::tests::mesh_shadings_with_many_patches_or_no_rows_finish`; (20) `raster::tests::function_limits_calculator_arithmetic_and_depth`, `raster::tests::function_limits_public_construction_work_and_depth`, `raster::tests::function_limits_stitching_cycles_are_refused_and_shared_children_work`, `raster::tests::function_limits_invalid_calculator_keeps_the_page_renderable`; (21) `raster::tests::cids_a_substitute_lacks_are_not_drawn_as_other_glyphs`; (22) `raster::tests::variation_sequences_draw_their_kanji_in_a_substitute`; (23) `raster::tests::axial_and_radial_shadings_draw_as_native_gradients`; (24) `raster::tests::tiles_match_full_render`, `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel` (both decode through the cache), `raster::tests::decoded_images_with_identical_dictionaries_keep_their_own_pixels`, `raster::tests::decoded_images_under_a_transfer_function_are_not_shared`; (25) `text::tests::actual_text_replaces_the_glyphs_it_marks_in_every_writing_direction`, `text::tests::actual_text_follows_the_pdf_rules_for_marked_content`; (26) `raster::tests::accented_letters_the_standard_stand_ins_lack_are_built_from_their_parts`, `raster::tests::comma_below_letters_in_standard_fonts_draw_a_comma_under_the_letter`; (27) `text::tests::truetype_without_an_encoding_reads_its_codes_as_win_ansi`; (28) `raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value`, `raster::tests::images_converted_through_a_table_keep_every_samples_colour`, `raster::tests::a_malformed_tint_for_an_unused_sample_keeps_the_used_colours`; (29) `raster::tests::reading_colors_keep_highlight_annotations_readable`; (30) `text::tests::the_annex_d2_alias_codes_read_as_an_ordinary_hyphen_and_space`; (31) `raster::tests::shadings_with_large_sampled_functions_paint_quickly`, `raster::tests::sampled_functions_interpolate_their_grid_as_before`; (32) `raster::tests::decoded_images_serve_every_size_they_are_drawn_at`; (33) `raster::tests::patch_meshes_keep_their_corner_colours`; (34) `raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value`; (35) `mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing}` (`cargo test -p hayro`); (36) `raster::tests::mesh_triangles_sharing_an_edge_leave_no_seam`; (38) `raster::tests::default_colour_spaces::device_rgb_operators_draw_in_default_rgb`, `raster::tests::default_colour_spaces::device_cmyk_operators_draw_in_default_cmyk`, `raster::tests::default_colour_spaces::device_gray_operators_draw_in_default_gray`, `raster::tests::default_colour_spaces::device_images_draw_in_the_default_colour_spaces`, `raster::tests::default_colour_spaces::device_shadings_draw_in_the_default_colour_spaces`, `raster::tests::default_colour_spaces::an_image_drawn_with_and_without_a_default_keeps_both_colours`, `raster::tests::default_colour_spaces::forms_inherit_the_default_colour_spaces_they_do_not_override`, `raster::tests::default_colour_spaces::unusable_defaults_leave_device_colour_alone`; (39) `raster::tests::knockout_groups_replace_what_their_elements_cover`; (40) `raster::tests::visibility_expressions_decide_what_a_membership_dictionary_shows`; (41) `raster::tests::a_multiply_highlight_inside_a_transparency_group_keeps_the_text`; (42) `raster::tests::rendering_modes_but_3_and_7_draw_type3_glyphs_as_filled`; (43) `raster::tests::lab_honours_the_space_white_point`, `raster::tests::indexed_palettes_over_lab_keep_their_colours`; (44) `text::tests::identity_to_unicode_names_give_text`; (46) `raster::tests::stencil_masks_are_drawn_at_the_fill_opacity_once`; (47) `raster::tests::a_five_point_polyline_is_not_painted_as_its_bounding_box`, `raster::tests::a_path_that_visits_every_corner_out_of_order_is_not_a_rectangle`, `raster::tests::a_rectangle_is_still_painted_as_a_rectangle`; (48) `raster::tests::icc_profile_cache::icc_profiles_with_identical_dictionaries_draw_their_own_colours`, `raster::tests::icc_profile_cache::icc_profiles_used_again_are_read_once`; (49) `raster::tests::a_font_whose_post_version_has_swapped_halves_is_still_used`; (50) `font::blob::tests::glyph_names_resolve_to_the_first_glyph_with_that_name`; (51) `context::saved_state_tests::repeated_saves_stay_bounded_and_balanced`; (52) `font::cid::tests::w2_run_past_the_largest_cid_keeps_the_rest_of_the_table` (hayro-interpret); (55) `raster::tests::devicen_inks::a_tint_transform_that_cannot_be_evaluated_draws_the_rest_of_the_page_and_warns`, `raster::tests::devicen_inks::a_tint_transform_that_cannot_be_read_falls_back_and_warns`; (56) `raster::tests::invisible_unknown_annotations_and_popups_paint_nothing`; (57) `substitute_widths::a_substitute_keeps_its_glyph_shapes_at_wider_advances`, `substitute_widths::a_substitute_still_narrows_glyphs_to_narrower_advances` (pdfcraft-render `tests/it`); (58) `cid_fallback_query::a_cid_font_fallback_request_carries_its_descendants_style`, `cid_fallback_query::a_cid_font_without_a_descriptor_asks_for_the_default_style` (pdfcraft-render integration tests); (59) `raster::tests::no_rotate_notes_stay_upright_at_their_corner_on_rotated_pages` (hayro-interpret); (64) `type3_text_clip::{clip_only_type3_text_is_not_rendered_and_leaves_no_clip, invisible_type3_text_is_not_rendered, other_modes_render_type3_glyphs_filled_and_never_clip, only_outline_glyphs_shape_a_clip_shared_with_type3_text, self_nesting_type3_text_in_a_clip_mode_stays_bounded}` (`cargo test -p pdfcraft-render --test it`), `raster::tests::rendering_modes_but_3_and_7_draw_type3_glyphs_as_filled`, `text::tests::clipping_mode_type3_text_is_kept_by_the_ocr_filter`; (66) `separation_all::{separation_all_fills_paint_the_complemented_tint_as_gray, separation_all_strokes_paint_the_complemented_tint_as_gray, separation_all_images_paint_each_sample_as_the_complemented_gray, separation_all_shadings_paint_the_complemented_gray, separation_none_paints_nothing, a_named_separation_still_paints_through_its_alternate, a_device_n_space_naming_all_still_paints_through_its_alternate}` (`cargo test -p pdfcraft-render --test it`); (63) `color::device_cmyk_gray_tests::{each_ink_darkens_by_its_iso_weight, ink_past_full_coverage_is_black_and_bad_components_read_as_zero}` (hayro-interpret), `it::luminosity_group_space::{black_ink_in_a_device_cmyk_group_masks_fully_and_no_ink_shows_fully, each_process_ink_in_a_device_cmyk_group_masks_by_its_weight, a_device_cmyk_backdrop_colour_masks_by_its_gray, gray_and_rgb_groups_mask_as_before, device_cmyk_on_the_page_keeps_its_display_colour}` (pdfcraft-render); (71) `raster::tests::an_indexed_lookup_table_is_read_in_every_legal_form` (pdfcraft-render); (72) `raster::tests::an_uncoloured_pattern_uses_its_underlying_space_in_every_form` (pdfcraft-render); (73) `raster::tests::the_jbig2_images_of_a_page_share_one_allowance`, `a_picture_whose_jbig2_mask_is_past_the_pages_allowance_is_not_drawn` and `a_pages_jbig2_allowance_ends_with_the_page` (pdfcraft-render); (74) `raster::tests::{a_chain_of_colour_spaces_ends, colour_spaces_are_read_to_the_limit_of_their_nesting, a_chain_built_on_spaces_already_read_ends, a_profile_read_at_the_limit_keeps_its_alternate_for_a_read_nearer_the_top}` (pdfcraft-render); (75) `raster::tests::a_soft_hyphen_a_substitute_lacks_draws_its_hyphen` (pdfcraft-render) | +| hayro-interpret | 0.7.0 | Apache-2.0 OR MIT | (1) select `/AP /N` appearance states via `/AS` (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to `/Encoding` glyph names, then printable ASCII; (3) `Type3Glyph::advance_width`; (4) `InterpreterSettings::ocg_overrides` for viewer layer toggles; (5) soft masks whose group lacks `/CS` were dropped (Chrome gradient text); (6) paint nesting cap `MAX_PAINT_NESTING` for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in `/W` and `/W2` clamped to `MAX_CID` (a `0 4294967295 w` range hung); (8) `InterpreterSettings::hide_comments` skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) `image_size_ok`: images and masks over 2^28 pixels are not decoded (a fuzzed `/W 4294967295` stencil mask allocated 4 GiB); (10) `MAX_NESTED_PAINTS`: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (`Charset::glyph_id` returns `Option`); (12) annotation appearances whose `/BBox` or `/Matrix` collapses to a line or a point are skipped (mapping them onto `/Rect` divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its `/QuadPoints`, `/C` and `/CA` with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font `/Widths` (`read_widths`): entries only for codes 0–255 (one entry per code below `/FirstChar` was allocated: `4294967295` asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a `/LastChar` below `/FirstChar` gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto `/Rect` corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from `/Rect` was drawn shifted); (16) a clip set by `W`/`W*` takes effect after whichever path-painting operator ends the path (`S`, `f`, `B`, … as well as `n`; ISO 32000-2 §8.5.4): it was applied only on `n`, so `re W* S` clipped nothing and the clip stayed pending for a later, unrelated `n`; (17) `InterpreterSettings::cancelled`: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) `MAX_PAGE_CONTENT`: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (`decode_content`; once it runs out, `InterpreterWarning::ContentTruncated` is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most `MAX_SAMPLE_ENTRIES` (2^21) table points and `MAX_SAMPLE_VALUES` (2^23) samples, read no further than `/Size` asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (`ShadingPattern::encode_within`, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (`TriangleSampler`), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most `MAX_MESH_PATCHES` (2^16) patches and `MAX_MESH_TRIANGLES` (2^20) triangles, a lattice needs two to 2^19 vertices per row (`/VerticesPerRow 0` pushed empty rows for ever; one huge row read the whole stream), and a page samples at most `MAX_PAGE_MESH_TRIANGLES` (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws `.notdef` for a CID of an Adobe character collection (Japan1, GB1, CNS1, Korea1) that it can't find by Unicode: hayro took the CID for a glyph index of the substitute, which numbers its glyphs its own way, so the character was drawn as an unrelated one (𠮷 as "Q" in BIZ UDPGothic); (22) a Type 0 font drawn with a substitute finds a character given with a variation selector (`/ToUnicode`, or Adobe-Japan1-UCS2, which maps the JIS X 0208 forms of the kanji JIS X 0213:2004 redrew that way: CID 1247 is 噂 U+E0100) through the substitute's variation sequences, else as the character alone: such a CID found no glyph and was drawn as an unrelated one; (23) backport of hayro 0.8's `gradient` module: `EncodedShadingPattern::as_svg_gradient` turns an axial or radial shading into stops of an sRGB gradient (within 0.01 per channel, a stop at every stitching-function bound via `Function::stitching_bounds`) when a native gradient draws it exactly (extended past the box drawn in; a radial one with no background colour); (24) `DecodedImageCache`: decoded image pixels shared across renders through the interpreter cache within a 64 MiB budget (an image-heavy page decoded its thousands of images again per tile and per scale), keyed by every decode input (`DecodedImageKey`: the stream's place and dictionary bytes, its object, mask/stencil/raster output, interpolation, decode target size, preview or final); images under a transfer function and inline images are not cached, an image over the budget is never kept, and a poisoned lock decodes uncached; (25) `Device::actual_text` hands the device a marked content sequence's `/ActualText` (inline or named in `/Properties`), so text extraction keeps the Thai, Indic and Arabic clusters Chrome marks this way (their shaped tone marks and vowels have no Unicode of their own); (26) non-embedded simple fonts drawn with a standard-font stand-in build the Latin Extended-A letters the stand-in lacks (`/Gbreve`, `/Idotaccent`, `/Scedilla`, `/aogonek`, `/ccaron`…) from the base letter and accent it has (`StandardKind::composite_glyph`, `ACCENTED_LETTERS`): PDFium's Foxit stand-ins hold only the Standard/WinAnsi/Mac Roman glyphs, so iText letters with Times New Roman unembedded showed blanks for Ğ, İ, Ş, ğ and ş; the comma-below letters (`/scommaaccent`, `/Tcommaaccent`, `/kcommaaccent`…, found by character, so `/uni0136` and the glyph list's `/Kcedilla` too) are the letter and the font's comma, smaller and under the baseline, since the stand-ins have no `commaaccent` (Romanian ș drew nothing and ț a cedilla in forms laid out from XFA; ģ, whose comma is turned and above, is not built); (27) an embedded TrueType font without a base encoding reads its codes as WinAnsi for text extraction (`EmbeddedKind::code_to_unicode_name`), matching pdftotext's output, instead of Standard: OCR text layers of scans write WinAnsi codes with no `/Encoding`, and Standard has no ü, ö or ß and reads é as Ø (glyph selection unchanged); (28) an image whose pixels hold at most 16 bits of samples (an 8-bit Separation or DeviceN image, a 1-, 2- or 4-bit gray or indexed one) is converted once per possible pixel value through a lookup table instead of once per pixel (`rgb_via_lookup`): a spot-colour tint transform is often a PostScript calculator program, run for every pixel, so a fillable character sheet's 3304 × 4105 DeviceN background took four seconds and 430 MB to draw a page; each entry is converted on its own (`convert_each`), so a tint transform giving the wrong number of components for a value the image never uses cannot shift the colours of the values it does; (29) `Device::begin_annotations`/`end_annotations` bracket the page's annotations, so a device can tell comment paints from document paints (reading colours recolour the document, never a comment, #1548); (30) the Annex D.2 alias codes read as the ink they draw: WinAnsiEncoding 0xA0 and 0xAD, and MacRomanEncoding 0xCA, carry the glyph names `space` and `hyphen` -- the names those tables also give 0x20 and 0x2D -- not `nbspace` and `sfthyphen`, whose glyph list values U+00A0 and U+00AD are invisible formatting characters. Extraction hid drawn ink behind them, so search and redaction matched neither (`XiaoBiaoSong.pdf` in the pdf.js corpus drew 104 soft hyphens; across the 989 files the soft hyphens we extract fall from 117 to 13 and the non-breaking spaces from 30 to 26, what remains coming from `/ToUnicode` maps, which say so explicitly). The base table only, in the text direction only: a `/Differences` name keeps its glyph list value and glyph lookup is untouched, both as pdftotext does; (31) a Type 0 sampled function keeps its samples in one flat array indexed by strides (`Type0::entry`) instead of a hash map keyed by each grid point: the map was built again, one hashed insert per sample, every time a shading using the function was painted (a 2,097,152-sample ramp painted 30 times took 38 s; `bug1721218` in the pdf.js corpus spent most of a 1.5 s page on it); the interpolation is unchanged, so results are bit for bit the same, a sample missing from a short stream still gives no value, and the maximum sample value is computed without overflowing at 32 bits per sample; (32) the decoded-image cache (24) keys a decode by the size the image is drawn at only when a JPX stream decodes it or its soft mask or mask (`decode_reads_target`): JPX is the only filter that decodes a reduced resolution for a size, so a JPEG or Flate scan was decoded again at every zoom step and kept once per zoom level; one decode now serves every size; `InterpreterCache::decoded_image_decodes` counts the cache's decodes; (33) Coons and tensor-product patches copy their vertices' colours less: each point of a patch's 20 × 20 triangle grid is interpolated once (`generate_patch_triangles`; each cell interpolated its four corners itself, 1,444 interpolations a patch for 400 points), a cell's second triangle takes the vertices its first copied, and the mesh sampler moves a patch's triangles into device space instead of copying each vertex (`TriangleSampler::sample_owned`); the same values, 10–16% faster on mesh-gradient pages; (34) `color::tint_transform_runs`: how many colours the calling thread has run through a Separation or DeviceN tint transform (counted once per batch in `convert_f32`), so a test counts patch (28)'s once-per-sample-value runs instead of timing them (its 10 s deadline failed on a busy FreeBSD VM in CI); (35) `EncodedShadingPattern::extent`: where a mesh shading (types 4–7) can paint at all, the pixels its triangles touched when it was encoded (`None` for the types that paint everywhere, and when the shading has a `/Background`), so the renderer samples a texture there alone and skips a mesh with nothing on the canvas; (36) a mesh shading's pixel whose corner no triangle holds strictly (on an edge two triangles share, or on the mesh's side) takes the triangle that covers its centre, edges included (`Triangle::covers`): testing only the corner left a white seam along every shared edge, and the colour is still taken at the corner; (38) the `/DefaultGray`, `/DefaultRGB` and `/DefaultCMYK` colour spaces of a resource dictionary (ISO 32000-2 §8.6.5.6) replace the device space content selects: with `g`/`G`, `rg`/`RG`, `k`/`K` and `cs`/`CS` (by name, or through a `/ColorSpace` resource that is a device space), and as the `/ColorSpace` of an image XObject, an inline image or a shading painted by `sh` or as a shading pattern (`DefaultColorSpaces`, looked up once per content window in the resources and those they inherit, so a form XObject without a default of its own draws in its page's, as Poppler and MuPDF draw it); a default that is a Lab, Indexed, Pattern or device space, has another number of components or is no colour space is ignored and device colour is drawn as before; device spaces inside other spaces (an Indexed base, a Separation or DeviceN alternate, a Pattern's underlying space), shading dictionaries, masks and transparency group spaces are unchanged, and the decoded-image cache (24) keys a decode by the default it was drawn in. hayro drew device colour as it was, so pages that calibrate it this way (common in PDF/A) came out in the wrong colours; (39) knockout groups (#2971): `Device::set_knockout` right after a `/K true` group is pushed; `Device::knockout_image` for an unmasked image before its group is pushed, when `Device::in_knockout_group`: the device clears its placement rectangle only if it will draw the image (on the canvas, not a preview's tiny image, and it decodes), decoded at its draw size so the draw reuses that decode; and, when `Device::knockout_erase_wanted`, a nested group's content drawn once between `begin_knockout_erase` and `end_knockout_erase` before the group itself, which then always gets a layer of its own; (40) an optional content membership dictionary's visibility expression (`/VE`: `/And`, `/Or`, `/Not`, nested up to 32 deep) decides its visibility ahead of `/OCGs` and `/P` (ISO 32000-2 §8.11.2.2); it was ignored, so content under it stayed visible with its layers hidden; a malformed expression, including one naming an object the document doesn't list in `/OCProperties /OCGs`, falls back to `/OCGs` and `/P` (#2986); (41) a transparency group that is neither isolated nor knockout, entered at full opacity with no soft mask and Normal blending, is drawn straight onto the page instead of into a layer of its own (§11.4.7: its elements composite with the backdrop; the layer made a `/BM /Multiply` highlight appearance an opaque block, #1968); (42) text in a Type 3 font is drawn filled whatever the text rendering mode, except mode 3 (invisible): only that mode has any effect on Type 3 fonts (ISO 32000-1 §9.3.6). Stroking ran the glyph procedure a second time, so `2 Tr` made bitmap fonts look bold, `1 Tr` painted an uncoloured glyph in the stroking colour, and `7 Tr` hid it (`7 Tr` since follows 64); (43) Lab colour spaces honour their own `/WhitePoint`: L\*a\*b\* goes to XYZ relative to that white, is Bradford-adapted to D65 and converted to sRGB (ISO 32000-2 §8.6.5.4) instead of through `LAB.icc` as if every white were D50 (a D65 blue came out 45 levels too red; the profile is removed); components normalised to [0, 1] (`manual_scale`, an `/Indexed` palette over Lab) map onto L\* 0..100 and the a\*/b\* `/Range`; (44) a Type 0 font whose `/ToUnicode` is the name `/Identity-H` or `/Identity-V` (not a CMap stream; WSCAD writes it) reads each code as its Unicode value, as Acrobat shows it (`Type0Font::identity_to_unicode`); the identity CMap the name resolved to maps codes to CIDs only, so such pages had no text to select or find; (46) an image XObject's group applies the fill opacity, so the fill alpha is reset to 1 before a stencil mask takes the fill colour as its paint, and a pattern paint drops the opacity it captured at `scn` (`Pattern::without_opacity`) (it was applied twice: a mask at `ca 0.5` was drawn at a quarter, #2975); (47) `path_as_rect`: a five-element path is a rectangle only when its four points are distinct, each step from one to the next runs along a single axis, and the last returns to the first. The length test admits one MoveTo and four LineTo as well as one MoveTo, three LineTo and a ClosePath, and asking only that every corner of the bounding box was visited says nothing about the order or the steps between, so a five-point polyline whose ends reached the corners was painted as that box -- a fourth side the path does not have -- and a clip of that shape was discarded; the coordinates are compared exactly, since the matrix is applied after this check and a `cm` of 10,000 turns `is_nearly_equal`'s 256th of a unit into tens of device units (#3327); (48) an `/ICCBased` colour space is cached by what its profile is read from (`IccProfileKey`): the stream object it names, looked up without reading the stream, and then the stream's dictionary and its own bytes (decrypted, not decoded), so a byte-for-byte copy of a profile in another object is read once, and a profile stream written directly in its array, which reports the object holding the array as every other stream written there does, is told apart by its bytes. hayro keyed it by the stream dictionary's bytes alone, so two different profiles with the same `/N` and `/Length` shared whichever was read first, and colours in the second were drawn through the first. `InterpreterCache::icc_profile_reads` counts the profiles read; (49) `OpenTypeFontBlob::new` also accepts a `post` table version 1.0, 2.0 or 3.0 written with its 16-bit halves swapped (`00 00 00 03`), as ScanSoft PDF Create! embeds every font; such fonts were refused and their text drawn with a substitute (Hebrew came out garbled or missing, #2946); `glyph_names` reads such a table from a copy with the version written correctly, so a 2.0 table's names serve `/Differences`. Any other unknown version is still refused (pdf.js 9462); (50) `CffFontBlob::glyph_index_by_name` looks a glyph name up in a name to glyph map built on first use (shared between clones) instead of searching the whole charset, with a string lookup per glyph, for every character code of a font; the first glyph with a name still wins; (51) at most 4096 graphics states saved at once, holding at most 64 MiB together (each `q` copies the whole state, dash array and clip list included, so short operators multiplied one large state); a `q` past either limit is counted but not copied, and its `Q` still removes the clips added since (from storytold/craft-libs#239); (52) a `/W2` run whose first CID is past `MAX_CID` stops there, as in `/W` (7), checked before the run's metrics are read: a first CID of 4294967295 inserted one entry and then overflowed, discarding the whole `/W2` table (from storytold/craft-libs#254); (55) a Separation or DeviceN tint transform that cannot be read keeps its colour space, and one that cannot be read or evaluated paints the alternate space's initial colour, as evaluation failures already did; that fallback, and a `cs`/`CS` colour space that cannot be read (drawn in DeviceGray as before), are now reported once per content stream as `InterpreterWarning::ColorFallback` (`RenderWarning::ColourFallback` in `pdfcraft-render`) instead of silently (#4260); (56) an annotation whose Invisible flag (bit 1) is set and whose `/Subtype` is none of the standard types of ISO 32000-2 Table 171 (`is_standard_annotation_type`; no `/Subtype` counts as unknown) is not drawn (Table 167), and a `/Popup` annotation's appearance is never painted on the page (§12.5.6.14: a popup has no appearance stream of its own; the viewer shows it as a window for its parent annotation). Both were drawn from their `/AP`; (57) a simple font drawn with a standard-font stand-in because the document names a font it doesn't embed (`StandardKind` with `fallback`) narrows a glyph to a narrower `/Widths` advance but no longer widens it to a wider one: the advance already places the next glyph (ISO 32000-2 §9.2.4; /Widths in Table 109), and widening drew a non-embedded `Arial,Bold` with 600-unit widths as a heavy fixed-pitch face whose `i` and `l` were solid blocks twice their width; an exact standard-14 font still matches its standard metrics both ways; (58) the fallback request for a Type 0 font that is not embedded (`FallbackFontQuery::with_descriptor`) takes its serif, italic and small-cap flags, `/FontWeight`, `/FontFamily` and `/FontStretch` from the descendant CIDFont's font descriptor (ISO 32000-2 §9.7.4.1, §9.7.6.1; §9.8): it read them from the Type 0 dictionary, which has none, so every such request asked for a regular sans-serif face and no resolver could tell a Song or Mincho font from a Hei or Gothic one, or see `/FontWeight 700`; (59) the NoRotate flag (`/F` bit 5) is honoured: the appearance, fitted to `/Rect`, is turned against the page's `/Rotate` about the rectangle's upper-left corner, which stays where it is, so it reads upright as the page is displayed (ISO 32000-2 §12.5.3). Sticky notes, which are saved NoRotate, were drawn turned with the page (#2583). Form fields and links stay in `/Rect`, where PdfCraft takes their click targets from it; (64) text in a Type 3 font follows ISO 32000-2 §9.3.6 in the clipping modes: under `7 Tr` it is not rendered (reported to the device as clipping text, `GlyphDrawMode::Clip`, so extraction keeps it), and in modes 4 to 7 a Type 3 glyph adds nothing to the clipping path (`clip_glyph`), so text shown only in a Type 3 font leaves no clip at `ET`. Since (42) `7 Tr` painted the glyphs in the fill colour (black by default), so a fill meant to follow showed them wherever it didn't cover them; 4 to 6 still render filled; (66) a Separation colour space whose colourant is the special name `All` paints every colourant at once (ISO 32000-2 §8.6.6.4): on screen and in an RGB print raster its tint t is the gray 1 − t, complemented because the device is additive, whatever its alternate space and tint transform say (hayro drew it through the tint transform, so a `/DeviceCMYK` transform giving cyan drew registration marks light cyan); `/None` still paints nothing, any other name still goes through its alternate, and a DeviceN space naming `All` keeps its alternate, since the clause gives the name that meaning in a Separation space; (63) a luminosity soft mask whose group colour space is `/DeviceCMYK` (by name, not an ICC-based or calibrated space) reads a `/DeviceCMYK` fill or stroke colour drawn in it, and its `/BC`, as their gray `1 - min(1, 0.3 c + 0.59 m + 0.11 y + k)` (ISO 32000-2 §10.4.2.3, §11.5.3: device colour converts to gray with no colour calibration; `Context::luminosity_in_device_cmyk`, `color::device_cmyk_to_gray`): it went through the CMYK display profile, where 100% K is a dark grey, so a 100% K mask let about an eighth of the masked paint through. CMYK images, shadings and tiling patterns inside such a mask, and CMYK on the page, still use the profile; (71) a lookup table in an indirect string reads: it was read as a stream first, which uses up a reference (ISO 32000-2 §8.6.6.3); (72) an uncoloured pattern's underlying space given as a name or a reference is used: it was skipped, so a CMYK cyan drew red (ISO 32000-2 §8.7.3.3); (73) `InterpreterSettings::jbig2_work`, set for each page in `interpret_page` and ended with it, and a JBIG2 image or an image with a JBIG2 mask past it left out with `ContentTruncated`; (74) colour spaces nest at most 32 levels, through every way one names another, so a chain of them ends instead of overflowing the stack; an ICC profile's `/Alternate` is read where it's met and isn't kept with the profile, or a page could build a chain past the limit on spaces already read; (75) a soft hyphen (U+00AD) that a CFF substitute for a CID font has no glyph for draws its hyphen, as Annex D.2 reads WinAnsi 0xAD in (30), instead of the CID taken as a glyph index (#5888); (89) a form XObject already being drawn is not drawn inside itself (`Context::enter_form`, reported as `ContentTruncated`, §8.10.1): a name in its content that reached it again, through its own resources or the page resources a form with its own still falls back on, drew it until the page's nested-paint budget was spent (PDFBOX-5782, 38 s); the forms being drawn are kept per page and thread, so the tiling patterns, Type 3 glyphs, soft masks and groups painted inside a form, which are interpreted in contexts of their own, see it too | (1) `raster::tests::widget_appearance_states`; (2, 3) `xtask text-oracle` on corpus/pdfjs issue918; (4) `raster::tests::layer_override_hides_content`; (5) `raster::tests::soft_mask_without_group_colour_space_is_applied`; (6) `raster::tests::self_referencing_tiling_pattern_terminates`, `raster::tests::self_referencing_type3_glyph_terminates`; (7) `raster::tests::huge_cid_width_ranges_terminate`; (8) `raster::tests::hiding_comments_keeps_fields`; (9) `raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding`; (10) `raster::tests::type3_glyphs_that_fan_out_into_themselves_finish`; (11) `raster::tests::type1_font_programs_with_long_runs_of_numbers_load_quickly`; (12) `raster::tests::degenerate_annotation_appearances_are_skipped`; (13) `raster::tests::highlights_without_an_appearance_are_drawn`; (14) `raster::tests::malformed_simple_font_widths_keep_later_widths_and_the_font`, `raster::tests::huge_simple_font_first_char_terminates`; (15) `raster::tests::annotation_appearance_box_maps_onto_rect`; (16) `raster::tests::clipping_path_applies_after_any_painting_operator`; (17) `raster::tests::dropping_the_pool_cancels_its_renders`, `raster::tests::the_watchdog_stops_the_render_it_gives_up_on`; (18) `raster::tests::page_contents_and_nested_forms_share_one_budget`, `raster::tests::content_painted_again_and_again_keeps_drawing`, `raster::tests::sampled_functions_with_billions_of_samples_are_refused`; (19) `raster::tests::mesh_shadings_far_off_the_page_render`, `raster::tests::many_overlapping_mesh_triangles_finish`, `raster::tests::mesh_shadings_with_many_patches_or_no_rows_finish`; (20) `raster::tests::function_limits_calculator_arithmetic_and_depth`, `raster::tests::function_limits_public_construction_work_and_depth`, `raster::tests::function_limits_stitching_cycles_are_refused_and_shared_children_work`, `raster::tests::function_limits_invalid_calculator_keeps_the_page_renderable`; (21) `raster::tests::cids_a_substitute_lacks_are_not_drawn_as_other_glyphs`; (22) `raster::tests::variation_sequences_draw_their_kanji_in_a_substitute`; (23) `raster::tests::axial_and_radial_shadings_draw_as_native_gradients`; (24) `raster::tests::tiles_match_full_render`, `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel` (both decode through the cache), `raster::tests::decoded_images_with_identical_dictionaries_keep_their_own_pixels`, `raster::tests::decoded_images_under_a_transfer_function_are_not_shared`; (25) `text::tests::actual_text_replaces_the_glyphs_it_marks_in_every_writing_direction`, `text::tests::actual_text_follows_the_pdf_rules_for_marked_content`; (26) `raster::tests::accented_letters_the_standard_stand_ins_lack_are_built_from_their_parts`, `raster::tests::comma_below_letters_in_standard_fonts_draw_a_comma_under_the_letter`; (27) `text::tests::truetype_without_an_encoding_reads_its_codes_as_win_ansi`; (28) `raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value`, `raster::tests::images_converted_through_a_table_keep_every_samples_colour`, `raster::tests::a_malformed_tint_for_an_unused_sample_keeps_the_used_colours`; (29) `raster::tests::reading_colors_keep_highlight_annotations_readable`; (30) `text::tests::the_annex_d2_alias_codes_read_as_an_ordinary_hyphen_and_space`; (31) `raster::tests::shadings_with_large_sampled_functions_paint_quickly`, `raster::tests::sampled_functions_interpolate_their_grid_as_before`; (32) `raster::tests::decoded_images_serve_every_size_they_are_drawn_at`; (33) `raster::tests::patch_meshes_keep_their_corner_colours`; (34) `raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value`; (35) `mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing}` (`cargo test -p hayro`); (36) `raster::tests::mesh_triangles_sharing_an_edge_leave_no_seam`; (38) `raster::tests::default_colour_spaces::device_rgb_operators_draw_in_default_rgb`, `raster::tests::default_colour_spaces::device_cmyk_operators_draw_in_default_cmyk`, `raster::tests::default_colour_spaces::device_gray_operators_draw_in_default_gray`, `raster::tests::default_colour_spaces::device_images_draw_in_the_default_colour_spaces`, `raster::tests::default_colour_spaces::device_shadings_draw_in_the_default_colour_spaces`, `raster::tests::default_colour_spaces::an_image_drawn_with_and_without_a_default_keeps_both_colours`, `raster::tests::default_colour_spaces::forms_inherit_the_default_colour_spaces_they_do_not_override`, `raster::tests::default_colour_spaces::unusable_defaults_leave_device_colour_alone`; (39) `raster::tests::knockout_groups_replace_what_their_elements_cover`; (40) `raster::tests::visibility_expressions_decide_what_a_membership_dictionary_shows`; (41) `raster::tests::a_multiply_highlight_inside_a_transparency_group_keeps_the_text`; (42) `raster::tests::rendering_modes_but_3_and_7_draw_type3_glyphs_as_filled`; (43) `raster::tests::lab_honours_the_space_white_point`, `raster::tests::indexed_palettes_over_lab_keep_their_colours`; (44) `text::tests::identity_to_unicode_names_give_text`; (46) `raster::tests::stencil_masks_are_drawn_at_the_fill_opacity_once`; (47) `raster::tests::a_five_point_polyline_is_not_painted_as_its_bounding_box`, `raster::tests::a_path_that_visits_every_corner_out_of_order_is_not_a_rectangle`, `raster::tests::a_rectangle_is_still_painted_as_a_rectangle`; (48) `raster::tests::icc_profile_cache::icc_profiles_with_identical_dictionaries_draw_their_own_colours`, `raster::tests::icc_profile_cache::icc_profiles_used_again_are_read_once`; (49) `raster::tests::a_font_whose_post_version_has_swapped_halves_is_still_used`; (50) `font::blob::tests::glyph_names_resolve_to_the_first_glyph_with_that_name`; (51) `context::saved_state_tests::repeated_saves_stay_bounded_and_balanced`; (52) `font::cid::tests::w2_run_past_the_largest_cid_keeps_the_rest_of_the_table` (hayro-interpret); (55) `raster::tests::devicen_inks::a_tint_transform_that_cannot_be_evaluated_draws_the_rest_of_the_page_and_warns`, `raster::tests::devicen_inks::a_tint_transform_that_cannot_be_read_falls_back_and_warns`; (56) `raster::tests::invisible_unknown_annotations_and_popups_paint_nothing`; (57) `substitute_widths::a_substitute_keeps_its_glyph_shapes_at_wider_advances`, `substitute_widths::a_substitute_still_narrows_glyphs_to_narrower_advances` (pdfcraft-render `tests/it`); (58) `cid_fallback_query::a_cid_font_fallback_request_carries_its_descendants_style`, `cid_fallback_query::a_cid_font_without_a_descriptor_asks_for_the_default_style` (pdfcraft-render integration tests); (59) `raster::tests::no_rotate_notes_stay_upright_at_their_corner_on_rotated_pages` (hayro-interpret); (64) `type3_text_clip::{clip_only_type3_text_is_not_rendered_and_leaves_no_clip, invisible_type3_text_is_not_rendered, other_modes_render_type3_glyphs_filled_and_never_clip, only_outline_glyphs_shape_a_clip_shared_with_type3_text, self_nesting_type3_text_in_a_clip_mode_stays_bounded}` (`cargo test -p pdfcraft-render --test it`), `raster::tests::rendering_modes_but_3_and_7_draw_type3_glyphs_as_filled`, `text::tests::clipping_mode_type3_text_is_kept_by_the_ocr_filter`; (66) `separation_all::{separation_all_fills_paint_the_complemented_tint_as_gray, separation_all_strokes_paint_the_complemented_tint_as_gray, separation_all_images_paint_each_sample_as_the_complemented_gray, separation_all_shadings_paint_the_complemented_gray, separation_none_paints_nothing, a_named_separation_still_paints_through_its_alternate, a_device_n_space_naming_all_still_paints_through_its_alternate}` (`cargo test -p pdfcraft-render --test it`); (63) `color::device_cmyk_gray_tests::{each_ink_darkens_by_its_iso_weight, ink_past_full_coverage_is_black_and_bad_components_read_as_zero}` (hayro-interpret), `it::luminosity_group_space::{black_ink_in_a_device_cmyk_group_masks_fully_and_no_ink_shows_fully, each_process_ink_in_a_device_cmyk_group_masks_by_its_weight, a_device_cmyk_backdrop_colour_masks_by_its_gray, gray_and_rgb_groups_mask_as_before, device_cmyk_on_the_page_keeps_its_display_colour}` (pdfcraft-render); (71) `raster::tests::an_indexed_lookup_table_is_read_in_every_legal_form` (pdfcraft-render); (72) `raster::tests::an_uncoloured_pattern_uses_its_underlying_space_in_every_form` (pdfcraft-render); (73) `raster::tests::the_jbig2_images_of_a_page_share_one_allowance`, `a_picture_whose_jbig2_mask_is_past_the_pages_allowance_is_not_drawn` and `a_pages_jbig2_allowance_ends_with_the_page` (pdfcraft-render); (74) `raster::tests::{a_chain_of_colour_spaces_ends, colour_spaces_are_read_to_the_limit_of_their_nesting, a_chain_built_on_spaces_already_read_ends, a_profile_read_at_the_limit_keeps_its_alternate_for_a_read_nearer_the_top}` (pdfcraft-render); (75) `raster::tests::a_soft_hyphen_a_substitute_lacks_draws_its_hyphen` (pdfcraft-render); (89) `raster::tests::a_form_that_draws_itself_is_drawn_once`, `raster::tests::a_form_that_reaches_itself_through_a_pattern_or_a_glyph_is_drawn_once` (pdfcraft-render) | | hayro | 0.7.1 | Apache-2.0 OR MIT | (1) `RenderSettings::{x_offset, y_offset}` to render one tile of a large page; (2) images over `MAX_IMAGE_PIXELS` (2^28) are skipped (a fuzzed `/W 4294967295` hung in resampling); (3) `tiling_cell_scale` keeps a tiling pattern's cell pixmap within 3000 px a side whatever its `/XStep`/`/YStep` (a fuzzed step far beyond `/BBox` allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed `/LW 9223372036854775807` allocated 10 GB in stroke expansion); (5) `usable_dash_pattern`: a stroke whose dash pattern could cut it into more than `MAX_DASHES_PER_STROKE` (250,000) pieces, counting kurbo's restart of the pattern at every subpath and `ClosePath`, is drawn solid (a fuzzed `/D [[11] 0]` on a line from x = 92234775807 allocated over 5 GB, `[0.000001] 0 d` on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid (`[-1 -1] 0 d` looped for ever in kurbo); (7) `render_size` and `render_into`: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); `render` is `render_into` on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) `may_paint_canvas`: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); `set_skip_offscreen_paths` turns this off for the regression test; (10) `/Interpolate` governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour (since (21), only inside Type 3 glyphs), planned without the Catmull-Rom filter limits (`image_resampling_size` takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) `may_paint_rect`: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (`InterpreterSettings::preview`: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under `PREVIEW_MIN_IMAGE_AREA` device pixels, while final renders always draw them; (13) reading colours (#897): `RenderSettings::color_replacement` takes a `ColorReplacement` (text colour, background, `only_black`, `line_art`) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; `render_pdf` gains a `color_replacement: Option` argument (pass `None` for true colours); the upstream-only `[dev-dependencies.log]` entry is removed from `Cargo.toml` (log is a normal dependency since the logging patch), so `cargo test -p hayro` runs this crate's unit tests; (14) a shrunk image that is interpolated (or, since (21), area-averaged) and whose axes lie along the device's (upright, mirrored or a quarter turn, as on a page under `/Rotate`) is snapped to whole device pixels: shrunk once to its rounded device rectangle (halves round up, so a tile snaps an edge where the whole page does) and drawn 1:1 from the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (`snap_minified_image`; scans looked soft next to Acrobat, and until #5530 a turned one still did, up to 200 grey levels off the same page drawn upright). An image counts as axis-aligned when the terms the snap drops move its far edges by at most a hundredth of a device pixel (`SNAP_SKEW`), not by a tolerance on the coefficients, which let a slight skew of a very wide image through; and the snapped 1:1 transform is used only when the shrink produced the snapped size, so a failed resize keeps the full transform instead of drawing the source raster at its source size; (15) a soft mask of another size than its image is filtered when shrunk, whatever its `/Interpolate` (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroke thinner than the shrink factor; (16) under a soft mask or a blend mode, axial and radial shadings are sampled instead of drawn as native gradients (#2128): vello_cpu 0.0.8 copies an opaque gradient over fully covered tiles without the mask or blend mode, so PowerPoint's faded gradient fills came out solid; (17) the renderer keeps annotation paints out of the reading-colours mapping between `begin_annotations` and `end_annotations` (#1548); (18) the page renders as one transparency group; under reading colours its background is painted inside it as the initial backdrop, so a Multiply highlight blends with the dark background instead of staying an opaque block, and otherwise the background stays outside, so a blend mode over an empty part of the page keeps its colour instead of blending with the white paper (#2970); (19) `set_paint` samples a mesh shading's texture over its triangles alone (`EncodedShadingPattern::extent`, padded rather than repeated) and skips a mesh with none on the canvas: a BricsCAD plan's 310 one-triangle `sh` each covered the whole tile, 2 s per 1024 px tile at any zoom (the watchdog gave the page up on a busy machine), now about 70 ms; (20) a tiling pattern whose step is too large for a cell pixmap, where what is painted (a fill, or a stroke with its width) can only show the first cell (PowerPoint's `/XStep 2147480100` for a picture fill drawn once), has its `/BBox` drawn at full scale with a transparent border and padded instead of tiled (`shows_one_cell`; a cell whose scale isn't finite and positive, from a `/BBox` too thin to divide by, keeps the old path: `usable_cell_scale`); shrinking the cell to fit the 3000 px limit of (3) left the picture under a pixel, so network diagrams lost their pictures (#3349); (21) an image that does not request interpolation and is drawn below its resolution, outside a Type 3 glyph, is shrunk by area averaging (`area_shrink`: each device pixel the coverage-weighted mean of the samples it covers, in 16.16 fixed point whose weights sum to exactly one, premultiplied when the image has a mask; a soft mask no larger than the image is first sampled onto the image's grid by `mask_on_image_grid` (nearest-neighbour, or bilinearly when the mask asks for `/Interpolate`, so its edge stays as smooth as an enlarged mask's), so its colour and mask are averaged together instead of the averaged colour of transparent samples showing under a mask layer at the mask's edges) to its snapped device rectangle, instead of nearest-neighbour: `/Interpolate` asks for smoothing only when an image is enlarged (ISO 32000-2 §8.9.5.3), and nearest-neighbour kept one sample in N per axis, so the text of a 300 dpi scan at 100% lost most of its strokes (sampling them again at a fractional offset dropped more). PicScale's `Box` and `Area` filters were not used: shrinking, both average a window four target pixels wide. On a 13-page 300 dpi scan every rendered sample at 96, 120 and 192 dpi is within 1 of the exact area average (it was up to 255 off, 17-23% of samples by more than 1). A snapped image whose shrink failed keeps the full transform, so it is not drawn at its source size. The shrink never enlarges: a side drawn at or above its resolution (an image shrunk across and stretched down, as a 204 × 98 dpi fax is) keeps its samples for the sampler. Each target row averages the whole source rows it covers (contiguous, cheap) and is then shrunk across with taps worked out once per image, which made the shrink about three times faster (a 300 dpi page at 96 dpi: 25 ms, nearest-neighbour 11 ms); working a row at a time (`area_shrink_rows`) keeps its work memory to a few rows. An image with a soft mask, a stencil's ink included, is premultiplied a source row at a time as it is averaged (`area_shrink_premultiplied`), never as a full-resolution RGBA copy: a 10718 × 13701 stencil shown at page size peaked at 1.16 GB with that copy and peaks at 1.03 GB without it, as with nearest-neighbour. A source with a side over `MAX_RESAMPLING_SOURCE_SIDE` (2^20) keeps (10)'s nearest-neighbour shrink, and `area_shrink` refuses it, so its accumulator row and tap table stay a few times one source row (a 2^27 × 2 image of zeros is a tiny file); (22) knockout groups: while one is innermost, fills, strokes, glyph outlines and shadings are drawn with `Compose::Copy` (weighted by coverage), so each element replaces what earlier elements of the group drew instead of compositing over it; an image or nested group, drawn into a layer of its own, first has what is under its shape cleared (`knockout_image` for an unmasked image's rectangle, only when `draw_image` will draw it, decoded at the size it is drawn; a knockout erase pass that draws a nested group's content with every paint clearing instead of painting, a shading fill's clear clipped to the shading's `/BBox` as its paint is) and then composites over it (#2971); (23) backport of hayro 0.8: shadings that `as_svg_gradient` can represent are painted as a `vello_cpu` linear or two-point radial gradient instead of a texture sampled pixel by pixel through the shading function (InDesign fades are luminosity soft masks of one full-page axial `sh`: 40 on a report page took 15 s at 144 dpi, now under 1 s); without upstream's half-pixel shift, which suits vello_cpu 0.3 but moves 0.0.8's gradients off pixel centres; (25) a tiling cell is drawn at the scale that makes one `/XStep` (`/YStep`) exactly the cell pixmap's whole number of pixels (`tiling_cell_pixels`, at least 1 and at most the cap), so the repeated pixmap puts cells at whole multiples of the step at any device scale; drawing at the unrounded scale repeated the cell up to half a pixel off at every step, and the error grew with distance from the pattern's origin (ISO 32000-2 8.7.3.1); (28) an image of page-scan size (`SMOOTHED_IMAGE_SIDE`: 512 samples or more on its shorter side) that does not request interpolation is drawn with `ImageQuality::Medium` (bilinear, as interpolated images are) when enlarged by more than 1% on either side (`SMOOTHED_IMAGE_SCALE`: a scan drawn at its resolution can come out a few ulps above 1, and blurred by half a pixel at a half-pixel offset), also when the other side is shrunk (the area shrink leaves that side at 1:1), outside a Type 3 glyph: scans rarely set `/Interpolate`, and drawn nearest-neighbour every sample of a zoomed scan became a block, so its text had staircase edges; icons, QR codes and pixel art keep their hard pixels. Bicubic (`High`) looked the same on a zoomed 300 dpi scan and drew four 1024 px tiles at 400% in 267 ms against 50 ms | (1) `raster::tests::tiles_match_full_render`; (2) `raster::tests::absurd_image_dimensions_are_skipped`; (3) `raster::tests::tiling_cells_stay_small_whatever_the_step`; (4) `raster::tests::absurd_predictor_columns_and_line_widths_render`; (5) `raster::tests::dash_patterns_that_cut_a_stroke_into_billions_of_pieces_render`; (6) `raster::tests::invalid_dash_patterns_terminate_and_are_drawn_solid`; (7) `raster::tests::pixels_are_a_plain_renders_bytes_in_a_buffer_a_gui_can_take_over`; (8) `raster::tests::image_resampling_rejects_anisotropic_target_growth`, `raster::tests::image_resampling_bounds_intermediate_before_planning`, `raster::tests::image_resampling_rejects_invalid_sources_and_scales`, `raster::tests::image_resampling_keeps_valid_sizes_and_exact_limits`, `raster::tests::resampling_fallback_preserves_original_geometry_and_pixels`, `raster::tests::image_resampling_wide_sources_shrink_before_backend_side_limits`, `raster::tests::image_resampling_wide_transparent_sources_still_shrink`, `raster::tests::image_resampling_padded_backend_dimensions_stay_checked`, `raster::tests::image_resampling_mismatched_alpha_mask_keeps_pixels_and_placement`; (9) `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel`; (10) `raster::tests::non_interpolated_images_are_not_smoothed_when_minified`, `raster::tests::non_interpolated_gray_images_are_not_smoothed_when_minified`, `raster::tests::non_integer_minification_averages_the_columns_each_pixel_covers`, `raster::tests::non_integer_minification_averages_colour_and_mask_together`, `raster::tests::non_interpolated_alpha_images_keep_their_colours_and_mask_when_minified`, `raster::tests::non_interpolated_type3_glyph_images_are_not_smoothed_when_minified`, `raster::tests::image_resampling_nearest_accepts_wide_sources_the_filter_refuses`, `raster::tests::wide_non_interpolated_sources_draw_after_shrinking`; (11) `raster::tests::skipping_paths_off_the_canvas_changes_no_pixel`; (12) `raster::tests::preview_renders_skip_tiny_images_but_finals_draw_everything`; (13) `color_replacement_tests::{black_maps_to_text_colour, white_maps_to_background, midgrey_interpolates, alpha_copied_verbatim, only_black_threshold, line_art_false_skips_paths_but_not_text}`, `color_replacement_render_tests::{color_replacement_maps_text_and_line_art, color_replacement_maps_stencil_mask_ink}`; (14) `raster::tests::minified_image_edges_snap_to_whole_pixels`, `raster::tests::a_tile_snaps_half_pixel_image_edges_as_the_whole_page_does` (`cargo test -p hayro`), `raster::tests::a_turned_or_mirrored_minified_image_snaps_to_whole_pixels`, `rotated_image_snap::a_shrunk_image_on_a_turned_page_is_the_upright_render_turned` (`cargo test -p pdfcraft-render --test it`), `renderer::snap_tests::a_skew_that_moves_a_far_edge_is_not_snapped` (`cargo test -p hayro`); (15) `raster::tests::shrunk_soft_mask_larger_than_its_image_keeps_thin_strokes`; (16) `raster::tests::soft_mask_applies_to_a_shading_pattern_fill`; (17), (18) `raster::tests::reading_colors_keep_highlight_annotations_readable`; (19) `mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing}` (`cargo test -p hayro`); (20) `raster::tests::a_tiling_pattern_drawn_once_shows_its_cell`, `raster::tests::a_large_step_still_tiles_where_the_next_cell_shows`, `raster::tests::a_tiling_cell_too_thin_to_scale_keeps_the_old_path`; (21) `raster::tests::non_interpolated_minification_keeps_thin_strokes`, `raster::tests::non_integer_minification_averages_the_columns_each_pixel_covers`, `raster::tests::non_integer_minification_averages_colour_and_mask_together`, `raster::tests::non_interpolated_minification_does_not_bleed_transparent_colour`, `raster::tests::a_smaller_soft_mask_does_not_bleed_transparent_colour_when_minified`, `raster::tests::an_interpolated_smaller_soft_mask_stays_smooth_when_its_image_is_minified`, `raster::tests::a_shrunk_stencil_mask_keeps_thin_strokes`, `raster::tests::minified_image_edges_snap_to_whole_pixels`, `raster::tests::an_image_shrunk_one_way_and_stretched_the_other_is_still_averaged`, `raster::tests::wide_non_interpolated_sources_draw_after_shrinking`, `renderer::area_shrink_tests` (`cargo test -p hayro`); (22) `raster::tests::knockout_groups_replace_what_their_elements_cover`; (23) `raster::tests::axial_and_radial_shadings_draw_as_native_gradients`; (25) `tiling_cell_placement::{far_cells_of_a_fill_sit_at_whole_steps, far_cells_sit_at_whole_steps_under_a_turned_scaled_matrix, far_cells_of_a_stroke_sit_at_whole_steps, a_cell_pixmap_spans_exactly_one_step}` (`cargo test -p pdfcraft-render --test it`); (28) `raster::tests::large_non_interpolated_images_are_smoothed_when_enlarged`, `raster::tests::small_non_interpolated_images_keep_hard_pixels_when_enlarged`, `raster::tests::a_scan_shrunk_one_way_is_smoothed_the_way_it_is_stretched`, `raster::tests::a_scan_drawn_a_rounding_error_above_its_resolution_stays_sharp` | | hayro-jbig2 | 0.3.0 | Apache-2.0 OR MIT | (1) at most 2^28 pixels per bitmap and page bitmap, and 2^25 cells per halftone grid and grayscale image (each side could be 65535: 4.3 billion pixels decoded one by one, and a 4-byte value per cell); (2) at most 2^26 symbol pixels per symbol dictionary (a fuzzed dictionary decoded for minutes); (3) a symbol dictionary's height classes are capped at one per new symbol and its export runs at `2 × symbols + 1`, each plus `EXTRA_EMPTY_STEPS` (1024): an empty height class or a zero-length export run doesn't advance, and past the end of its data the arithmetic decoder can decode them for ever (one changed byte hung the renderer); (4) a segment header's long-form referred-to segment count is capped at `u16::MAX` and a larger count is refused before its retention or referred-to bytes are read: the 29-bit count went straight to `Vec::with_capacity`, a reservation of about 2 GiB whose failure aborts the process (from storytold/craft-libs#176); (5) what a page's segments hold for later ones (intermediate regions, pattern and symbol dictionaries with their headers and retained contexts, and Huffman tables) is at most 64 MiB, two square bitmaps of 16,384 a side (`MAX_RETAINED_WORDS`), and a dictionary's new and exported symbols are counted against it as they are made: each was bounded on its own, but a few dozen bytes an intermediate region held 33 MB each, and 8 MiB symbols exported again and again killed the process, so a few kilobytes could exhaust memory; (6) a segment may refer to at most 2^20 symbols, checked before any are collected: references may repeat, and 65,535 of them to one dictionary, in a 78 KB stream, asked for 17 GB; (7) a pattern dictionary keeps at most 16 MiB of shifted copies, which only speed up drawing, and none that wouldn't fit what the page has left (`MAX_SHIFTED_WORDS`); (8) a custom Huffman table has at most 2^16 lines, and a stream with its globals at most 2^20 segments and 2^22 references, checked as they are read: both grew before anything counted them, and PDFs of a kilobyte or so reached 5 GB and more behind nested Flate; (9) a page past the bounds of (1) is refused when the stream is opened: a caller sizes its output from the header before the page bitmap is made, and an Indexed image in a 743-byte PDF asked for 137 GB, which aborts; (10) a stream may ask for at most 2^30 units of decoding work, four of the largest pages, about 6 s to 17 s (`MAX_WORK`): decoding a bitmap costs a pixel each and 4 a row, copying one a word each and 1 a row, and each integer an item reads costs 38, what one takes at worst, so one with no pixels pays too, before it is decoded: each bitmap was bounded, but nothing bounded how many there were, and a 795-byte PDF took 265 s; (11) `Image::decode_within` spends from what a run of images may still do, starting with its page's pixels, and `charge_reading` charges reading a stream and its globals before anything is parsed, so an operation over a document can bound what `MAX_WORK` bounds for one image, and images that fail or decode nothing add up too: a file of thousands of images made to reach that limit took as many times as long; (12) adding to a location, a reference offset, a symbol's corner or a Huffman range line near the end of an `i32`, and negating a decoded integer of 2^31, overflowed: a panic wherever overflow is checked. The refinement context, a text region's corners, the integer decoder and the Huffman table wrap, as a release build did, and `Bitmap::combine` saturates, which comes to the same, as a location that overflows is past the end of any page. What is drawn doesn't change; (13) a bitmap filled with a default pixel of 1 (a page, or an intermediate text or halftone region) had 1s past its width in the last word of each row, and a refinement, which reads its reference a word at a time, read them as pixels where the standard has 0 (6.3.5.2): refining such a reference not a multiple of 32 wide decoded to other pixels than were coded, release builds too. `Bitmap::clear_padding` clears those bits when a bitmap is filled with 1s, as cutting symbols and patterns from a collective bitmap already did. Ghostscript reads the test's page and text streams as they decode now; (14) a region's location is unsigned (5.4.4), so one past `i32::MAX` is off the page, where the decoder took it for a signed one: a region there was drawn back onto the page from the left or the top (at 0xFFFFFFFC, four pixels left of it), and a refinement there, of the page or of a region at the same place (7.4.7.5), was refused as an overflow, and the page with it. `Bitmap::combine_region` keeps such a location past the page. A refinement of another region refines it where it is, as Table 35 has it (GRREFERENCEDX and GRREFERENCEDY 0), whatever the two locations say; a refinement of the page, see (16); (15) a refinement whose reference is offset right of it (GRREFERENCEDX above 0, which a refined text instance has from a wider bitmap, Table 12, and a refined symbol from an RDX above 0, Table 18) read the reference's first 32 columns as 0: `RefinementContextGatherer::maybe_reload_buffers` emptied its buffers left of the reference and didn't load them again at its first column. Such a refinement decoded to other pixels than were coded, or failed and lost the page. Left of the reference the buffers now keep the words of its first column each row starts with; with an offset of 0 or less nothing changes; (16) a refinement of the page refines "the current contents of the page buffer, restricted to the area of the page buffer specified by this segment's region segment information field" (7.4.7.4), at an offset of 0 (Table 35): `Bitmap::area` copies that area a word at a time, 0 off the page and past its width whatever the padding of its rows holds, and every refinement region reads its reference at offset 0. The page around the area was read too, so a refinement of the page decoded to other pixels than were coded, and one wider or taller than the page was refused. And `RefinementContextGatherer::start_row` loads the reference's words a column left of where a row starts in it, as the reload would, not from column 0: with the reference offset 32 or more left of the refinement (GRREFERENCEDX below −31, as a refined text instance has from a narrower bitmap or an RDX below −31, and a refined symbol from such an RDX), each row's first context read 0 for the reference column it starts at, and from 33 left for the one before it too, and the refinement decoded to other pixels or failed | (1) `raster::tests::jbig2_regions_of_billions_of_pixels_are_refused`, `raster::tests::jbig2_halftone_grids_past_their_bound_are_refused`; (2) the nightly `cargo xtask fuzz` finding, which now finishes in about a second; (3) `raster::tests::jbig2_symbol_dictionaries_that_never_advance_terminate`; (4) `segment::tests::test_referred_to_count_is_capped` (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`); (5) `raster::tests::jbig2_regions_held_past_a_pages_bound_are_refused`, `jbig2_symbols_without_pixels_count_towards_a_pages_bound`, `jbig2_retained_contexts_count_towards_a_pages_bound`, `jbig2_huffman_tables_count_towards_a_pages_bound`, `jbig2_new_symbols_count_towards_a_pages_bound`, `jbig2_new_symbols_are_held_to_what_a_page_has_left`, `jbig2_pattern_dictionaries_count_towards_a_pages_bound`, and in the crate (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`) `decode::symbol::tests::exported_symbols_are_counted_before_they_are_copied`, `tests::the_room_left_shrinks_as_segments_are_stored`; (6) `raster::tests::jbig2_segments_referring_to_too_many_symbols_are_refused`; (7) `raster::tests::jbig2_pattern_dictionaries_too_large_to_shift_still_draw`, `jbig2_shifted_patterns_never_refuse_an_image`; (8) `raster::tests::jbig2_huffman_tables_of_too_many_lines_are_refused`, `jbig2_streams_of_too_many_segments_are_refused`, `jbig2_streams_of_too_many_references_are_refused`, and in the crate `tests::the_segments_of_a_stream_and_its_globals_count_together`; (9) `raster::tests::jbig2_pages_past_the_bounds_are_refused_before_anything_is_sized_from_them`, `tests::a_page_past_the_bounds_is_refused_when_opened`; (10) `raster::tests::jbig2_decoding_work_is_counted_item_by_item`; (11) `jbig2::decode::tests::a_run_of_images_spends_its_allowance_down_to_the_step` (pdfcraft-filters); (12) `raster::tests::jbig2_placements_at_the_end_of_an_i32_draw_nothing`; (13) `jbig2::decode::tests::a_refinement_reads_nothing_past_its_references_width` (pdfcraft-filters), and in the crate `tests::a_bitmap_of_1s_has_nothing_past_its_width`; (14) `jbig2::decode::tests::a_region_located_past_i32_max_is_off_the_page` (pdfcraft-filters), and in the crate `tests::a_region_located_past_i32_max_is_off_the_page` and `tests::a_refinement_of_a_region_decodes_the_same_wherever_they_are`; (15) `jbig2::decode::tests::a_refinement_reads_a_reference_right_of_it_from_its_first_column` (pdfcraft-filters), and in the crate `tests::a_refinement_reads_a_reference_right_of_it_from_its_first_column`; (16) `jbig2::decode::tests::a_refinement_refines_the_page_in_its_area_and_any_reference_where_it_is` (pdfcraft-filters), and in the crate `tests::a_refinement_refines_the_page_in_its_area_and_any_reference_where_it_is` and `tests::a_page_area_is_the_page_inside_it_and_0_outside` | | lopdf | 0.45.0 | MIT | (1) PNG predictor rows longer than the data are refused before their buffers are allocated (a fuzzed `/Columns 4294967295` allocated two 4 GiB rows), and empty frames return without row buffers; (2) predictor row widths use saturating arithmetic or reject overflow; (3) sub-byte TIFF Predictor 2 scratch is bounded by the samples present in the input instead of the declared `/Colors`, aligned with [lopdf#589](https://github.com/J-F-Liu/lopdf/pull/589); (4) the object lookup keeps its cycle history through compressed objects, requires an ordinary in-file container, and caps the chain of indirect lookups at 16 (a chain of new objects, each a `/Length` compressed in a container whose own `/Length` names the next, or a stream whose `/Length` names the next stream, overflowed the stack: a dev build's 2 MiB thread within 800 hops); (5) object-stream members are parsed within their own extent (up to the next distinct offset), an entry whose offset repeats an earlier entry's is skipped (so a repeated object number can resolve to its earlier entry), and offsets are processed in index order; (6) `find_latest_trailer`: a rebuild takes a `/Type /XRef` stream dictionary as the trailer when the file has no `trailer` keyword (the latest in the file wins, as a later `trailer` would), so a concatenated or truncated file keeps its /Info and /ID instead of failing to load; a candidate whose /Root is the catalog beats one whose /Root is another dictionary, and an in-file /Root must match its generation (#2365) | (1) `inspect::tests::png_predictor_rows_longer_than_the_data_are_refused_before_allocating`, `inspect::tests::empty_png_predictor_frames_do_not_allocate_rows`; (2) `inspect::tests::tiff_subbyte_predictor_row_width_overflow_is_refused`; (3) `inspect::tests::tiff_subbyte_predictor_scratch_is_bounded_by_available_samples`, `inspect::tests::tiff_subbyte_predictors_keep_components_rows_and_padding`; (4) `inspect::tests::compressed_object_self_container_does_not_recurse`, `inspect::tests::compressed_object_mutual_containers_do_not_recurse`, `inspect::tests::object_stream_length_cycle_does_not_recurse`, `inspect::tests::compressed_object_container_chain_does_not_recurse`, `inspect::tests::compressed_length_chain_does_not_recurse`, `inspect::tests::plain_stream_length_chain_does_not_recurse`, `inspect::tests::compressed_objects_and_indirect_stream_length_still_load`, `inspect::tests::compressed_length_in_a_container_with_an_indirect_length_still_loads`; (5) `inspect::tests::object_stream_member_extents_bound_owned_strings`, `inspect::tests::object_stream_member_extents_preserve_index_order_and_whitespace`; (6) `inspect::tests::rebuilt_file_keeps_the_info_of_its_cross_reference_stream`, `inspect::tests::rebuilt_file_prefers_a_trailer_whose_root_is_the_catalog` | diff --git a/vendor/hayro-interpret/src/context.rs b/vendor/hayro-interpret/src/context.rs index de3907b1ff..a281cd04b9 100644 --- a/vendor/hayro-interpret/src/context.rs +++ b/vendor/hayro-interpret/src/context.rs @@ -37,6 +37,12 @@ std::thread_local! { // PdfCraft patch (18): whether this page's content budget ran out, reported as // `InterpreterWarning::ContentTruncated`. static CONTENT_TRUNCATED: core::cell::Cell = const { core::cell::Cell::new(false) }; + // PdfCraft patch (89): the form XObjects being drawn on this page, outermost first, across + // the contexts of the tiling patterns, Type 3 glyphs, soft masks and groups painted while + // they draw. A form already on it draws itself and is skipped. Its length is bounded by the + // nesting limits: every form on it is nested in the ones before it. + static FORMS_DRAWING: RefCell> = + const { RefCell::new(Vec::new()) }; } /// PdfCraft patch: decoded content-stream bytes one page may hold at once: its contents, and @@ -218,6 +224,8 @@ impl<'a> Context<'a> { NESTED_PAINTS.with(|n| n.set(0)); CONTENT_LEFT.with(|c| c.set(MAX_PAGE_CONTENT)); CONTENT_TRUNCATED.with(|truncated| truncated.set(false)); + // PdfCraft patch (89): no form of another page is being drawn. + FORMS_DRAWING.with(|forms| forms.borrow_mut().clear()); // PdfCraft patch (73): and fresh JBIG2 decoding, which every image of it shares. hayro_syntax::set_page_jbig2_work(settings.jbig2_work); // PdfCraft patch (55): a fallback left over from another page is not this page's. @@ -267,6 +275,56 @@ impl<'a> Context<'a> { } } + /// PdfCraft patch (89): begins drawing the form `id`; `false` when that form is already + /// being drawn, so drawing it again would never end (ISO 32000-2 §8.10.1: a form is a + /// self-contained description; one that contains itself describes nothing finite). The + /// skip is reported as `InterpreterWarning::ContentTruncated`. A form without an object + /// number can't be told apart and is only bounded by the nesting limits. + /// + /// The forms being drawn are those of this page's context and of every context painting + /// inside it (a tiling pattern, a Type 3 glyph, a soft mask or a group), so a form that + /// reaches itself through one of those is skipped too, while a form drawn again after it + /// has finished draws as usual. + pub(crate) fn enter_form( + &mut self, + id: Option, + ) -> bool { + let Some(id) = id else { + return true; + }; + // The stack is never borrowed across a call, so the borrow succeeds; if it didn't, + // the form draws, bounded by the nesting limits as before this patch. + let entered = FORMS_DRAWING.with(|forms| { + let Ok(mut forms) = forms.try_borrow_mut() else { + return None; + }; + if forms.contains(&id) { + return Some(false); + } + forms.push(id); + Some(true) + }); + if entered == Some(false) { + warn!("form XObject draws itself; the inner drawing is skipped"); + CONTENT_TRUNCATED.with(|truncated| truncated.set(true)); + return false; + } + true + } + + /// PdfCraft patch (89): ends drawing a form begun with `enter_form`. + pub(crate) fn leave_form(&mut self, id: Option) { + if let Some(id) = id { + FORMS_DRAWING.with(|forms| { + if let Ok(mut forms) = forms.try_borrow_mut() + && forms.last() == Some(&id) + { + forms.pop(); + } + }); + } + } + pub(crate) fn save_state(&mut self) { let Some(cur) = self.states.last() else { warn!("attempted to save state without existing state"); diff --git a/vendor/hayro-interpret/src/x_object.rs b/vendor/hayro-interpret/src/x_object.rs index 6fcb34a5d7..aa05c4f455 100644 --- a/vendor/hayro-interpret/src/x_object.rs +++ b/vendor/hayro-interpret/src/x_object.rs @@ -121,6 +121,24 @@ pub(crate) fn draw_form_xobject<'a, 'b>( x_object: &'b FormXObject<'a>, context: &mut Context<'a>, device: &mut impl Device<'a>, +) { + // PdfCraft patch (89): a form already being drawn is not drawn inside itself. A name its + // content uses can reach it again (through its own resources, or the page's, which a form + // with resources of its own still falls back on), and every such level used to run until + // the page's nested-paint budget was spent. + let id = x_object.dict.obj_id(); + if !context.enter_form(id) { + return; + } + draw_form_xobject_once(resources, x_object, context, device); + context.leave_form(id); +} + +fn draw_form_xobject_once<'a, 'b>( + resources: &Resources<'a>, + x_object: &'b FormXObject<'a>, + context: &mut Context<'a>, + device: &mut impl Device<'a>, ) { if !context.ocg_state.is_visible() { return; From 3ae9093d6198fe329ef6352ac674468213cb2ff8 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:30:37 +0800 Subject: [PATCH 07/75] M2: `info` lists the repairs made while opening, as Document Properties does (misc.cli-inspect) (#6518) * M2: `info` lists the repairs made while opening, as Document Properties does (misc.cli-inspect) `pdfcraft-cli info` now prints a `repairs` array: the notes about damage repaired while opening that the app shows in Document Properties and doc_info reports. It opens the document as the app does, once (the session the dynamic XFA summary already used), so notes the app adds after reading the objects, such as a page tree rebuilt from the pages found by scanning, are listed too. A file the app cannot open for editing falls back to what reading its objects repaired; a sound file prints an empty array. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M2: `info` runs none of the document's scripts while inspecting it (misc.cli-inspect) The session `info` opens to read the app's repair notes and the dynamic XFA layout now has JavaScript turned off, so an XFA form's initialize and calculate scripts, which the app runs on opening, do not run during inspection of an untrusted file. Before this branch the session was opened for dynamic XFA forms only, and their scripts ran there too. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn * M2: `info` reads the standards and repairs from the session it opens, and a laid-out form keeps its repair notes (misc.cli-inspect) `info` now takes the declared standards (`Document::standards`, as the Standards panel reads them) and the repair notes from the session it opened, and reads the file once more only when the app cannot open it. When opening lays out a dynamic XFA form, or writes any other revision, the notes about damage repaired while reading the file now stay with the re-read document, so Document Properties and `info` both list them. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA --------- Co-authored-by: mtlcl410 Co-authored-by: Claude Opus 5.5 --- apps/pdfcraft-cli/src/commands.rs | 45 ++++++++----- apps/pdfcraft-cli/tests/it/info_repairs.rs | 75 ++++++++++++++++++++++ apps/pdfcraft-cli/tests/it/main.rs | 1 + apps/pdfcraft-cli/tests/it/xfa.rs | 43 ++++++++++++- crates/engine/src/open.rs | 13 ++-- parity/acrobat-features.toml | 4 +- 6 files changed, 156 insertions(+), 25 deletions(-) create mode 100644 apps/pdfcraft-cli/tests/it/info_repairs.rs diff --git a/apps/pdfcraft-cli/src/commands.rs b/apps/pdfcraft-cli/src/commands.rs index 3118c901b4..c64f9e1c0f 100644 --- a/apps/pdfcraft-cli/src/commands.rs +++ b/apps/pdfcraft-cli/src/commands.rs @@ -8,6 +8,9 @@ pub(crate) fn read(path: &str) -> Result>, String> { std::fs::read(path).map(Arc::new).map_err(|e| format!("{path}: {e}")) } +/// `info FILE`: a JSON summary of the document as the app opens it (pages, fields, metadata, the +/// standards it declares, and the notes about damage repaired while opening it), with none of +/// its scripts run. pub(crate) fn info(args: &[String]) -> Result<(), CliError> { validate_options("info", args, INFO_OPTIONS)?; let path = *positional(args, INFO_OPTIONS).first().ok_or("info: missing file")?; @@ -18,24 +21,33 @@ pub(crate) fn info(args: &[String]) -> Result<(), CliError> { let mut info = inspect(bytes.clone(), render_password.as_deref()).map_err(|e| e.to_string())?; let pdf_pages = info.pages.len(); let acroform_fields = info.fields.len(); + // The document as the app opens it: the notes about damage repaired while opening, which the + // app shows in Document Properties, and for a dynamic XFA form the laid-out pages and fields. + // Inspection runs none of the document's scripts: an XFA form's initialize and calculate + // scripts, which the app runs on opening, are left alone, as with JavaScript turned off. + let mut session = pdfcraft_engine::Session::new(); + session.set_javascript(false); + let opened = session.open("info.pdf", None, bytes.clone(), password).ok().and_then(|id| session.get(id)); // Dynamic XFA forms: say what laying the template out gives (what the app shows). - let xfa_layout = (info.xfa == Some(pdfcraft_render::Xfa::Dynamic)) - .then(|| { - let mut s = pdfcraft_engine::Session::new(); - let id = s.open("info.pdf", None, bytes.clone(), password).ok()?; - let d = s.get(id)?; - d.xfa.as_ref().map(|x| { - info = d.info.clone(); - // The in-memory layout is an appended revision, not the file being inspected. - info.file_size = bytes.len(); - serde_json::json!({ "pages": x.pages, "fields": x.fields, "warnings": x.warnings }) - }) + let xfa_layout = opened.filter(|_| info.xfa == Some(pdfcraft_render::Xfa::Dynamic)).and_then(|d| { + d.xfa.as_ref().map(|x| { + info = d.info.clone(); + // The in-memory layout is an appended revision, not the file being inspected. + info.file_size = bytes.len(); + serde_json::json!({ "pages": x.pages, "fields": x.fields, "warnings": x.warnings }) }) - .flatten(); - // What the document declares as a standard, read the way the Standards panel reads it, so - // `info` and the app agree. A file that declares nothing leaves these null and empty. - let declared = - pdfcraft_cos::Document::open_with_password(bytes.clone(), password).map(|doc| pdfcraft_preflight::declared(&doc)).unwrap_or_default(); + }); + // What the document declares as a standard, as the Standards panel reads it, and the notes + // Document Properties shows, which include what the app repaired after reading the objects (a + // page tree rebuilt from the pages found by scanning), both from the session's own reading. A + // file the app cannot open is read once more for what the object reader declares and repaired. + // A file that declares nothing leaves the standards null and empty. + let (declared, repairs) = match opened { + Some(d) => (d.standards(), d.repair_log()), + None => pdfcraft_cos::Document::open_with_password(bytes.clone(), password) + .map(|cos| (pdfcraft_preflight::declared(&cos), cos.repair_log().to_vec())) + .unwrap_or_default(), + }; let json = serde_json::json!({ "file": path, "pdf_version": info.pdf_version, @@ -56,6 +68,7 @@ pub(crate) fn info(args: &[String]) -> Result<(), CliError> { "custom_properties": info.custom.len(), "page_labels": info.pages.iter().take(8).map(|p| p.label.clone()).collect::>(), "warnings": info.warnings, + "repairs": repairs, }); stdout_line(format_args!("{}", serde_json::to_string_pretty(&json).unwrap_or_default()))?; Ok(()) diff --git a/apps/pdfcraft-cli/tests/it/info_repairs.rs b/apps/pdfcraft-cli/tests/it/info_repairs.rs new file mode 100644 index 0000000000..72e8bed1a3 --- /dev/null +++ b/apps/pdfcraft-cli/tests/it/info_repairs.rs @@ -0,0 +1,75 @@ +//! `info` lists the notes about damage repaired while opening, as the app shows them in Document +//! Properties and as `doc_info` reports them in `repairs`: a rebuilt cross-reference index, and a +//! page tree rebuilt from the pages found by scanning (a note the app adds after reading). + +use std::process::Command; + +const BIN: &str = env!("CARGO_BIN_EXE_pdfcraft-cli"); + +/// A PDF with `pages` pages, each drawing a rectangle (a page found by scanning has content), and +/// `catalog` as object 1; `startxref` points at its cross-reference table, or at `bad_startxref`. +fn fixture(catalog: &str, pages: usize, bad_startxref: Option) -> Vec { + let content = "0 0 0 rg 10 10 40 20 re f"; + let kids: String = (0..pages).map(|i| format!("{} 0 R ", 4 + i)).collect(); + let mut objs = vec![ + catalog.to_string(), + format!("<< /Type /Pages /Kids [{kids}] /Count {pages} >>"), + format!("<< /Length {} >>\nstream\n{content}\nendstream", content.len()), + ]; + objs.extend((0..pages).map(|_| "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 120 60] /Contents 3 0 R >>".to_string())); + let mut out = b"%PDF-1.7\n".to_vec(); + let mut offsets = Vec::new(); + for (i, o) in objs.iter().enumerate() { + offsets.push(out.len()); + out.extend_from_slice(format!("{} 0 obj\n{o}\nendobj\n", i + 1).as_bytes()); + } + let xref = out.len(); + out.extend_from_slice(format!("xref\n0 {}\n0000000000 65535 f \n", objs.len() + 1).as_bytes()); + for off in &offsets { + out.extend_from_slice(format!("{off:010} 00000 n \n").as_bytes()); + } + let start = bad_startxref.unwrap_or(xref); + out.extend_from_slice(format!("trailer\n<< /Size {} /Root 1 0 R >>\nstartxref\n{start}\n%%EOF\n", objs.len() + 1).as_bytes()); + out +} + +const CATALOG: &str = "<< /Type /Catalog /Pages 2 0 R >>"; + +fn info_of(bytes: &[u8], name: &str) -> serde_json::Value { + let dir = std::env::temp_dir().join(format!("pdfcraft-info-repairs-{}-{name}", std::process::id())); + std::fs::create_dir_all(&dir).unwrap(); + let path = dir.join("f.pdf"); + std::fs::write(&path, bytes).unwrap(); + let out = Command::new(BIN).args(["info", path.to_str().unwrap()]).output().unwrap(); + assert!(out.status.success(), "{}", String::from_utf8_lossy(&out.stderr)); + let json = serde_json::from_slice(&out.stdout).expect("info prints JSON"); + let _ = std::fs::remove_dir_all(&dir); + json +} + +fn repairs(json: &serde_json::Value) -> Vec { + json["repairs"].as_array().unwrap_or_else(|| panic!("no repairs array: {json}")).iter().map(|l| l.as_str().unwrap().to_string()).collect() +} + +#[test] +fn info_lists_a_rebuilt_cross_reference_index() { + // `startxref` points into the middle of an object, so the index is rebuilt from the headers. + let json = info_of(&fixture(CATALOG, 1, Some(20)), "xref"); + assert_eq!(json["pages"], serde_json::json!(1), "{json}"); + let lines = repairs(&json); + assert!(lines.iter().any(|l| l.contains("rebuilt cross-reference index")), "{lines:?}"); +} + +#[test] +fn info_lists_a_page_tree_the_app_rebuilt_by_scanning() { + // The catalog names a page tree object the file does not have (#1656). + let json = info_of(&fixture("<< /Type /Catalog /Pages 9 0 R >>", 2, None), "pagetree"); + let lines = repairs(&json); + assert!(lines.iter().any(|l| l.contains("page tree is missing")), "{lines:?}"); +} + +#[test] +fn info_lists_no_repairs_for_a_sound_file() { + let json = info_of(&fixture(CATALOG, 1, None), "sound"); + assert_eq!(json["repairs"], serde_json::json!([]), "{json}"); +} diff --git a/apps/pdfcraft-cli/tests/it/main.rs b/apps/pdfcraft-cli/tests/it/main.rs index 30a6e25882..7b7292fe45 100644 --- a/apps/pdfcraft-cli/tests/it/main.rs +++ b/apps/pdfcraft-cli/tests/it/main.rs @@ -8,6 +8,7 @@ mod check; mod edit; mod extract; mod help; +mod info_repairs; mod info_standards; mod mcp_compact; mod options; diff --git a/apps/pdfcraft-cli/tests/it/xfa.rs b/apps/pdfcraft-cli/tests/it/xfa.rs index 1b5ff7bc9d..4d5df67535 100644 --- a/apps/pdfcraft-cli/tests/it/xfa.rs +++ b/apps/pdfcraft-cli/tests/it/xfa.rs @@ -7,6 +7,11 @@ use std::sync::Arc; const BIN: &str = env!("CARGO_BIN_EXE_pdfcraft-cli"); fn fixture(test: &str) -> PathBuf { + fixture_with(test, "") +} + +/// The two-page form, with `event` inside its `name` field. +fn fixture_with(test: &str, event: &str) -> PathBuf { let xdp = r#" "#; +"# + .replace("{event}", event); let objects = [ "<< /Type /Catalog /Pages 2 0 R /NeedsRendering true /AcroForm << /XFA 5 0 R /Fields [] >> >>".to_string(), "<< /Type /Pages /Kids [3 0 R] /Count 1 >>".into(), @@ -67,6 +73,21 @@ fn info_reports_the_laid_out_pages_and_fields() { assert_eq!(std::fs::read(path).unwrap(), original, "inspection must not write the layout to the input"); } +/// `info` inspects an untrusted file: a form's initialize script, which the app runs on opening +/// and which here would change the field's value, does not run. +#[test] +fn info_runs_none_of_the_forms_scripts() { + let script = r#""#; + let path = fixture_with("info-script", script); + let output = Command::new(BIN).arg("info").arg(&path).output().unwrap(); + success(&output); + let info: serde_json::Value = serde_json::from_slice(&output.stdout).unwrap(); + assert_eq!(info["xfa_layout"]["pages"], 2, "the form is still laid out: {info}"); + let warnings = info["warnings"].as_array().unwrap(); + assert!(!warnings.iter().any(|w| w.as_str().is_some_and(|w| w.contains("scripts changed it on opening"))), "{warnings:?}"); + assert_eq!(info["repairs"], serde_json::json!([]), "{info}"); +} + #[test] fn text_reads_the_layout_including_a_page_absent_from_the_placeholder() { let path = fixture("text"); @@ -134,3 +155,21 @@ fn a_laid_out_form_keeps_the_original_copy_permissions() { assert_eq!(info["file_size"], original.len()); assert_eq!(std::fs::read(path).unwrap(), original); } + +/// A damaged dynamic XFA form: `info` lists the cross-reference index rebuilt while opening it, +/// as Document Properties does after the form is laid out as one more revision. +#[test] +fn info_lists_the_repairs_of_a_laid_out_form() { + let path = fixture("info-damaged"); + let bytes = std::fs::read(&path).unwrap(); + let at = bytes.windows(10).rposition(|w| w == b"startxref\n").unwrap() + 10; + let mut damaged = bytes[..at].to_vec(); + damaged.extend_from_slice(b"20\n%%EOF\n"); + std::fs::write(&path, &damaged).unwrap(); + let output = Command::new(BIN).arg("info").arg(&path).output().unwrap(); + success(&output); + let info: serde_json::Value = serde_json::from_slice(&output.stdout).unwrap(); + assert_eq!(info["xfa_layout"]["pages"], 2, "the form is still laid out: {info}"); + let repairs = info["repairs"].as_array().unwrap(); + assert!(repairs.iter().any(|l| l.as_str().is_some_and(|l| l.contains("rebuilt cross-reference index"))), "{repairs:?}"); +} diff --git a/crates/engine/src/open.rs b/crates/engine/src/open.rs index 80a5ad30c7..facf058695 100644 --- a/crates/engine/src/open.rs +++ b/crates/engine/src/open.rs @@ -125,11 +125,17 @@ impl Session { // pages and fields so the rest of the engine works on it. The original bytes stay: the // laid-out form is one appended revision. let opts = SaveOptions { mod_date: context.now.map(pdfcraft_cos::pdf_date), ..SaveOptions::default() }; - // Write `work` as one more revision and reopen it: what the document then is. + // Write `work` as one more revision and reopen it: what the document then is. The notes + // about damage repaired while reading the file stay with it (Document Properties), once. let rebase = |work: &pdfcraft_cos::Document| -> Result<(Arc>, DocInfo, pdfcraft_cos::Document), String> { let new_bytes = write_incremental(work, &opts).map(Arc::new).map_err(|e| e.to_string())?; let new_info = inspect(new_bytes.clone(), render_password.as_deref()).map_err(|e| e.to_string())?; - let new_cos = pdfcraft_cos::Document::open_with_password(new_bytes.clone(), password).map_err(|e| e.to_string())?; + let mut new_cos = pdfcraft_cos::Document::open_with_password(new_bytes.clone(), password).map_err(|e| e.to_string())?; + for line in work.repair_log() { + if !new_cos.repair_log().contains(line) { + new_cos.note_repair(line.clone()); + } + } Ok((new_bytes, new_info, new_cos)) }; let (bytes, info, cos, xfa) = match (info.xfa, cos) { @@ -254,9 +260,6 @@ impl Session { }; match rebuilt { Some((b, i, mut c)) => { - for line in cos.repair_log() { - c.note_repair(line.clone()); - } c.note_repair(format!("the page tree is missing; its {} pages were found by scanning the file", i.pages.len())); (b, i, Ok(Ok(c))) } diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index d8918e5bd6..9b341fcd8b 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -9637,8 +9637,8 @@ milestone = "M2" status = "shipped" commands = [] tools = ["doc_info", "page_render", "text_extract"] -evidence = ["crates/automation/tests/it/automation.rs::open_inspect_render_and_find"] -notes = "" +evidence = ["crates/automation/tests/it/automation.rs::open_inspect_render_and_find", "apps/pdfcraft-cli/tests/it/info_repairs.rs::info_lists_a_page_tree_the_app_rebuilt_by_scanning"] +notes = "`info` lists the repair notes the app shows in Document Properties (`repairs`)." [[feature]] id = "misc.cli-edit" From c02341f6c0e0afe5b9c458794fcf236e16c34f90 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:31:00 +0800 Subject: [PATCH 08/75] M2: a wide stroke of a path with `y` curves is drawn, not dropped (vello_common patch 4) (#6526) A cubic whose second control point is its end point (as the `y` operator builds, ISO 32000-2 8.5.2.1 Table 58) has a zero derivative at its end. kurbo 0.13.1's offset curve looks for cusps with a root search on the curvature, which grows without bound towards that end; at some widths the search stops on the end point and normalising the zero tangent there gives NaN. vello_common's fill drops any outline with NaN, so the whole stroke vanished (a two-curve `y` path drew nothing at 16, 32 and 63 pt; a PDFBox tracker file lost its 32 pt "XX" outline). The NaN fallback that #4913 added to `flatten::stroke` now also moves such a control point the stroke tolerance towards the previous control point before stroking again (the curve moves by at most 4/9 of the tolerance). Outlines that come back finite take the old path unchanged. The vendor ledger row also gets back #4913's entry, which a merge had dropped. Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA Co-authored-by: mtlcl410 Co-authored-by: Claude Opus 5.5 --- crates/render/tests/it/main.rs | 1 + .../tests/it/stroke_zero_end_tangent.rs | 61 ++++++++++++ vendor/README.md | 2 +- vendor/vello_common/src/flatten.rs | 98 ++++++++++++++++++- 4 files changed, 158 insertions(+), 4 deletions(-) create mode 100644 crates/render/tests/it/stroke_zero_end_tangent.rs diff --git a/crates/render/tests/it/main.rs b/crates/render/tests/it/main.rs index 7e8307d908..591fc79b3e 100644 --- a/crates/render/tests/it/main.rs +++ b/crates/render/tests/it/main.rs @@ -13,6 +13,7 @@ mod search; mod separation_all; mod stray_q_in_a_form; mod stroke_degenerate_cubic; +mod stroke_zero_end_tangent; mod substitute_widths; mod tiling_cell_placement; mod type3_text_clip; diff --git a/crates/render/tests/it/stroke_zero_end_tangent.rs b/crates/render/tests/it/stroke_zero_end_tangent.rs new file mode 100644 index 0000000000..50a22f5c7b --- /dev/null +++ b/crates/render/tests/it/stroke_zero_end_tangent.rs @@ -0,0 +1,61 @@ +//! A contributor-original synthetic page: a wide stroke of a path built with `y` (the second +//! control point on the end point, ISO 32000-2 §8.5.2.1 Table 58) was not drawn at all at some widths, +//! because the stroker put NaN into its outline and the rasterizer dropped the whole path. A file +//! in the PDFBox issue tracker lost the 32 pt outline of its "XX" this way. +use pdfcraft_render::{PageRenderer, RenderConfig, RenderRequest}; +use std::sync::Arc; + +/// A one-page 200 × 200 pt PDF drawing `content`, with a cross-reference table. +fn pdf(content: &str) -> Vec { + let objects = [ + "<< /Type /Catalog /Pages 2 0 R >>".to_string(), + "<< /Type /Pages /Kids [3 0 R] /Count 1 >>".to_string(), + "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /Contents 4 0 R >>".to_string(), + format!("<< /Length {} >>\nstream\n{content}\nendstream", content.len()), + ]; + let mut bytes = b"%PDF-1.7\n".to_vec(); + let mut offsets = Vec::new(); + for (i, object) in objects.iter().enumerate() { + offsets.push(bytes.len()); + bytes.extend_from_slice(format!("{} 0 obj\n{object}\nendobj\n", i + 1).as_bytes()); + } + let xref = bytes.len(); + bytes.extend_from_slice(format!("xref\n0 {}\n0000000000 65535 f \n", objects.len() + 1).as_bytes()); + for offset in offsets { + bytes.extend_from_slice(format!("{offset:010} 00000 n \n").as_bytes()); + } + bytes.extend_from_slice(format!("trailer << /Size {} /Root 1 0 R >>\nstartxref\n{xref}\n%%EOF\n", objects.len() + 1).as_bytes()); + bytes +} + +/// The red channel of the pixel at (`x`, `y`) pt from the top left of `content` rendered at 1×. +fn red_at(content: &str, x: usize, y: usize) -> u8 { + let mut renderer = PageRenderer::new(Arc::new(pdf(content)), RenderConfig::default()); + let out = renderer.render(RenderRequest { page: 0, scale: 1.0, ..Default::default() }); + assert!(out.error.is_none(), "{:?}", out.error); + let at = (y * out.width as usize + x) * 4; + *out.rgba.get(at).expect("the pixel lies on the page") +} + +#[test] +fn a_wide_stroke_of_a_path_with_y_curves_is_drawn() { + // The path's exact coordinates are the trigger; `cm` moves the point where its two curves + // meet, (18.9, 236.3), to about (98.9, 99.7) from the top left, inside every stroke width + // below whatever the caps. The `v` path is the same two curves run backwards, a guard that + // draws with or without the fix. + let y = "21.4166 234.5928 m 19.9009 236.0303 18.9478 236.3428 y 19.2603 238.4209 20.8697 238.4209 y S"; + let v = "20.8697 238.4209 m 19.2603 238.4209 18.9478 236.3428 v 19.9009 236.0303 21.4166 234.5928 v S"; + let mut dropped = Vec::new(); + for path in [y, v] { + for width in [4, 8, 16, 24, 32, 48, 63, 64] { + for (join, cap) in [(0, 0), (1, 1), (2, 2)] { + let content = format!("0 1 0 RG {width} w {join} j {cap} J 1 0 0 1 80 -136 cm {path}"); + let red = red_at(&content, 98, 99); + if red > 64 { + dropped.push(content); + } + } + } + } + assert!(dropped.is_empty(), "strokes not drawn: {dropped:#?}"); +} diff --git a/vendor/README.md b/vendor/README.md index 4733eba49c..b398891efa 100644 --- a/vendor/README.md +++ b/vendor/README.md @@ -46,7 +46,7 @@ Other resource categories retain their existing parent-chain lookup. | hayro-syntax | 0.7.2 | Apache-2.0 OR MIT | (1) Page-tree cycle guard in `resolve_pages` (visited set + depth cap): a `/Kids` loop among compressed objects overflowed the stack and aborted the process; (2) `ccitt_size_ok`: CCITT images wider than 2^20 columns or over 2^28 pixels are not decoded (a fuzzed `/Columns 4294967295` allocated 4 GiB); (3) Flate/LZW predictor row widths are saturating `usize`, a zero width is refused and a row longer than the data is not allocated (a fuzzed `/Columns 9223372036854775807` wrapped to a 2^61-byte allocation and aborted); (4) the inline-image end-of-data check (`EI`) doesn't nest inside its own look-ahead and examines at most 256 candidates and steps (50 images with "EI" in their data ran for over ten minutes); an EI followed by a delimiter (EI%..., EI[...) ends the image like one followed by white space; only a regular character after EI skips the candidate; (5) a stream whose `/Length` is wrong is recovered only when `stream` is the next token — scanning ahead to a later object's `stream` made an indirect appearance-state dictionary (Quartz checkboxes: `/AP /N` → `<< /Yes … /Off … >>`) render as a non-form, so the widget was skipped; (6) JPX rescaling uses an exact integer sample range for 1–32-bit components (32 shifted past u32, and floating-point maxima at 25–31 bits wrapped to zero), and refuses zero-sized rows; (7) `decode_limit`: Flate (through flate2 and the fallback inflater), LZW, RunLength and ASCII85 never write past `MAX_DECODED_STREAM` (256 MiB), or, for the last filter of an image of known size, past twice its pixel bytes (a predictor tag byte per byte), at most 1 GiB; `Stream::decoded_within` for budgeted callers; a page's `/Contents` array shares one `MAX_DECODED_STREAM` (a 2 MB page whose content stream or 8 × 8 image inflated to 2 GB took the renderer past 3 GB in under a second; naming the stream twenty times in `/Contents`, 5 GB); (8) high-precision numbers (jsPDF MediaBox coordinates with over 19 digits) parse without overflowing the u64 mantissa: fractional digits past it are dropped, integer digits past it scale the value (a real, kept finite); (9) large dictionary indexes are built on first use and shared across repeated reads within each PDF (#307); the validated-byte cache is bounded, reader contexts remain separate, repairs and inline/content dictionaries bypass sharing, and small dictionaries stay eager; (10) decoded object-stream slots publish once under contention, cold loading rejects recursive dependencies, and xref repair serializes without holding recursive read guards; this permits one parser per native render pool with private worker caches; (11) `sync.rs`: a poisoned lock is recovered instead of unwrapped (a worker panic no longer cascades into every reader of the same PDF), and xref repair takes a blocking write lock instead of asserting a non-blocking one; (12) `SegmentList` (`util.rs`) returns `None` for an out-of-range index instead of panicking; (13) a page's `/UserUnit` scales `render_dimensions` and `initial_transform`, so the page and everything on it are sized in points, as Acrobat shows them; it is read as Acrobat reads it (the page's own entry, not inherited; a number from 1 up, at most 75,000; anything else counts as 1), and a page without it keeps exactly the transform it had; (14) revision 2–4 passwords are tried as given and then in PDFDocEncoding (`legacy_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Convert to PDFDocEncoding`): passwords arrive as UTF-8 text, so a password with characters outside ASCII (`Çok güzel`, `ılık€`) written in PDFDocEncoding, as ISO 32000-2 §7.6.4.3.2 asks and qpdf and Acrobat write it, was always wrong; (15) page content is read a piece at a time (`page::ContentReader`, `Page::content_reader`): plain streams and Flate streams without a predictor are decoded as they are read, with a budget of 64 times their compressed size and never less than `MAX_DECODED_STREAM` (a stream that expands past it is cut and reported); other streams are decoded whole, as before, and share `MAX_DECODED_STREAM`. The content tokenizer reads a window at a time (`UntypedIter::new_window`, `TypedIter::needs_more`, `TypedIter::committed`): an instruction that runs past the window is not decided until more data follows, which `Reader` records by flagging reads and `at_end` past the end of the data, including through the clones the parser makes. The Flate decoder keeps its own 32 KiB history, so it pauses at any output size. `ContentReader::new` reuses raw/decrypted stream bytes for budget accounting and piece creation, and `decoded_raw_within` applies the existing Whole-stream filters and limits to those bytes without decrypting twice; (16) revision 5–6 (AES-256) passwords are tried in their SASLprep form (RFC 4013, then UTF-8, at most 127 bytes, as ISO 32000-2 §7.6.4.3.3 Algorithm 2.A asks and pdfcraft-crypt writes them) and then, if that differs, as given (`aes256_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Do the above.`; `stringprep`, already in the tree through pdfcraft-crypt, is an optional dependency enabled by `std`, and without `std` passwords are used as given): the renderer used the password as typed, so one that SASLprep changes (a decomposed `café`, a no-break space, the fullwidth `ab` a Japanese IME types in full-width mode) was wrong, although pdfcraft-cos accepted it: protecting a document with such an open password was rolled back ("the password is incorrect"), and a file with such a password did not open with it.; (17) CCITT `/Rows 0` means the height is not predetermined (ISO 32000-2 Table 11), not zero rows: scanned Group 4 pages written that way rendered blank; (18) The streaming Flate decoder (`lzw_flate::flate::fallback`) copies a back-reference in runs of up to its distance (`extend_from_within`), not byte by byte. The copy reads only bytes already in the output, and a copy whose source does not exist still writes nothing, so the bytes are those a byte-by-byte copy writes. Inflating the A1 sample's content stream (278.6 MB) takes about 14% less time (0.66 s to 0.56 s); the output is identical; (19) CCITT `/Rows` below the image's `/Height` decodes to the height: a plotter driver wrote `/Rows` equal to `/Columns`, so only the first rows were decoded and the rest drawn black (#3308) (20) `stream` followed by spaces or tabs and then the line break (`stream \r\n`, as Microsoft Reporting Services writes it) starts the data after the line break; the content stream was taken to start at the space and the page rendered blank.; (21) a JP2 colour specification naming an enumerated colour space the decoder refuses (CIEJab in a veraPDF fixture, YCbCr, PhotoYCC, YPbPr, bilevel…) no longer fails the image: it is decoded again from a copy whose `colr` names the space the decoder assumes without one, greyscale, sRGB or CMYK by component count, as MuPDF and Poppler draw it (`without_refused_enumerated_colour`; the page drew blank, #3306); (22) an object stream's index of `objnum offset` pairs is parsed once per stream and kept beside its decoded data (`ObjectStreamIndexes`, `XRef::with_object_stream_indexes`), and an offset past `usize` is malformed instead of wrapping: `XRef::get_with` parsed the whole index again for every object it fetched from the stream, so reading all of a stream of N objects took N² steps (a Firefox page whose 65,542 objects share one stream took 37 s to draw and 40 s to extract its text); (23) the inline-image end-of-data look-ahead no longer rejects a candidate `EI` because a byte above 0x7F lies between it and the next one: when the candidate is the end, those bytes are the content stream that follows the image, and its strings carry such bytes (WinAnsi `(Grüße) Tj`, the two-byte codes of any Type 0 font), so the operator check that follows decides alone, bounded as in (4) (a scanned page whose OCR text held an accented word and a later capital "EI" — EINGANG, BEI, KLEIN — drew and extracted nothing after the image: 40 of 275 pages across five real documents); (24) `xref::prev_offset`: a trailer `/Prev` written as a reference `N G R` is read as the offset that object holds when the section's own entries place it (iText's `indirectPrev.pdf`; each section's entries are read before its `Prev` is followed), and otherwise as offset N (tiff2pdf; JHOVE's PDF-HUL-157 files, #2366), as `pdfcraft-cos` and other readers take it, so the earlier sections' pages are seen instead of one; (25) direct arrays and dictionaries share a reader-local nesting cap of 128, counted while skipping and while parsing a dictionary (its lenient repair reads, rather than skips, an object found where a key belongs, so `<< << …` recursed through that read); deeper objects return `None` instead of overflowing the stack (including unused trailer values, trailer entries, content operands and marked-content properties); a chain of 65 direct stitching functions now stops at this cap, before the 64-function guard; (26) an xref stream subsection must be backed by its decoded data: all-zero `/W` widths (an entry reads no bytes) and a declared count longer than the data are refused before any entry is inserted, where a tiny stream with a large `/Size` or `/Index` count inserted that many entries (from storytold/craft-libs#239); a subsection running past object number `i32::MAX` is refused instead of overflowing or turning negative; (27) an xref stream's type field is read big-endian at its full `/W[0]` width, where only its first byte was read and wider fields shifted every later entry, and `/W[0]` over 8 bytes is refused like `/W[1]` (from storytold/craft-libs#241); (28) a content instruction keeps up to 256 operands (`MAX_OPERANDS`) instead of 10, and past that drops the oldest instead of ending the content stream: the eleventh operand of a twelve-ink DeviceN `scn` (ISO 32000-2 §8.6.6.5 allows any number of components) blanked the rest of the page (#4260); (29) a `/JBIG2Globals` stream that is itself JBIG2 isn't decoded: globals are raw segments, and a chain of 16,000 such streams overflowed the stack, which aborts the process; (30) a stream's own crypt filter: `/Crypt` in its chain read as the filter it names, `/Identity` without one, instead of the document's filter and then a filter that can't be decoded, which left such a stream unreadable in any spelling (ISO 32000-2 §7.4.10); (31) `set_page_jbig2_work`: the JBIG2 images decoded on a thread spend one allowance, set per page, 2^30 steps (`MAX_PAGE_JBIG2_WORK`), so a page of many images made to reach their own limit stops at one, and what they cost to read counts, whether they decode or not; unbounded until a page sets it; (32) a group of array items read together (a tuple: an image's `/Decode`, `/W`, `/W2`, and a function's `/Domain`, `/Range`, `/Encode` and `/Decode`) can start with a reference: it was taken for the whole group and lost, which ended the array, so an image with `/Decode [7 0 R 0]` didn't draw and CID widths fell back to `/DW`; (33) an object that holds only a reference stands for what that names, stored on its own or in an object stream, up to 32 references: read as a number it was the reference's own object number, so a `/Decode` or `/W` entry naming such an object took a wrong value, and as anything else it was lost. Past 32, as any cycle goes, it reads as a missing object. An object stream isn't reached that way, as its checks are of the object named; reading its index counts against the 64 nested loads and reads its dictionary alone, as a chain of them, each with a member of the next for its `/N`, `/First` or `/Length`, ran out of stack; and a comment inside a reference counts as white space; (34) a table rebuilt by scanning the file (`fallback_xref_map`) finds an object by its number when its header names another generation, including one no generation may be (`3 18446744073709551616 obj` in `poppler-85140-0.pdf`, #5580): the header is read leniently only in that scan (`read_object_header`), the object is also filed under generation 0 unless a real `N 0 obj` holds that key (a later header for the number moves the stand-in), `XRef::get_with` matches by object number only in a repaired table, and `XRef::objects` yields each location once; (35) PNG predictors (`lzw_flate::unfilter_png`) are undone a byte row at a time on slices, as the PNG filters are defined, instead of through the bit-level readers TIFF predictors need: a 993 × 936 RGB image took about 63 ms to unfilter whatever its row filters, and now takes about 3 ms (Up) to 16 ms (Paeth) in a release build (#5430); a row ending part-way through a pixel (three 4-bit samples, an odd width) keeps its last byte, where the bit-level readers dropped it and each later row started a byte early, shearing the picture; (42) TIFF predictor 2 on 8- and 16-bit samples (`lzw_flate::undo_tiff_bytes`) is undone a byte row at a time on slices, each sample added to the same component of the pixel to its left, wrapping at the sample's width (16-bit samples big-endian), instead of through the bit-level readers 1-, 2- and 4-bit samples still need; the bytes are those the bit-level path wrote, and a partial last row is still dropped. A page of 45 full-page 2445 × 4724 8-bit grey images so predicted (PDFBOX-1350) took about 23 s to render at 96 dpi and now takes about 4 s, with the same pixels | (1) `cos/tests/writer.rs::page_tree_loops_in_object_streams_do_not_crash_readers`; (2) `raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding`; (3) `raster::tests::absurd_predictor_columns_and_line_widths_render`; (4) `raster::tests::inline_images_with_ei_in_their_data_parse_in_linear_time`, `content::tests::inline_image_ei_may_be_followed_by_a_delimiter`; (5) `raster::tests::indirect_appearance_state_dict_is_not_another_stream`; (6) `raster::tests::jpx_stream_bit_depth_boundaries_do_not_panic`, `raster::tests::jpx_32_bit_image_does_not_prevent_other_page_content_from_rendering`, `raster::tests::jpx_gray_rgb_and_alpha_keep_eight_bit_samples`, `raster::tests::jpx_streams_reject_out_of_range_bit_depths`; (7) `raster::tests::decompression_bombs_stop_at_the_stream_limit`, `raster::tests::page_contents_and_nested_forms_share_one_budget`; (8) `inspect::tests::jspdf_high_precision_mediabox_does_not_overflow_dimensions`, `object::number::tests::high_precision_float_does_not_overflow_mantissa` (hayro-syntax); (9) `object::dict::lazy_index_tests`, `raster::tests::large_resource_indexes_preserve_pixels_and_text`; (10) `data::tests`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`, `raster::tests::shared_parser_is_lazy_and_initializes_once_under_contention`; (11) `data::tests::panicking_initializer_wakes_waiters_and_prevents_retries`, `xref::tests::concurrent_cold_object_stream_reads_all_succeed`; (12) `util::tests::segment_list_rejects_out_of_range_indices`; (13) `raster::tests::user_unit_scales_the_page_and_everything_on_it`; (14) `crypto::pdfcraft_password_tests`, `pdfcraft-engine` `tests::legacy_passwords_with_turkish_characters_open`; (15) `content::tests::windows_keep_every_instruction_at_every_split`, `content::tests::windows_keep_every_instruction_in_pieces_of_any_size`, `content::tests::windows_match_the_slice_on_random_content`, `content::tests::windows_drop_blanks_and_keep_a_comment_that_may_go_on`, `page::content_reader_tests::content_past_the_cap_is_read_to_the_end`, `page::content_reader_tests::a_stream_that_expands_past_the_budget_is_cut_there`, `page::content_reader_tests::a_small_expansion_under_the_cap_is_not_cut`, `raster::tests::page_content_past_the_old_cap_renders`, `raster::tests::content_reader_returns_the_same_bytes_in_any_pieces`, `raster::tests::content_bombs_stop_at_the_budget_and_say_so`, `page::content_reader_tests::plain_and_whole_pieces_keep_their_bytes`, `page::content_reader_tests::repeated_content_reference_counts_compressed_bytes_once`, `page::content_reader_tests::whole_cap_still_counts_later_unique_streams`; (16) `crypto::pdfcraft_password_tests::aes256_passwords_are_tried_in_their_saslprep_form_then_as_given`, `crypto::pdfcraft_password_tests::aes256_passwords_open_as_typed_and_in_their_saslprep_form`, `pdfcraft-engine` `tests::aes256_open_passwords_that_saslprep_changes_protect_and_reopen_as_typed`, `tests::aes256_files_whose_passwords_saslprep_changes_open_as_typed`; (17) `raster::tests::group4_fax_with_rows_zero_decodes_to_the_image_height`; (18) `filter::lzw_flate::tests::back_reference_longer_than_its_distance_repeats_its_source`, `back_reference_cut_by_the_limit_resumes_in_the_next_call` and `back_reference_reads_bytes_written_by_an_earlier_call`; each fails when the copy returns the wrong remaining count; (19) `raster::tests::group4_fax_with_rows_below_its_height_decodes_to_the_image_height`; (20) `raster::tests::space_before_the_line_break_after_stream_still_renders`; (21) `raster::tests::a_jp2_image_in_a_refused_enumerated_colour_space_is_drawn`, `filter::jpx::tests::refused_enumerated_colour_spaces_become_the_component_counts_default`, `filter::jpx::tests::what_the_decoder_handles_or_cant_be_read_is_left_alone` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (22) `raster::tests::objects_in_one_large_object_stream_are_read_in_linear_time`; (23) `text::tests::text_after_an_inline_image_survives_a_non_ascii_byte_before_a_later_ei`, `content::tests::an_inline_image_ends_at_its_own_ei_when_later_text_holds_a_non_ascii_byte` (hayro-syntax); (24) `xref::tests::prev_as_reference_reaches_the_earlier_section`, `xref::tests::prev_as_reference_to_an_object_holding_the_offset_reads_that_offset`, `pdfcraft-cos` `document::tests::a_prev_written_as_a_reference_still_reaches_the_earlier_sections`, `document::tests::a_prev_reference_to_an_object_holding_the_offset_reads_that_offset`; (25) `raster::tests::deeply_nested_direct_objects_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_content_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_in_the_trailer_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_as_marked_content_properties_do_not_overflow_the_stack`, `raster::tests::dictionaries_nested_without_keys_are_read_up_to_the_nesting_cap`, `raster::tests::nesting_at_the_cap_fits_a_512_kib_stack` (native only), `raster::tests::function_limits_depth_guard_refuses_a_chain_of_references`; (26) `xref::tests::xref_stream_entries_need_backing_bytes`; (27) `xref::tests::xref_stream_type_field_uses_the_full_w0_width`, `xref::tests::xref_stream_loader_rejects_oversized_w0` (`cargo test --manifest-path vendor/hayro-syntax/Cargo.toml`); (28) `content::ops::tests::scn_with_twelve_components_keeps_them_all`, `content::ops::tests::operands_past_the_cap_keep_the_latest_in_order` (hayro-syntax), `raster::tests::devicen_inks::a_twelve_ink_devicen_colour_paints`, `raster::tests::devicen_inks::a_four_ink_devicen_colour_paints`, `raster::tests::devicen_inks::an_operator_with_hundreds_of_operands_does_not_end_the_content`; (29) `raster::tests::jbig2_globals_that_are_jbig2_themselves_are_not_decoded`, `raster::tests::jbig2_globals_are_read`, `raster::tests::the_jbig2_images_of_a_page_share_one_allowance`, `raster::tests::an_image_that_fails_spends_the_pages_allowance_too`; (30) `raster::tests::a_stream_with_a_crypt_filter_of_its_own_is_drawn`; (32) `raster::tests::a_decode_array_with_references_in_it_is_read`, `raster::tests::a_functions_arrays_can_start_with_a_reference` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret); (33) `raster::tests::a_reference_to_a_reference_is_what_it_names`, `raster::tests::a_decode_array_with_references_in_it_is_read` (pdfcraft-render), `font::cid::width_tests::a_width_group_can_start_with_a_reference` (hayro-interpret), `xref::tests::an_object_stream_isnt_reached_through_a_reference`, `xref::tests::a_chain_of_object_streams_ends_at_the_limit_of_nested_loads`, `xref::tests::a_chain_of_object_streams_takes_no_more_work_for_two_ways_on`, `xref::tests::a_reference_held_in_either_storage_is_followed`, `xref::tests::a_page_a_reference_is_held_to_is_found_once` (hayro-syntax); (34) `raster::tests::a_rebuilt_table_finds_an_object_whose_header_names_another_generation`, `raster::tests::a_rebuilt_tables_generation_0_stand_in_is_not_a_second_object` (pdfcraft-render), `xref::tests::an_object_stream_and_its_first_member_are_both_enumerated` (hayro-syntax); (35) `filter::lzw_flate::pdfcraft_png_predictor_tests` (hayro-syntax: every filter type at 1–16 bits and 1–4 colours against the PNG specification's byte-by-byte definition, a row ending part-way through a pixel, Average and Paeth over 6-byte pixels), `raster::tests::png_predicted_images::a_four_bit_rgb_image_of_odd_width_with_up_predicted_rows_is_not_sheared`; (42) `filter::lzw_flate::pdfcraft_tiff_predictor_tests` (hayro-syntax: random 8- and 16-bit rows at 1–7 colours and 1–50 columns against the bit-level path, partial and empty rows, and a row counter showing whole-byte samples skip the bit-level path), `raster::tests::tiff_predicted_images` | | winit | 0.30.13 | Apache-2.0 | (1) on Windows 11 (build 22000 and later), `WM_DPICHANGED` applies the rect Windows suggests. It used to nudge the window onto the monitor `MonitorFromWindow` names, which during a drag is still the monitor being left; the nudge sent another `WM_DPICHANGED`, so a window dragged to a monitor with another scale factor grew and stayed on the wrong monitor at the wrong scale (#324). Backport of winit master 488c036a ([winit#4341](https://github.com/rust-windowing/winit/pull/4341)), which no 0.30 release has; Windows 10 keeps the old adjustment, as on winit master. (2) on Wayland, a seat's text input is destroyed when the seat goes away, not whenever it loses a capability: losing any capability (a touch screen or tablet unplugged, a remote-desktop session ending) destroyed it, and only an added capability recreated it, so IME stopped working until restart. Port of [winit#4747](https://github.com/rust-windowing/winit/pull/4747); (3) on Windows, an IME composition reports its cursor from `GCS_CURSORPOS` only when `WM_IME_COMPOSITION` says it is set, and at the end of the composition otherwise: the Microsoft Korean IME never sets it, so the cursor read as 0 (egui then drew no caret while composing; it now draws it after the syllable). Port of [winit#4746](https://github.com/rust-windowing/winit/pull/4746); (4) on Windows, a hidden window is not maximized until it is shown, and a hidden window shown while maximized gets one `ShowWindow(SW_MAXIMIZE)`: `SW_MAXIMIZE` shows a hidden window, so PdfCraft, which eframe keeps hidden until its first frame is drawn, appeared unpainted (white) when it restored a maximized window, and could lose its focus. Port of [winit#4587](https://github.com/rust-windowing/winit/pull/4587) (the later, focus-safe form of [winit#4575](https://github.com/rust-windowing/winit/pull/4575)); (5) on X11, when allowing or disallowing IME replaces the focused window's input context (egui does it each time a text field gains or loses focus), the new context is focused: no focus event follows, so it stayed unfocused until the window lost and regained focus. Port of [winit#4727](https://github.com/rust-windowing/winit/pull/4727); (6) on macOS, `insertText` commits through IME when a composition was started (`pending_commit`, set by `setMarkedText`) rather than when marked text is left: input methods that end a composition with `unmarkText` before `insertText` lost the text, and Apple Korean dropped the key typed after a syllable (`한` then `5` or `;`); other `insertText` calls stay `KeyboardInput`; a commit outside a key press (a candidate clicked with the mouse, dictation) clears the marked text and leaves no key to swallow, as in [winit#4748](https://github.com/rust-windowing/winit/pull/4748) (without it the next key, such as Backspace, was lost). `validAttributesForMarkedText` lists the underline and marked-clause attributes, without which some input methods don't compose. Port of [winit#4650](https://github.com/rust-windowing/winit/pull/4650), the 0.30 backport of [winit#4651](https://github.com/rust-windowing/winit/pull/4651) (merged on winit master), with the underline attribute spelled as AppKit's `NSUnderlineStyleAttributeName` (`NSUnderline`); (7) on macOS, disallowing IME with a composition in progress (focus leaving a text field) tells the input method to discard its marked text (`discardMarkedText`): it kept composing it once IME was allowed again, so the abandoned syllable came back (`하`, then `가` typed in the next field, became `학`, then `하` + `가`; Japanese `か` then `さ` became `傘`). Port of [winit#4745](https://github.com/rust-windowing/winit/pull/4745); (8) on macOS, text inserted outside a key press without a composition (the emoji picker) is reported as `Ime::Commit`; it was dropped. Port of the emoji-picker part of [winit#4749](https://github.com/rust-windowing/winit/pull/4749) (0.30 backport of [winit#4748](https://github.com/rust-windowing/winit/pull/4748)); its `selectedRange` change for dictation is left out because it switches on press-and-hold for every held key (an accent menu instead of key repeat, and the chosen accent added after the repeated letters); (9) on macOS, a key whose text was committed through IME isn't also sent as `KeyboardInput` unless AppKit forwards it as a command (Enter): Apple Korean answers Space with no syllable in progress with `setMarkedText(" ")`, `insertText(" ")`, `setMarkedText("")`, and the last call reset winit's state, so the space arrived twice. Same goal as [winit#4478](https://github.com/rust-windowing/winit/pull/4478), whose changes don't reach this call order (macOS 14); to be proposed upstream there; (10) on macOS, when the first key in a fresh app reaches Apple Korean's `insertText` as a raw compatibility jamo instead of starting a composition, that key is interpreted once more, synchronously, before it is sent as `KeyboardInput` (g k s typed ㅎㅏㄴ instead of 한); only an unmodified, non-repeated first key whose one callback inserts that same jamo, with IME enabled and nothing composed, focus and input source unchanged, is retried (`ime_startup.rs`). Port of [winit#4744](https://github.com/rust-windowing/winit/pull/4744), with its policy tests.; (11) on X11 without `Xft.dpi`, the scale factor comes from the X screen's configured DPI (its size in pixels and millimetres) before the output's physical density from RandR: Xorg can deliberately run a dense panel at 96 DPI, and only winit applications then looked zoomed; `Xft.dpi` still wins, `WINIT_X11_SCALE_FACTOR` still overrides, and a screen reporting no size keeps the RandR fallback. PdfCraft patch (#1150, replacing the launcher pin proposed in #908). (12) on macOS, `WinitView` answers `mouseDownCanMoveWindow` with the window's `movable_by_window_background` attribute (NO unless asked): NSView answers YES for a view that isn't opaque, so with the transparent title bar and full-size content view PdfCraft uses, AppKit moved the window for any drag that began in the 28 pt title-bar band before the app saw it, and document tabs (which live in that band) could not be dragged to reorder them. The empty strip still moves the window through `ViewportCommand::StartDrag`. PdfCraft patch. The published crate's asset files (`docs/res/*.svg`, `examples/data/*.png`) are left out, so its `window` example doesn't build; (13) on Wayland, pixel-only wheel, wheel-tilt and continuous-source scroll events report `Moved` rather than starting an unterminated finger gesture, so egui follows modifier release instead of keeping Ctrl zoom latched (#991). Finger gestures and unknown sources retain their existing phase policy | (1) `pdfcraft` `tests::winit_carries_the_windows_11_monitor_scale_fix` keeps the patch in the build; the drag itself needs two monitors with different scale factors on Windows 11; (2) `pdfcraft` `tests/winit_wayland_text_input.rs`; the capability change itself needs a Wayland session and an input method; (3) `pdfcraft` `tests/winit_windows_ime_cursor.rs`; composing itself needs the Microsoft Korean IME; (4) `pdfcraft` `tests/winit_windows_hidden_maximized.rs`; the start-up itself needs Windows and a window that was closed maximized; (5) `pdfcraft` `tests/winit_x11_ime_focus.rs`; switching fields itself needs an X11 session and an input method (IBus, fcitx5); (6) `pdfcraft` `tests/winit_macos_ime_commit.rs`; composing itself needs macOS and an input method; (7) `pdfcraft` `tests/winit_macos_ime_discard.rs`; composing itself needs macOS and an input method; (8) `pdfcraft` `tests/winit_macos_emoji_picker.rs`; the picker itself needs macOS; (9) `pdfcraft` `tests/winit_macos_korean_space.rs`; composing itself needs macOS and Apple Korean; (10) `pdfcraft` `tests/winit_macos_korean_first_key.rs` and `ime_startup.rs`'s own tests (run them with `rustc --test`); typing itself needs macOS and Apple Korean; (11) `pdfcraft` `tests/winit_x11_dpi.rs` keeps the patch in the build, and `util::randr` tests the policy (run them in the vendored crate); the scale itself needs an X11 session; (12) `pdfcraft` `tests/winit_macos_title_bar_drag.rs` keeps the patch in the build; the drag itself needs macOS and a pointer; (13) `pdfcraft` `tests/winit_wayland_scroll.rs` compiles the production phase policy and exercises modifier release through egui; live COSMIC confirmation remains open | | egui-wgpu | 0.36.2 | MIT OR Apache-2.0 | (1) `winit::surface_fit`: a window's surface is configured within the device's `max_texture_dimension_2d`, both sides shrunk by one factor that egui then draws at (`pixels_per_point × render_scale`), so the whole window is drawn into the surface (on DX12 it is stretched over the window and lines up with the pointer; wgpu's GL backend copies it unscaled into a corner); the longer side becomes exactly the limit; the depth and MSAA textures follow, and a recreated surface starts again from the window's own size. egui-wgpu configured the window's size as is, and `Surface::configure` panics beyond the limit: a window restored at another monitor's scale (opening an Outlook attachment on a mixed-DPI laptop asked an 8192-pixel device for 8600 × 3423, #577) or stretched across monitors closed PdfCraft, before its window appeared when it happened at start-up ([emilk/egui#8361](https://github.com/emilk/egui/issues/8361)). The published crate leaves out egui's licence files; `LICENSE-APACHE` and `LICENSE-MIT` are from the 0.36.2 tag; (2) `winit::catch_errors`: a new window's surface is configured inside wgpu error scopes (validation, out of memory, internal), and a failure is returned as `WgpuError::ConfigureSurface` (for the root window at start-up eframe returns it as `Error::Wgpu`, which PdfCraft retries with OpenGL); nothing is kept for the window. wgpu's default handler panicked instead, and eframe configures the root surface before the app is created, so a driver that can't configure one closed PdfCraft at launch (a 2015 Intel GPU through wgpu's GL backend timed out waiting for the GPU to go idle, `GpuWaitTimeout`, #519). Unchanged, so still a panic: an error in a later reconfigure (resize, a lost surface), a device lost during the first configure (wgpu reports it only to a callback), and a window shown at zero size (first configured on resize) | (1) `pdfcraft` `tests::a_window_larger_than_the_gpu_limit_gets_a_surface_within_it`, `tests::surface_fit_never_empties_or_overflows_a_surface`, and `tests::vendor_provenance_egui_wgpu_surface_fit_identity`, which keeps the patch in the build and in the paths that size a surface; (2) `pdfcraft` `tests::a_gpu_error_while_a_window_is_set_up_is_returned_not_a_panic` (a real device: an error inside `catch_errors` is returned, not a panic) and `tests::vendor_provenance_egui_wgpu_catch_errors_identity`, which keeps it around `add_surface`'s configure | -| vello_common | 0.0.8 | Apache-2.0 OR MIT | (1) `clip::intersect` (the clip step every clipped path passes through) starts its row walk at the first row both paths have strips in, not the earlier of the two first rows. The walk visited every row from the top of the clip for each path (on the A1 sample's full-page canvas, about half its strip rows per stroke), and rows where one path has no strips produce no output, so the strips are unchanged (vello_common 0.1 starts at the later row too). Its cost grew with the page, and a full-page render of the A1 sample was about 30% faster; the output is byte-identical | (1) `clip::tests::intersect_walks_only_rows_both_paths_have_strips` counts the rows the walk visits (`cargo test --manifest-path vendor/vello_common/Cargo.toml`) and fails without the patch; the render crate's pixel tests run through the patched crate | +| vello_common | 0.0.8 | Apache-2.0 OR MIT | (1) `clip::intersect` (the clip step every clipped path passes through) starts its row walk at the first row both paths have strips in, not the earlier of the two first rows. The walk visited every row from the top of the clip for each path (on the A1 sample's full-page canvas, about half its strip rows per stroke), and rows where one path has no strips produce no output, so the strips are unchanged (vello_common 0.1 starts at the later row too). Its cost grew with the page, and a full-page render of the A1 sample was about 30% faster; the output is byte-identical; (4) `flatten::stroke`: when kurbo 0.13.1's stroker returns an outline containing NaN, which `fill` would drop whole, the path is stroked again from a copy (`robust_for_stroking`, one element out per element in) in which each curve whose control points lie within the stroke tolerance of its chord is a line (the #4907 fix, a nearly straight cubic given a NaN join, kurbo issues #617 and #519; its row here was lost in a merge and is restored with this entry) and each cubic whose second control point is its end point, as `y` builds (ISO 32000-2 §8.5.2.1 Table 58), has that control point moved the tolerance towards its first (`open_zero_end_handle`; the curve moves by at most 4/9 of the tolerance): kurbo's cusp search on such a curve's offset can stop on the end point, where the derivative is zero, and normalising it gives NaN, so at 16, 32 and 63 pt a two-curve `y` path drew nothing and a file in the PDFBox tracker lost its 32 pt "XX" outline. Only an exact zero triggers it (none of 192,000 random strokes with a handle 1e-9 long, nor of 256,000 `v` strokes, came back with NaN; 1,073 of 192,000 random `y`/`v` strokes at 1–64 pt did before and none after). A finite outline takes the old path unchanged | (1) `clip::tests::intersect_walks_only_rows_both_paths_have_strips` counts the rows the walk visits (`cargo test --manifest-path vendor/vello_common/Cargo.toml`) and fails without the patch; the render crate's pixel tests run through the patched crate; (4) `flatten::tests::a_tiny_degenerate_cubic_does_not_drop_the_stroke` and the render crate's `stroke_degenerate_cubic` fail without the near-line replacement, `flatten::tests::a_wide_stroke_of_a_curve_with_a_zero_end_tangent_is_drawn` (widths 1–64, every join and cap, `y` and the mirrored `v`) and the render crate's `stroke_zero_end_tangent` fail without the moved control point; `a_finite_outline_is_stroked_as_before` and `a_path_with_nan_is_still_dropped` pass either way | | hayro-cmap | 0.1.0 | Apache-2.0 OR MIT | (1) malformed parts of a CMap (bad bfchar/bfrange entries, unresolvable `usecmap`, unparseable tail) are skipped instead of discarding the whole map, so search and redaction find text the edit layer can read (#3177); (2) a binary CMap's `usecmap` counts against the 16-level nesting limit as a text one does, where it passed its own depth and a self-referencing binary CMap recursed until the stack overflowed (from storytold/craft-libs#254) | (1) `hayro-cmap::tests::bad_bfchar_entry_is_skipped`, `unresolvable_usecmap_is_skipped`; (2) `tests::self_referencing_binary_usecmap_stops_at_the_nesting_limit` | | hayro-postscript | 0.1.0 | Apache-2.0 OR MIT | (1) the token scanner's `object::read` skips `<<` and `>>` in a loop; it called itself once per delimiter, so a CMap stream holding a long run of `<` or `>` overflowed the stack and aborted the process (from storytold/craft-libs#263) | (1) `hayro-postscript::object::tests::long_open_dict_run_does_not_overflow`, `long_close_dict_run_does_not_overflow`, `dict_delimiters_are_skipped`, `nested_dict_delimiters_are_skipped` | diff --git a/vendor/vello_common/src/flatten.rs b/vendor/vello_common/src/flatten.rs index afacf7ecbc..25697c7032 100644 --- a/vendor/vello_common/src/flatten.rs +++ b/vendor/vello_common/src/flatten.rs @@ -240,8 +240,11 @@ pub fn stroke( // PdfCraft patch: kurbo's stroker can put NaN into the outline of a curve that is nearly a // line (all control points within a rounding error of the chord), and `fill` then drops the // whole outline. Stroke it again with such curves as lines. A finite outline is unchanged. + // PdfCraft patch (4): it can also put NaN into the offset of a cubic whose second control + // point is its end point (as `y` builds), so that control point is moved off it in the same + // pass. if !stroke_ctx.output().is_finite() { - let path = near_line_curves_as_lines(path, tolerance); + let path = robust_for_stroking(path, tolerance); expand_stroke(&path, style, tolerance, stroke_ctx); } fill( @@ -256,8 +259,10 @@ pub fn stroke( // PdfCraft patch. /// `path` with each quadratic or cubic whose control points all lie within -/// `tolerance` of its chord replaced by a line to its end point. One element out per element in. -fn near_line_curves_as_lines(path: impl IntoIterator, tolerance: f64) -> BezPath { +/// `tolerance` of its chord replaced by a line to its end point, and (patch (4)) each other +/// cubic whose second control point is its end point given a short end handle instead +/// ([`open_zero_end_handle`]). One element out per element in. +fn robust_for_stroking(path: impl IntoIterator, tolerance: f64) -> BezPath { let mut out = BezPath::new(); let mut start = kurbo::Point::ZERO; let mut last = start; @@ -270,6 +275,9 @@ fn near_line_curves_as_lines(path: impl IntoIterator, tolerance: { PathEl::LineTo(p3) } + PathEl::CurveTo(p1, p2, p3) => { + PathEl::CurveTo(p1, open_zero_end_handle(last, p1, p2, p3, tolerance), p3) + } el => el, }; match el { @@ -285,6 +293,33 @@ fn near_line_curves_as_lines(path: impl IntoIterator, tolerance: out } +// PdfCraft patch (4). +/// The second control point of the cubic `p0 p1 p2 p3`, moved `tolerance` towards `p1` (or `p0` +/// when `p1` is that close too) when it sits on the end point `p3`, as a `y` curve's does. +/// +/// There the curve's derivative vanishes at its end. kurbo finds the cusps of a wide stroke's +/// offset curve with a root search on the curvature, which grows without bound towards such an +/// end; at some widths the search returns the end point itself, and the unit tangent kurbo takes +/// there is `0 / 0`. Only an exact zero does this: a handle even 1e-9 long strokes finitely, as +/// does a first control point on the start point (`v`). The moved control point moves the curve +/// by at most `4 / 9 * tolerance`. +fn open_zero_end_handle( + p0: kurbo::Point, + p1: kurbo::Point, + p2: kurbo::Point, + p3: kurbo::Point, + tolerance: f64, +) -> kurbo::Point { + if p2 != p3 { + return p2; + } + [p1, p0] + .into_iter() + .map(|q| q - p3) + .find(|v| v.hypot() > tolerance) + .map_or(p2, |v| p3 + v * (tolerance / v.hypot())) +} + /// Whether `p` is within `tolerance` of the segment from `a` to `b`. False for non-finite input. fn near_segment(p: kurbo::Point, a: kurbo::Point, b: kurbo::Point, tolerance: f64) -> bool { let ab = b - a; @@ -390,6 +425,63 @@ mod tests { assert!(lines.is_empty()); } + /// The outline of `path` stroked with `style`, drawn translated by (80, -136) so that the + /// test paths below land inside the 200 × 200 cull box, and whether kurbo's outline was finite. + fn stroke_style_lines(path: &BezPath, style: &Stroke) -> (Vec, bool) { + let mut lines = Vec::new(); + let mut ctx = StrokeCtx::default(); + stroke( + Level::baseline(), + path, + style, + Affine::translate((80.0, -136.0)), + &mut lines, + &mut FlattenCtx::default(), + &mut ctx, + RectU16::new(0, 0, 200, 200), + ); + let finite = ctx.output().is_finite(); + (lines, finite) + } + + // A file in the PDFBox issue tracker strokes "XX" outlines at 32 pt whose subpaths include + // `y` curves (the second control point on the end point, so the end tangent vanishes), and at + // 16, 32 and 63 pt kurbo's offset of one of them came back with NaN, so the whole outline was + // dropped. These exact coordinates are the trigger. `v` (the first control point on the start + // point) is the mirrored case, the same two curves run backwards; kurbo strokes it finitely, + // and it is checked as a guard. + #[test] + fn a_wide_stroke_of_a_curve_with_a_zero_end_tangent_is_drawn() { + use crate::kurbo::{Cap, Join}; + let a = Point::new(21.4166, 234.5928); + let b = Point::new(18.9478, 236.3428); + let e = Point::new(20.8697, 238.4209); + let mut y = BezPath::new(); + y.move_to(a); + y.curve_to(Point::new(19.9009, 236.0303), b, b); + y.curve_to(Point::new(19.2603, 238.4209), e, e); + let mut v = BezPath::new(); + v.move_to(e); + v.curve_to(e, Point::new(19.2603, 238.4209), b); + v.curve_to(b, Point::new(19.9009, 236.0303), a); + + let mut failures = Vec::new(); + for (name, path) in [("y", &y), ("v", &v)] { + for width in 1..=64 { + for join in [Join::Miter, Join::Round, Join::Bevel] { + for cap in [Cap::Butt, Cap::Round, Cap::Square] { + let style = Stroke::new(f64::from(width)).with_join(join).with_caps(cap); + let (lines, finite) = stroke_style_lines(path, &style); + if lines.is_empty() || !finite { + failures.push((name, width, join, cap)); + } + } + } + } + } + assert!(failures.is_empty(), "strokes dropped: {failures:?}"); + } + // A regression guard: it passes with or without the patch. #[test] fn a_finite_outline_is_stroked_as_before() { From 90def111c90ef93cc77f52b1444534aba4ac533e Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:31:41 +0800 Subject: [PATCH 09/75] fix(engine): map URL text line rects with view_rect_to_user and filter non-finite coordinates (#6536) * fix(engine): map URL text line rects with view_rect_to_user and filter non-finite coordinates * test(engine): assert concrete rotated user rects and overflow guards (#6469) --- crates/engine/src/session/view.rs | 121 +++++++++++++++++++++++++++--- 1 file changed, 110 insertions(+), 11 deletions(-) diff --git a/crates/engine/src/session/view.rs b/crates/engine/src/session/view.rs index 5ef135d6c8..711779903a 100644 --- a/crates/engine/src/session/view.rs +++ b/crates/engine/src/session/view.rs @@ -120,17 +120,7 @@ impl Session { let rects: Vec<[f64; 4]> = text .line_rects(glyphs) .into_iter() - .map(|v| { - let q = doc.info.pages[page].view_rect_to_quad(v); - let xs = [q[0], q[2], q[4], q[6]]; - let ys = [q[1], q[3], q[5], q[7]]; - [ - xs.iter().copied().fold(f64::MAX, f64::min), - ys.iter().copied().fold(f64::MAX, f64::min), - xs.iter().copied().fold(f64::MIN, f64::max), - ys.iter().copied().fold(f64::MIN, f64::max), - ] - }) + .filter_map(|v| line_user_rect(&doc.info.pages[page], v)) .filter(|r| { !doc.links().iter().any(|l| l.page == page && l.rect[0] < r[2] && r[0] < l.rect[2] && l.rect[1] < r[3] && r[1] < l.rect[3]) }) @@ -313,3 +303,112 @@ impl Session { &self.docs } } + +/// Convert a view-space text line rectangle into a normalized, finite PDF user-space rectangle. +fn line_user_rect(page: &pdfcraft_render::PageInfo, v: [f32; 4]) -> Option<[f64; 4]> { + if !v.iter().all(|x| x.is_finite()) { + return None; + } + let r = page.view_rect_to_user(v).map(f64::from); + r.iter().all(|x| x.is_finite()).then_some(r) +} + +#[cfg(test)] +mod tests { + use super::line_user_rect; + use crate::{Edit, Session}; + use pdfcraft_render::PageInfo; + use std::sync::Arc; + + /// Build a sample `PageInfo` with a non-origin crop box `[10.0, 20.0, 210.0, 320.0]` (`200 × 300`). + fn sample_page(rotation: u16) -> PageInfo { + let (width, height) = if rotation % 180 == 90 { (300.0, 200.0) } else { (200.0, 300.0) }; + PageInfo { width, height, label: String::new(), crop: [10.0, 20.0, 210.0, 320.0], rotation } + } + + #[test] + fn line_user_rect_rejects_non_finite_coordinates() { + let page = sample_page(0); + assert_eq!(line_user_rect(&page, [f32::NAN; 4]), None); + assert_eq!(line_user_rect(&page, [f32::NAN, 10.0, 50.0, 20.0]), None); + assert_eq!(line_user_rect(&page, [10.0, 20.0, f32::INFINITY, 30.0]), None); + assert_eq!(line_user_rect(&page, [10.0, f32::NEG_INFINITY, 50.0, 30.0]), None); + + for bad_crop_box in [[f32::NAN, 0.0, 200.0, 300.0], [0.0, f32::INFINITY, 200.0, 300.0], [0.0, 0.0, f32::NEG_INFINITY, 300.0]] { + let mut bad_crop = sample_page(0); + bad_crop.crop = bad_crop_box; + assert_eq!(line_user_rect(&bad_crop, [10.0, 20.0, 50.0, 40.0]), None); + } + + // Finite inputs whose crop span overflows f32 during view-to-user mapping must also be rejected. + let overflow_page = PageInfo { width: 1.0, height: 1.0, label: String::new(), crop: [-f32::MAX, 0.0, f32::MAX, 300.0], rotation: 0 }; + assert_eq!(line_user_rect(&overflow_page, [1.0, 10.0, 2.0, 20.0]), None); + } + + #[test] + fn line_user_rect_matches_view_rect_to_user_across_rotations() { + let v = [12.0, 24.0, 96.0, 48.0]; + let cases = [ + (0, [22.0, 272.0, 106.0, 296.0]), + (90, [34.0, 32.0, 58.0, 116.0]), + (180, [114.0, 44.0, 198.0, 68.0]), + (270, [162.0, 224.0, 186.0, 308.0]), + ]; + for (rotation, expected) in cases { + let page = sample_page(rotation); + let got = line_user_rect(&page, v).expect("finite view rect maps to user space"); + for (g, e) in got.into_iter().zip(expected) { + assert!((g - e).abs() < 1e-4, "rotation {rotation}: got {got:?}, expected {expected:?}"); + } + } + } + + #[test] + fn find_urls_returns_finite_user_rects_on_rotated_page() { + let mut objs: Vec> = vec![ + b"<< /Type /Catalog /Pages 2 0 R >>".to_vec(), + b"<< /Type /Pages /Kids [4 0 R] /Count 1 /MediaBox [0 0 200 300] >>".to_vec(), + b"<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>".to_vec(), + b"<< /Type /Page /Parent 2 0 R /Contents 5 0 R /Resources << /Font << /F1 3 0 R >> >> >>".to_vec(), + ]; + let body = "BT /F1 12 Tf 20 150 Td (Visit https://example.com today) Tj ET"; + objs.push(format!("<< /Length {} >>\nstream\n{body}\nendstream", body.len()).into_bytes()); + let mut pdf = b"%PDF-1.7\n".to_vec(); + let mut offsets = Vec::new(); + for (i, o) in objs.iter().enumerate() { + offsets.push(pdf.len()); + pdf.extend_from_slice(format!("{} 0 obj\n", i + 1).as_bytes()); + pdf.extend_from_slice(o); + pdf.extend_from_slice(b"\nendobj\n"); + } + let xref = pdf.len(); + pdf.extend_from_slice(format!("xref\n0 {}\n0000000000 65535 f \n", objs.len() + 1).as_bytes()); + for o in offsets { + pdf.extend_from_slice(format!("{o:010} 00000 n \n").as_bytes()); + } + pdf.extend_from_slice(format!("trailer\n<< /Size {} /Root 1 0 R >>\nstartxref\n{xref}\n%%EOF\n", objs.len() + 1).as_bytes()); + + let mut session = Session::new(); + let id = session.open("urls.pdf", None, Arc::new(pdf), None).expect("opens"); + let expected_rect = [46.004, 147.6, 153.38, 159.6]; + let base_urls = session.find_urls(id); + assert_eq!(base_urls.len(), 1); + let base_rect = base_urls[0].1[0]; + for (got, exp) in base_rect.into_iter().zip(expected_rect) { + assert!((got - exp).abs() < 1e-2, "0 deg: got {base_rect:?}, expected {expected_rect:?}"); + } + + for step in [90, 180, 270] { + session.apply(id, Edit::RotatePages { pages: vec![0], degrees: 90 }).expect("rotates"); + let urls = session.find_urls(id); + assert_eq!(urls.len(), 1); + assert_eq!(urls[0].0, 0); + assert_eq!(urls[0].2, "https://example.com"); + assert_eq!(urls[0].1.len(), 1); + let r = urls[0].1[0]; + for (got, exp) in r.into_iter().zip(expected_rect) { + assert!((got - exp).abs() < 1e-2, "rotation {step} deg: got {r:?}, expected {expected_rect:?}"); + } + } + } +} From ca21968e6474d4a573cbb52ef4246694f907cfd5 Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:31:54 +0800 Subject: [PATCH 10/75] fix(js): decode ', numeric, and HTML named character references in FormCalc Decode (#6538) * fix(js): decode ', numeric, and HTML named character references in FormCalc Decode * test(js): verify all HTML named entities and 8-byte name boundary (#6464) * docs(js): remove duplicate decode_entities summary and document LATIN1_ENTITIES order (#6464) --- crates/js/src/formcalc.rs | 38 ++++++++ crates/js/src/formcalc/builtins.rs | 5 +- crates/js/src/formcalc/words.rs | 137 +++++++++++++++++++++++++++++ 3 files changed, 178 insertions(+), 2 deletions(-) diff --git a/crates/js/src/formcalc.rs b/crates/js/src/formcalc.rs index cfad619938..06ff21f438 100644 --- a/crates/js/src/formcalc.rs +++ b/crates/js/src/formcalc.rs @@ -245,4 +245,42 @@ mod tests { assert_eq!(num_text(-1e20), "-100000000000000000000"); assert_eq!(num_text(f64::MAX).len(), 309); } + + #[test] + fn decode_xml_and_html_entities() { + let eval = |script: &str| { + let root = XfaNode::default(); + let out = run_formcalc(script, &XfaEvent::default(), &XfaDoc::default(), &root, Limits::default()); + assert!(out.error.is_none(), "{:?}", out.error); + out.result.unwrap_or_default() + }; + + // `'` in both xml and html + assert_eq!(eval(r#"Decode("It's <ok> & "fine"", "xml")"#), "It's & \"fine\""); + assert_eq!(eval(r#"Decode("It's", "html")"#), "It's"); + + // Decimal and hex numeric character references (case-insensitive #x/#X and hex digits, leading zeros) + assert_eq!(eval(r#"Decode("ABCDe€😀", "xml")"#), "ABCDe€😀"); + assert_eq!(eval(r#"Decode("ABCé", "html")"#), "ABCé"); + + // Invalid / surrogate / out-of-range / zero / malformed numeric references stay literal without panicking + assert_eq!( + eval(r#"Decode("� � � � � &#; &#x; A G; &;", "xml")"#), + "� � � � � &#; &#x; A G; &;" + ); + + // Single-pass decoding: `&lt;` and `&lt;` must decode to `<`, never `<` + assert_eq!(eval(r#"Decode("&lt; &#65; &lt; &apos;", "xml")"#), "< A < '"); + assert_eq!(eval(r#"Decode("&nbsp; &eacute;", "html")"#), "  é"); + + // HTML named entities decode in "html" mode but remain literal in "xml" mode + assert_eq!(eval(r#"Decode("café ©€—αΩ♥", "html")"#), "café\u{00A0}©€—αΩ♥"); + assert_eq!(eval(r#"Decode("café ©€", "xml")"#), "café ©€"); + // Case sensitivity of HTML named entities (`É` vs `é`, `″` vs `′`, `‡` vs `†`) + assert_eq!(eval(r#"Decode("Éé″′‡†&bogus;", "html")"#), "Éé″′‡†&bogus;"); + // Non-ASCII characters after `&` or `&#` and long runs of `&` without `;` + assert_eq!(eval(r#"Decode("&é; &#é; &#xé; &é", "html")"#), "&é; &#é; &#xé; &é"); + let amps = "&".repeat(5_000); + assert_eq!(eval(&format!(r#"Decode("{amps}'", "xml")"#)), format!("{amps}'")); + } } diff --git a/crates/js/src/formcalc/builtins.rs b/crates/js/src/formcalc/builtins.rs index c3b7c7f1f2..e9e3bf90be 100644 --- a/crates/js/src/formcalc/builtins.rs +++ b/crates/js/src/formcalc/builtins.rs @@ -2,7 +2,7 @@ use super::ast::Expr; use super::dates::{MAX_DAYS, civil_from_days, days_from_civil, format_date, parse_date}; use super::parser::parse; use super::value::{Val, fixed, is_numeric_text, parse_num}; -use super::words::word_num; +use super::words::{decode_entities, word_num}; use super::{Interp, MAX_STRING, R, STRING_TOO_LONG, XfaKind, is_limit}; impl Interp<'_> { @@ -200,7 +200,8 @@ impl Interp<'_> { let kind = if args.len() > 1 { self.arg_str(args, 1)?.to_ascii_lowercase() } else { "url".into() }; Val::Str(match (lower.as_str(), kind.as_str()) { ("encode", "html" | "xml") => s.replace('&', "&").replace('<', "<").replace('>', ">").replace('"', """), - ("decode", "html" | "xml") => s.replace("<", "<").replace(">", ">").replace(""", "\"").replace("&", "&"), + ("decode", "xml") => decode_entities(&s, false), + ("decode", "html") => decode_entities(&s, true), ("encode", _) => s .bytes() .map(|b| if b.is_ascii_alphanumeric() || b"-_.~".contains(&b) { (b as char).to_string() } else { format!("%{b:02X}") }) diff --git a/crates/js/src/formcalc/words.rs b/crates/js/src/formcalc/words.rs index d0d119abb5..a6a5141920 100644 --- a/crates/js/src/formcalc/words.rs +++ b/crates/js/src/formcalc/words.rs @@ -73,3 +73,140 @@ pub(super) fn word_num(n: f64, cents: bool) -> String { } out } + +/// Decode XML (`html == false`) or HTML (`html == true`) character and entity references in a +/// single pass so `&lt;` becomes `<` rather than `<`, and invalid or unrecognized +/// references remain literal without panicking. +/// +/// When `html` is `false`, only the five predefined XML entities (`<`, `>`, `&`, +/// `"`, `'`) and numeric character references (`&#...;`, `&#x...;`, `&#X...;`) are +/// decoded. When `html` is `true`, W3C HTML 4 / XHTML 1.0 named entities are decoded as well. +pub(super) fn decode_entities(s: &str, html: bool) -> String { + let mut out = String::with_capacity(s.len()); + let mut rest = s; + while let Some(at) = rest.find('&') { + out.push_str(rest.get(..at).unwrap_or_default()); + rest = rest.get(at..).unwrap_or_default(); + match entity_at(rest, html) { + Some((c, len)) => { + out.push(c); + rest = rest.get(len..).unwrap_or_default(); + } + None => { + out.push('&'); + rest = rest.get(1..).unwrap_or_default(); + } + } + } + out.push_str(rest); + out +} + +/// Parse a character reference starting at the leading `&` of `s`, returning the decoded `char` +/// and the byte length of the reference (including `&` and `;`). +fn entity_at(s: &str, html: bool) -> Option<(char, usize)> { + if let Some(body) = s.strip_prefix("&#") { + let (radix, digits, prefix_len) = match body.strip_prefix(['x', 'X']) { + Some(hex) => (16u32, hex, 3usize), + None => (10u32, body, 2usize), + }; + let mut value = 0u32; + let mut any_digit = false; + for (i, c) in digits.char_indices() { + match c.to_digit(radix) { + Some(d) => { + any_digit = true; + value = value.checked_mul(radix)?.checked_add(d)?; + if value > 0x10_FFFF { + return None; + } + } + None if c == ';' && any_digit && value != 0 => { + return Some((char::from_u32(value)?, prefix_len + i + 1)); + } + None => return None, + } + } + return None; + } + let body = s.strip_prefix('&')?; + let semi = body.bytes().take(9).position(|b| b == b';')?; + let name = body.get(..semi).filter(|n| !n.is_empty() && n.bytes().all(|b| b.is_ascii_alphanumeric()))?; + let ch = match name { + "lt" => '<', + "gt" => '>', + "amp" => '&', + "quot" => '"', + "apos" => '\'', + _ if html => html_named_entity(name)?, + _ => return None, + }; + Some((ch, 1 + semi + 1)) +} + +/// Look up a W3C HTML 4 / XHTML 1.0 named entity (excluding the five XML entities). +fn html_named_entity(name: &str) -> Option { + if let Some(idx) = LATIN1_ENTITIES.iter().position(|&n| n == name) { + return char::from_u32(0xA0 + idx as u32); + } + HTML_ENTITIES.iter().find(|&&(n, _)| n == name).map(|&(_, c)| c) +} + +/// ISO-8859-1 named entities in code-point order `U+00A0..=U+00FF`. Entries must remain ordered +/// by successive code points starting at `0xA0` and must not be reordered alphabetically, since +/// [`html_named_entity`] maps each entry's index to `0xA0 + idx as u32`. +#[rustfmt::skip] +const LATIN1_ENTITIES: [&str; 96] = [ + "nbsp", "iexcl", "cent", "pound", "curren", "yen", "brvbar", "sect", "uml", "copy", "ordf", "laquo", "not", "shy", "reg", "macr", + "deg", "plusmn", "sup2", "sup3", "acute", "micro", "para", "middot", "cedil", "sup1", "ordm", "raquo", "frac14", "frac12", "frac34", "iquest", + "Agrave", "Aacute", "Acirc", "Atilde", "Auml", "Aring", "AElig", "Ccedil", "Egrave", "Eacute", "Ecirc", "Euml", "Igrave", "Iacute", "Icirc", "Iuml", + "ETH", "Ntilde", "Ograve", "Oacute", "Ocirc", "Otilde", "Ouml", "times", "Oslash", "Ugrave", "Uacute", "Ucirc", "Uuml", "Yacute", "THORN", "szlig", + "agrave", "aacute", "acirc", "atilde", "auml", "aring", "aelig", "ccedil", "egrave", "eacute", "ecirc", "euml", "igrave", "iacute", "icirc", "iuml", + "eth", "ntilde", "ograve", "oacute", "ocirc", "otilde", "ouml", "divide", "oslash", "ugrave", "uacute", "ucirc", "uuml", "yacute", "thorn", "yuml", +]; + +#[rustfmt::skip] +const HTML_ENTITIES: &[(&str, char)] = &[ + ("OElig", '\u{0152}'), ("oelig", '\u{0153}'), ("Scaron", '\u{0160}'), ("scaron", '\u{0161}'), ("Yuml", '\u{0178}'), ("fnof", '\u{0192}'), ("circ", '\u{02C6}'), ("tilde", '\u{02DC}'), + ("Alpha", '\u{0391}'), ("Beta", '\u{0392}'), ("Gamma", '\u{0393}'), ("Delta", '\u{0394}'), ("Epsilon", '\u{0395}'), ("Zeta", '\u{0396}'), ("Eta", '\u{0397}'), ("Theta", '\u{0398}'), + ("Iota", '\u{0399}'), ("Kappa", '\u{039A}'), ("Lambda", '\u{039B}'), ("Mu", '\u{039C}'), ("Nu", '\u{039D}'), ("Xi", '\u{039E}'), ("Omicron", '\u{039F}'), ("Pi", '\u{03A0}'), + ("Rho", '\u{03A1}'), ("Sigma", '\u{03A3}'), ("Tau", '\u{03A4}'), ("Upsilon", '\u{03A5}'), ("Phi", '\u{03A6}'), ("Chi", '\u{03A7}'), ("Psi", '\u{03A8}'), ("Omega", '\u{03A9}'), + ("alpha", '\u{03B1}'), ("beta", '\u{03B2}'), ("gamma", '\u{03B3}'), ("delta", '\u{03B4}'), ("epsilon", '\u{03B5}'), ("zeta", '\u{03B6}'), ("eta", '\u{03B7}'), ("theta", '\u{03B8}'), + ("iota", '\u{03B9}'), ("kappa", '\u{03BA}'), ("lambda", '\u{03BB}'), ("mu", '\u{03BC}'), ("nu", '\u{03BD}'), ("xi", '\u{03BE}'), ("omicron", '\u{03BF}'), ("pi", '\u{03C0}'), + ("rho", '\u{03C1}'), ("sigmaf", '\u{03C2}'), ("sigma", '\u{03C3}'), ("tau", '\u{03C4}'), ("upsilon", '\u{03C5}'), ("phi", '\u{03C6}'), ("chi", '\u{03C7}'), ("psi", '\u{03C8}'), + ("omega", '\u{03C9}'), ("thetasym", '\u{03D1}'), ("upsih", '\u{03D2}'), ("piv", '\u{03D6}'), ("ensp", '\u{2002}'), ("emsp", '\u{2003}'), ("thinsp", '\u{2009}'), ("zwnj", '\u{200C}'), + ("zwj", '\u{200D}'), ("lrm", '\u{200E}'), ("rlm", '\u{200F}'), ("ndash", '\u{2013}'), ("mdash", '\u{2014}'), ("lsquo", '\u{2018}'), ("rsquo", '\u{2019}'), ("sbquo", '\u{201A}'), + ("ldquo", '\u{201C}'), ("rdquo", '\u{201D}'), ("bdquo", '\u{201E}'), ("dagger", '\u{2020}'), ("Dagger", '\u{2021}'), ("bull", '\u{2022}'), ("hellip", '\u{2026}'), ("permil", '\u{2030}'), + ("prime", '\u{2032}'), ("Prime", '\u{2033}'), ("lsaquo", '\u{2039}'), ("rsaquo", '\u{203A}'), ("oline", '\u{203E}'), ("frasl", '\u{2044}'), ("euro", '\u{20AC}'), ("image", '\u{2111}'), + ("weierp", '\u{2118}'), ("real", '\u{211C}'), ("trade", '\u{2122}'), ("alefsym", '\u{2135}'), ("larr", '\u{2190}'), ("uarr", '\u{2191}'), ("rarr", '\u{2192}'), ("darr", '\u{2193}'), + ("harr", '\u{2194}'), ("crarr", '\u{21B5}'), ("lArr", '\u{21D0}'), ("uArr", '\u{21D1}'), ("rArr", '\u{21D2}'), ("dArr", '\u{21D3}'), ("hArr", '\u{21D4}'), ("forall", '\u{2200}'), + ("part", '\u{2202}'), ("exist", '\u{2203}'), ("empty", '\u{2205}'), ("nabla", '\u{2207}'), ("isin", '\u{2208}'), ("notin", '\u{2209}'), ("ni", '\u{220B}'), ("prod", '\u{220F}'), + ("sum", '\u{2211}'), ("minus", '\u{2212}'), ("lowast", '\u{2217}'), ("radic", '\u{221A}'), ("prop", '\u{221D}'), ("infin", '\u{221E}'), ("ang", '\u{2220}'), ("and", '\u{2227}'), + ("or", '\u{2228}'), ("cap", '\u{2229}'), ("cup", '\u{222A}'), ("int", '\u{222B}'), ("there4", '\u{2234}'), ("sim", '\u{223C}'), ("cong", '\u{2245}'), ("asymp", '\u{2248}'), + ("ne", '\u{2260}'), ("equiv", '\u{2261}'), ("le", '\u{2264}'), ("ge", '\u{2265}'), ("sub", '\u{2282}'), ("sup", '\u{2283}'), ("nsub", '\u{2284}'), ("sube", '\u{2286}'), + ("supe", '\u{2287}'), ("oplus", '\u{2295}'), ("otimes", '\u{2297}'), ("perp", '\u{22A5}'), ("sdot", '\u{22C5}'), ("lceil", '\u{2308}'), ("rceil", '\u{2309}'), ("lfloor", '\u{230A}'), + ("rfloor", '\u{230B}'), ("lang", '\u{2329}'), ("rang", '\u{232A}'), ("loz", '\u{25CA}'), ("spades", '\u{2660}'), ("clubs", '\u{2663}'), ("hearts", '\u{2665}'), ("diams", '\u{2666}'), +]; + +#[cfg(test)] +mod tests { + use super::{HTML_ENTITIES, LATIN1_ENTITIES, decode_entities}; + + #[test] + fn every_listed_html_entity_decodes_including_eight_byte_boundary() { + for (i, &name) in LATIN1_ENTITIES.iter().enumerate() { + let expected = char::from_u32(0xA0 + i as u32).unwrap().to_string(); + let input = format!("&{name};"); + assert_eq!(decode_entities(&input, true), expected, "entity {name}"); + assert_eq!(decode_entities(&input, false), input, "xml mode leaves {name} literal"); + } + for &(name, ch) in HTML_ENTITIES { + assert!(name.len() <= 8, "entity name {name} exceeds 8-byte parser limit"); + let input = format!("&{name};"); + assert_eq!(decode_entities(&input, true), ch.to_string(), "entity {name}"); + assert_eq!(decode_entities(&input, false), input, "xml mode leaves {name} literal"); + } + assert_eq!(decode_entities("ϑ", true), "\u{03D1}"); + assert_eq!(decode_entities("&thetasymx;", true), "&thetasymx;"); + } +} From 1f48670fcef2e5b9f5294650cb7dc51cdf5a3cff Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:32:08 +0800 Subject: [PATCH 11/75] refactor(cos): consolidate duplicate civil_from_days onto pdfcraft_cos::civil_from_days (#6540) Fixes #6466 --- apps/pdfcraft/src/logging.rs | 23 ++++-------------- crates/cos/src/lib.rs | 2 +- crates/cos/src/writer.rs | 20 ++++++++++------ crates/engine/src/lib.rs | 2 +- crates/sign/src/der.rs | 14 ++--------- .../ui-egui/src/workflows/combine_ui/parts.rs | 23 ++++-------------- xtask/src/demo_pdf/pdf.rs | 24 ++----------------- 7 files changed, 29 insertions(+), 79 deletions(-) diff --git a/apps/pdfcraft/src/logging.rs b/apps/pdfcraft/src/logging.rs index bef27de169..4c0f39e85c 100644 --- a/apps/pdfcraft/src/logging.rs +++ b/apps/pdfcraft/src/logging.rs @@ -102,25 +102,12 @@ pub fn timestamp(t: SystemTime) -> String { format!("{y:04}-{m:02}-{d:02}T{:02}:{:02}:{:02}.{:03}Z", rem / 3600, rem % 3600 / 60, rem % 60, since.subsec_millis()) } -/// Days since 1970-01-01 → proleptic Gregorian (year, month, day). -/// -/// `timestamp` clamps a pre-1970 clock to the epoch, so `days` is an unsigned count that can run -/// past `NaiveDate`'s ±262143-year range. The Gregorian calendar repeats every 400 years -/// (146_097 days), so chrono converts the day within one cycle and whole cycles go back into the -/// year (an `i128`, then saturated into `i64`), keeping chrono as the only calendar logic and -/// reproducing the old `civil_from_days` result exactly without hand-rolled day math. 1970-01-01 -/// is `num_days_from_ce` 719_163. +/// Days since 1970-01-01 → proleptic Gregorian (year, month, day): `pdfcraft_engine::civil_from_days`, +/// which keeps the date past chrono's range by whole 400-year cycles. `days` is a `u64` second +/// count over 86_400, at most about 2.1e14, so it always fits an `i64`. fn civil_from_days(days: u64) -> (i64, u32, u32) { - use chrono::Datelike; - let (cycles, rest) = (days / 146_097, days % 146_097); - // `rest + 719_163` is 719_163..865_259, always a valid CE day. - match i32::try_from(rest + 719_163).ok().and_then(chrono::NaiveDate::from_num_days_from_ce_opt) { - Some(d) => { - let y = i128::from(d.year()) + i128::from(cycles) * 400; - (i64::try_from(y).unwrap_or(i64::MAX), d.month(), d.day()) - } - None => (1970, 1, 1), // unreachable - } + // The fallback is unreachable: `days` is at most `u64::MAX / 86_400`, far below `i64::MAX`. + pdfcraft_engine::civil_from_days(i64::try_from(days).unwrap_or(i64::MAX)) } fn numbered(dir: &Path, n: usize) -> PathBuf { diff --git a/crates/cos/src/lib.rs b/crates/cos/src/lib.rs index 32a6f6d8b3..0d0fa33f61 100644 --- a/crates/cos/src/lib.rs +++ b/crates/cos/src/lib.rs @@ -30,7 +30,7 @@ pub use parser::{Lexer, is_whitespace, parse_indirect}; pub use pdfcraft_crypt::{ Algorithm, Auth, Method as CryptMethod, NewEncryption, Permissions, SecurityHandler, effective_password, pdfdoc_byte, pdfdoc_char, }; -pub use writer::{SaveOptions, pdf_date, reachable_objects, serialize, write_full, write_incremental}; +pub use writer::{SaveOptions, civil_from_days, pdf_date, reachable_objects, serialize, write_full, write_incremental}; #[derive(Debug, thiserror::Error, Clone, PartialEq)] pub enum CosError { diff --git a/crates/cos/src/writer.rs b/crates/cos/src/writer.rs index 75d49dbd76..3037669aea 100644 --- a/crates/cos/src/writer.rs +++ b/crates/cos/src/writer.rs @@ -919,14 +919,14 @@ pub fn pdf_date(unix_secs: i64) -> String { format!("D:{y:04}{m:02}{d:02}{:02}{:02}{:02}Z", secs / 3600, secs / 60 % 60, secs % 60) } -/// Days since 1970-01-01 → proleptic Gregorian (year, month, day). +/// Days since 1970-01-01 → proleptic Gregorian (year, month, day). Total for any `i64` count: +/// dates past `NaiveDate`'s ±262143-year range keep their proleptic Gregorian year instead of +/// falling back to the epoch. /// -/// A `from_timestamp` epoch fallback would change dates past `NaiveDate`'s ±262143-year range, -/// which the old hand-rolled converter formatted directly. The Gregorian calendar repeats every -/// 400 years (146_097 days), so chrono converts the day within one cycle and whole cycles go back -/// into the year (an `i128` so the extremes can't overflow), keeping chrono as the only calendar -/// logic and reproducing the old result exactly. 1970-01-01 is `num_days_from_ce` 719_163. -fn civil_from_days(days: i64) -> (i64, u32, u32) { +/// The Gregorian calendar repeats every 400 years (146_097 days), so chrono converts the day +/// within one cycle and whole cycles go back into the year (an `i128` so the extremes can't +/// overflow), keeping chrono as the only calendar logic. 1970-01-01 is `num_days_from_ce` 719_163. +pub fn civil_from_days(days: i64) -> (i64, u32, u32) { use chrono::Datelike; let (cycles, rest) = (days.div_euclid(146_097), days.rem_euclid(146_097)); // `rest + 719_163` is 719_163..865_259, always a valid CE day. @@ -1110,6 +1110,12 @@ mod tests { // its sign, exactly as before). assert_eq!(pdf_date(10_000_000_000_000), "D:3188570520174640Z"); assert_eq!(pdf_date(-10_000_000_000_000), "D:-3149180813061320Z"); + assert_eq!(crate::civil_from_days(0), (1970, 1, 1)); + assert_eq!(crate::civil_from_days(-1), (1969, 12, 31)); + assert_eq!(crate::civil_from_days(11_016), (2000, 2, 29)); + assert_eq!(crate::civil_from_days(19_782), (2024, 2, 29)); + assert_eq!(crate::civil_from_days(i64::MAX), (25_252_734_927_768_524, 7, 27)); + assert_eq!(crate::civil_from_days(i64::MIN), (-25_252_734_927_764_585, 6, 7)); } #[test] diff --git a/crates/engine/src/lib.rs b/crates/engine/src/lib.rs index baa06b02fd..e6e21cb1eb 100644 --- a/crates/engine/src/lib.rs +++ b/crates/engine/src/lib.rs @@ -86,7 +86,7 @@ pub use pdfcraft_organize::Standard; pub use pdfcraft_organize::check_custom_key; pub use pdfcraft_organize::view::{InitialView, Layout as InitialLayout, Magnification, Navigation}; -pub use pdfcraft_cos::Algorithm; +pub use pdfcraft_cos::{Algorithm, civil_from_days}; pub use pdfcraft_create::{CONVERTIBLE, CreateError, ImageResolution, SourceKind, decode_text, source_kind}; pub use pdfcraft_edit::added::{ CHAR_SPACING_LIMIT as ADDED_CHAR_SPACING_LIMIT, paste_place as added_paste_place, paste_rect as added_paste_rect, text_rect as added_text_rect, diff --git a/crates/sign/src/der.rs b/crates/sign/src/der.rs index 0cac252c33..f9f2f3e6d9 100644 --- a/crates/sign/src/der.rs +++ b/crates/sign/src/der.rs @@ -349,19 +349,9 @@ impl Time { /// year is cast exactly as the old code did (`as u32`, wrapping) — no well-formed signing time /// is anywhere near that range, and this preserves byte-for-byte the old output everywhere. pub fn from_unix(t: i64) -> Time { - use chrono::Datelike; let (days, secs) = (t.div_euclid(86_400), t.rem_euclid(86_400)); - let (cycles, rest) = (days.div_euclid(146_097), days.rem_euclid(146_097)); - // `rest + 719_163` is 719_163..865_259, always a valid CE day. - let (year, month, day) = match i32::try_from(rest + 719_163).ok().and_then(chrono::NaiveDate::from_num_days_from_ce_opt) { - Some(d) => { - let y = i128::from(d.year()) + i128::from(cycles) * 400; - // Match the old `as u32` cast: saturate i128→i64 first (the old year was i64), then wrap into u32. - (i64::try_from(y).unwrap_or(if y < 0 { i64::MIN } else { i64::MAX }) as u32, d.month(), d.day()) - } - None => (1970, 1, 1), // unreachable - }; - Time { year, month, day, hour: (secs / 3600) as u32, minute: (secs % 3600 / 60) as u32, second: (secs % 60) as u32 } + let (year, month, day) = pdfcraft_cos::civil_from_days(days); + Time { year: year as u32, month, day, hour: (secs / 3600) as u32, minute: (secs % 3600 / 60) as u32, second: (secs % 60) as u32 } } pub fn add_seconds(&self, s: i64) -> Time { diff --git a/crates/ui-egui/src/workflows/combine_ui/parts.rs b/crates/ui-egui/src/workflows/combine_ui/parts.rs index a64d0eeb34..4527ec061d 100644 --- a/crates/ui-egui/src/workflows/combine_ui/parts.rs +++ b/crates/ui-egui/src/workflows/combine_ui/parts.rs @@ -274,23 +274,10 @@ pub(super) fn utc_stamp(time: SystemTime) -> String { format!("{y:04}-{m:02}-{d:02} {:02}:{:02} UTC", rem / 3600, rem % 3600 / 60) } -/// Days since 1970-01-01 → proleptic Gregorian (year, month, day). -/// -/// A pre-1970 time already clamps to the epoch (`unwrap_or(0)`), so `days` is an unsigned count -/// that can run past `NaiveDate`'s ±262143-year range, where the old hand-rolled converter still -/// produced a date. The Gregorian calendar repeats every 400 years (146_097 days), so chrono -/// converts the day within one cycle and whole cycles go back into the year (an `i128`, then -/// saturated into `i64`), keeping chrono as the only calendar logic and reproducing the old -/// result exactly. 1970-01-01 is `num_days_from_ce` 719_163. +/// Days since 1970-01-01 → proleptic Gregorian (year, month, day): `pdfcraft_cos::civil_from_days`, +/// which keeps the date past chrono's range by whole 400-year cycles. `days` is a `u64` second +/// count over 86_400, at most about 2.1e14, so it always fits an `i64`. fn civil_from_days(days: u64) -> (i64, u32, u32) { - use chrono::Datelike; - let (cycles, rest) = (days / 146_097, days % 146_097); - // `rest + 719_163` is 719_163..865_259, always a valid CE day. - match i32::try_from(rest + 719_163).ok().and_then(chrono::NaiveDate::from_num_days_from_ce_opt) { - Some(d) => { - let y = i128::from(d.year()) + i128::from(cycles) * 400; - (i64::try_from(y).unwrap_or(i64::MAX), d.month(), d.day()) - } - None => (1970, 1, 1), // unreachable - } + // The fallback is unreachable: `days` is at most `u64::MAX / 86_400`, far below `i64::MAX`. + pdfcraft_cos::civil_from_days(i64::try_from(days).unwrap_or(i64::MAX)) } diff --git a/xtask/src/demo_pdf/pdf.rs b/xtask/src/demo_pdf/pdf.rs index b51ea1a643..dcdc156683 100644 --- a/xtask/src/demo_pdf/pdf.rs +++ b/xtask/src/demo_pdf/pdf.rs @@ -77,29 +77,9 @@ impl Stamp { pub fn from_unix(secs: i64) -> Self { let (days, rem) = (secs.div_euclid(86_400), secs.rem_euclid(86_400)); let (h, mi, s) = (rem / 3600, rem % 3600 / 60, rem % 60); - let (y, m, d) = Self::civil_from_days(days); + let (y, m, d) = pdfcraft_cos::civil_from_days(days); // `{y:04}` rather than `%Y`, which writes a `+` before years past 9999. - Stamp { unix: secs, pdf: format!("D:{y:04}{m:02}{d:02}{h:02}{mi:02}{s:02}Z"), iso: format!("{y:04}-{m:02}-{d:02}T{h:02}:{mi:02}:{s:02}Z") } - } - - /// Days since 1970-01-01 → proleptic Gregorian (year, month, day). - /// - /// Past `NaiveDate`'s ±262143-year range the old hand-rolled converter still produced a date, - /// which both stamps formatted directly; a `from_timestamp` epoch fallback would change it. - /// The Gregorian calendar repeats every 400 years (146_097 days), so chrono converts the day - /// within one cycle and whole cycles go back into the year (an `i128` so the extremes can't - /// overflow), keeping chrono as the only calendar logic. 1970-01-01 is `num_days_from_ce` 719_163. - fn civil_from_days(days: i64) -> (i64, u32, u32) { - use chrono::Datelike; - let (cycles, rest) = (days.div_euclid(146_097), days.rem_euclid(146_097)); - // `rest + 719_163` is 719_163..865_259, always a valid CE day. - match i32::try_from(rest + 719_163).ok().and_then(chrono::NaiveDate::from_num_days_from_ce_opt) { - Some(d) => { - let y = i128::from(d.year()) + i128::from(cycles) * 400; - (i64::try_from(y).unwrap_or(if y < 0 { i64::MIN } else { i64::MAX }), d.month(), d.day()) - } - None => (1970, 1, 1), // unreachable - } + Stamp { unix: secs, pdf: pdfcraft_cos::pdf_date(secs), iso: format!("{y:04}-{m:02}-{d:02}T{h:02}:{mi:02}:{s:02}Z") } } /// The same moment shifted by `minutes` (used to order threaded replies). From ee2398f7107cbdf06a7ba5425efd69d0fc00e7a1 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:37:56 +0800 Subject: [PATCH 12/75] M11: PDF/A verify checks that predefined XMP properties have their value types (optimize.pdfa-validate) (#6220) --- crates/preflight/src/lib.rs | 102 +++- crates/preflight/src/tests.rs | 353 +++++++++++++- crates/preflight/src/xmp.rs | 13 +- crates/preflight/src/xmp_types.rs | 780 ++++++++++++++++++++++++++++++ parity/acrobat-features.toml | 4 +- 5 files changed, 1228 insertions(+), 24 deletions(-) create mode 100644 crates/preflight/src/xmp_types.rs diff --git a/crates/preflight/src/lib.rs b/crates/preflight/src/lib.rs index 5fdec67fac..1cbe9a6ed3 100644 --- a/crates/preflight/src/lib.rs +++ b/crates/preflight/src/lib.rs @@ -23,6 +23,7 @@ pub mod icc; mod xmp; +mod xmp_types; use pdfcraft_cos::{Dict, Document, ObjRef, Object, PdfString, Stream}; @@ -63,18 +64,25 @@ impl Level { /// another language says it from [`pattern`](Issue::pattern) and [`value`](Issue::value) instead: /// the pattern is wording of ours, listed in [`WORDINGS`], and the value — when there is one — is /// the document's own name for something, a PDF keyword that reads the same in every language. +/// A wording that names more than one thing has a named hole, such as `{object}`, for each, so a +/// language can put them in its own order; [`values`](Issue::values) fills those. /// [`Issue::said`] does that. #[derive(Clone, Debug, PartialEq)] pub struct Issue { /// The ISO 19005-2 clause. pub clause: &'static str, pub message: String, - /// The problem as it is worded here, with `{}` where the document's own name goes. + /// The problem as it is worded here, with `{}` or a named hole where the document's own name goes. pub pattern: &'static str, /// What goes in the pattern's `{}`: an annotation subtype, an action, a font, a page number. pub value: Option, + /// What goes in each of the pattern's named holes, by the hole's name (`"{object}"`, braces + /// included). + pub values: Vec<(&'static str, String)>, /// 0-based page, when the problem is on a page. pub page: Option, + /// The indirect object the problem is in, when that is what locates it: a metadata stream. + pub object: Option, /// Whether [`convert`] can fix it. pub fixable: bool, } @@ -86,14 +94,43 @@ impl Issue { /// font called `{}` cannot turn into anything else. With [`in_english`] this is /// [`message`](Issue::message). pub fn said(&self, words: Words) -> String { - let said = words(self.pattern); - match &self.value { - Some(v) => said.replacen("{}", v, 1), - None => said, - } + fill(&words(self.pattern), self.value.as_deref(), &self.values) } } +/// `wording` with its holes filled: the first `{}` with `value`, and each named hole, such as +/// `{object}`, with its entry in `values`. A hole with nothing for it stays as it is written. +/// +/// The values go in during one pass over the wording, so a value with braces in it is never read +/// as a hole. +fn fill(wording: &str, value: Option<&str>, values: &[(&str, String)]) -> String { + let mut out = String::with_capacity(wording.len().saturating_add(16)); + let mut value = value; + let mut rest = wording; + while let Some(open) = rest.find('{') { + let (before, from_brace) = rest.split_at(open); + out.push_str(before); + // A brace with nothing closing it is just text. `}` is one byte, so the end of the hole is a + // char boundary; `get` rather than slicing all the same. + let Some(hole) = from_brace.find('}').and_then(|close| from_brace.get(..=close)) else { + out.push_str(from_brace); + return out; + }; + if hole == "{}" + && let Some(v) = value.take() + { + out.push_str(v); + } else if let Some((_, v)) = values.iter().find(|(name, _)| *name == hole) { + out.push_str(v); + } else { + out.push_str(hole); + } + rest = from_brace.get(hole.len()..).unwrap_or_default(); + } + out.push_str(rest); + out +} + /// How an interface says one of our wordings in its own language: given the English in [`WORDINGS`], /// the same wording in that language. [`in_english`] keeps the English. pub type Words<'a> = &'a dyn Fn(&str) -> String; @@ -103,7 +140,8 @@ pub fn in_english(wording: &str) -> String { wording.to_owned() } -/// Every problem [`verify`] can report, in English, with `{}` where the document's own name goes. +/// Every problem [`verify`] can report, in English, with `{}` or a named hole, such as `{object}`, +/// where the document's own name goes. /// /// An interface that shows them in another language translates these, which is the whole reason they /// are gathered here; `wordings_are_complete` fails if a rule says something that is not on the @@ -113,6 +151,8 @@ pub const WORDINGS: &[&str] = &[ "The document has no XMP metadata", "The metadata doesn't identify the file as {}", "The title in the metadata and the document information differ", + "The XMP property {property} in metadata stream {object} should have the value type {type}", + "The XMP property {property} in metadata stream {object}, written {written}, should have the value type {type}", // The document as a whole. "The document is encrypted", "Device colour is used, but there is no PDF/A output intent", @@ -427,15 +467,24 @@ thread_local! { /// Check `doc` against `level`. pub fn verify(doc: &Document, level: Level) -> Vec { - let mut out = Vec::new(); + // Shared by `issue` and `located_issue`, each of which takes the list out, adds one issue and + // puts it back. + let out = std::cell::Cell::new(Vec::new()); + let push = |i: Issue| { + let mut list = out.take(); + list.push(i); + out.set(list); + }; // Each problem is worded here and named by the document there: the wording goes on the list in // `WORDINGS` for an interface to translate, and the name goes in its `{}` as it stands. let mut issue = |clause: &'static str, pattern: &'static str, value: Option, page: Option, fixable: bool| { - let message = match &value { - Some(v) => pattern.replacen("{}", v, 1), - None => pattern.to_owned(), - }; - out.push(Issue { clause, message, pattern, value, page, fixable }); + let message = fill(pattern, value.as_deref(), &[]); + push(Issue { clause, message, pattern, value, values: Vec::new(), page, object: None, fixable }); + }; + // A problem in one indirect object, worded with named holes. + let located_issue = |clause: &'static str, pattern: &'static str, values: Vec<(&'static str, String)>, object: ObjRef, fixable: bool| { + let message = fill(pattern, None, &values); + push(Issue { clause, message, pattern, value: None, values, page: None, object: Some(object), fixable }); }; let cat = catalog(doc); // 6.1.3 / 6.6: identification and metadata. @@ -460,6 +509,32 @@ pub fn verify(doc: &Document, level: Level) -> Vec { } // 6.2.2: output intent when device colour is used. let objs = objects(doc); + // 6.6.2.3.1: the value types of predefined XMP properties, in every metadata stream. + let mut mistyped = Vec::new(); + xmp_types::for_each_packet(&objs, |object, packet| { + mistyped.extend(xmp_types::mistyped(packet).into_iter().map(|((name, written), ty)| (object, name, written, ty))); + }); + for (object, name, written, ty) in mistyped { + let mut values = vec![("{property}", name.clone()), ("{object}", object.to_string()), ("{type}", ty)]; + if written == name { + located_issue( + "6.6.2.3.1", + "The XMP property {property} in metadata stream {object} should have the value type {type}", + values, + object, + false, + ); + } else { + values.push(("{written}", written)); + located_issue( + "6.6.2.3.1", + "The XMP property {property} in metadata stream {object}, written {written}, should have the value type {type}", + values, + object, + false, + ); + } + } let used = uses_device_colour(doc, &objs); match pdfa_intent_profile(doc) { None if used.any() => { @@ -516,6 +591,7 @@ pub fn verify(doc: &Document, level: Level) -> Vec { } check_objects(doc, level, &objs, &page_of, &mut issue); // One message per kind of problem is enough for repeated ones (as preflight summaries do). + let mut out = out.into_inner(); let mut seen = std::collections::HashSet::new(); out.retain(|i| seen.insert((i.clause, i.message.clone(), i.page))); out diff --git a/crates/preflight/src/tests.rs b/crates/preflight/src/tests.rs index 52e6903e60..e691adae84 100644 --- a/crates/preflight/src/tests.rs +++ b/crates/preflight/src/tests.rs @@ -78,12 +78,14 @@ fn wordings_are_complete() { let source = include_str!("lib.rs"); let mut said = Vec::new(); let mut rest = source; - while let Some(at) = rest.find("issue(\"") { - rest = &rest[at + 7..]; + // `issue(` and `located_issue(`, written on one line or, when rustfmt breaks it, on several. + while let Some(at) = rest.find("issue(") { + rest = &rest[at + 6..]; + let Some(clause) = rest.trim_start().strip_prefix('"') else { continue }; // Step over the clause, then take the wording that follows it. - let Some(end) = rest.find('"') else { break }; - rest = &rest[end + 1..]; - let Some(text) = rest.strip_prefix(", \"") else { continue }; + let Some(end) = clause.find('"') else { break }; + rest = &clause[end + 1..]; + let Some(text) = rest.strip_prefix(',').map(str::trim_start).and_then(|r| r.strip_prefix('"')) else { continue }; let Some(end) = text.find('"') else { break }; said.push(&text[..end]); rest = text; @@ -488,3 +490,344 @@ fn an_output_intent_profile_header_is_read_within_a_size_cap() { assert_eq!(pdfa_intent_profile(&doc), Some(n), "a {size}-byte profile"); } } + +const XMP_NS: &str = concat!( + r#"xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:dc="http://purl.org/dc/elements/1.1/" "#, + r#"xmlns:xmp="http://ns.adobe.com/xap/1.0/" xmlns:xmpMM="http://ns.adobe.com/xap/1.0/mm/" "#, + r#"xmlns:stRef="http://ns.adobe.com/xap/1.0/sType/ResourceRef#" xmlns:xmpDM="http://ns.adobe.com/xmp/1.0/DynamicMedia/" "#, + r#"xmlns:crs="http://ns.adobe.com/camera-raw-settings/1.0/" xmlns:xmpTPg="http://ns.adobe.com/xap/1.0/t/pg/" "#, + r#"xmlns:stDim="http://ns.adobe.com/xap/1.0/sType/Dimensions#" xmlns:pdf="http://ns.adobe.com/pdf/1.3/""# +); + +/// The mistyped predefined properties of a packet whose one description holds `body` (its +/// attributes, then `>` and its elements). +fn mistyped(body: &str) -> Vec { + let xmp = format!( + r#""# + ); + let parsed = xmp::parse(&xmp).expect("parses"); + xmp_types::mistyped(&parsed).into_keys().map(|(name, _)| name).collect() +} + +/// [`mistyped`] for the property `p` written as an element holding `v`. +fn element(p: &str, v: &str) -> Vec { + mistyped(&format!("><{p}>{v}")) +} + +/// PDF/A-2 6.6.2.3.1, simple values: their lexical forms (ISO 16684-1 8.2.1), as elements and attributes. +#[test] +fn xmp_simple_values_have_their_lexical_forms() { + // Simple types, as elements and as attributes. + assert!(element("crs:Contrast", "-25").is_empty()); + assert_eq!(element("crs:Contrast", "25.0"), ["crs:Contrast"]); + assert_eq!(element("crs:Contrast", "4/1"), ["crs:Contrast"]); + assert!(element("crs:Exposure", "+.5").is_empty() && element("crs:Exposure", "5").is_empty()); + assert_eq!(element("crs:Exposure", "5."), ["crs:Exposure"]); + assert_eq!(element("crs:Exposure", "1e3"), ["crs:Exposure"]); + assert!(element("pdf:Trapped", "True").is_empty()); + assert_eq!(element("pdf:Trapped", "true"), ["pdf:Trapped"]); + assert_eq!(mistyped(r#"crs:AutoContrast="Yes">"#), ["crs:AutoContrast"]); + assert!(mistyped(r#"crs:AutoContrast="False">"#).is_empty()); + for good in + ["2016", "2016-02", "2016-02-01", "2016-02-01T13:19Z", "2016-02-01T13:19:21+01:00", "2016-02-01T13:19:21.5-08:00", "2016-02-01T13:19:21"] + { + assert!(element("xmp:CreateDate", good).is_empty(), "{good}"); + } + for bad in [ + "16-02-01", + "2016-13-01", + "2016-02-32", + "2016-02T13:19", + "2016-02-01T25:00", + "2016-02-01T13:19:21+1:00", + "Feb 1 2016", + "2016-02-01T13:19Z01:00", + ] { + assert_eq!(element("xmp:CreateDate", bad), ["xmp:CreateDate"], "{bad}"); + } + assert!(element("xmpDM:videoPixelAspectRatio", "1001/30000").is_empty()); + assert_eq!(element("xmpDM:videoPixelAspectRatio", "1.5"), ["xmpDM:videoPixelAspectRatio"]); + // The whole text of a value counts, however a comment splits it. + assert_eq!(element("crs:Contrast", "5.0"), ["crs:Contrast"]); + assert!(element("pdf:Trapped", "True").is_empty()); + // rdf:resource makes a URI, which only a text type takes. + assert_eq!(mistyped(r#">"#), ["crs:Contrast"]); + // A GUID is a simple value but not a URI (ISO 16684-1 8.2.2.3). + assert!(element("xmpMM:DocumentID", "uuid:1234").is_empty()); + assert_eq!(mistyped(r#">"#), ["xmpMM:DocumentID"]); + assert_eq!( + mistyped(r#">"#), + ["xmpMM:DerivedFrom"] + ); +} + +/// PDF/A-2 6.6.2.3.1, arrays: the kind, every item, and the languages of a language alternative. +#[test] +fn xmp_arrays_have_their_kind_and_item_types() { + // Text can't be an array or a structure; a simple value can be given as rdf:resource. + assert_eq!(element("crs:RawFileName", "a"), ["crs:RawFileName"]); + assert_eq!(mistyped(r#">1"#), ["crs:RawFileName"]); + assert!(mistyped(r#">"#).is_empty()); + // Arrays: the kind and every item. + assert!(element("dc:creator", "AB").is_empty()); + assert_eq!(element("dc:creator", "A"), ["dc:creator"]); + assert_eq!(element("dc:creator", "A"), ["dc:creator"]); + assert_eq!(element("dc:date", "2016-02-01yesterday"), ["dc:date"]); + assert_eq!(element("dc:date", "yesterday"), ["dc:date"]); + // Only white space around an array's items. + assert_eq!(element("dc:creator", "strayA"), ["dc:creator"]); + assert_eq!(element("dc:creator", "strayA"), ["dc:creator"]); + // rdf:parseType or rdf:resource on the property makes its value something other than the array. + assert_eq!(mistyped(r#">A"#), ["dc:creator"]); + assert_eq!(mistyped(r#">A"#), ["dc:creator"]); + assert!(element("dc:creator", "\n \n A\n \n").is_empty()); + // An rdf:resource element is empty. + assert_eq!(mistyped(r#">x"#), ["dc:source"]); + // Language alternatives: an rdf:Alt whose items all carry xml:lang. + assert!(element("dc:title", r#"T"#).is_empty()); + assert_eq!(element("dc:title", "T"), ["dc:title"]); + assert_eq!(element("dc:title", "T"), ["dc:title"]); +} + +/// PDF/A-2 6.6.2.3.1, structures in their three RDF/XML forms, nested, and qualified values. +#[test] +fn xmp_structures_have_their_field_types() { + // Structures in their three forms, with each known field checked. + assert!(mistyped(r#">d2016"#).is_empty()); + assert!(mistyped(r#">"#).is_empty()); + assert!(mistyped(r#">"#).is_empty()); + assert_eq!(element("xmpMM:DerivedFrom", "uuid:1234"), ["xmpMM:DerivedFrom"]); + assert_eq!(mistyped(r#">letter"#), ["xmpMM:DerivedFrom"]); + assert_eq!( + mistyped(r#">letter"#), + ["xmpMM:DerivedFrom"] + ); + assert_eq!(mistyped(r#">letter"#), ["xmpMM:DerivedFrom"]); + // Predefined properties inside a Pantry item keep their types. + let pantry = |v: &str| { + mistyped(&format!( + r#">{v}"# + )) + }; + assert!(pantry("5").is_empty()); + assert_eq!(pantry("high"), ["xmpMM:Pantry"]); + assert_eq!( + mistyped(r#">soon"#), + ["xmpMM:DerivedFrom"] + ); + assert_eq!(mistyped(r#">"#), ["xmpMM:DerivedFrom"]); + assert_eq!(mistyped(r#">8.5 in"#), ["xmpTPg:MaxPageSize"]); + // Nested: an array of structures holding a structure. + let marker = |start: &str| format!(r#">{start}"#); + assert!( + mistyped(&marker( + r#"501/25"# + )) + .is_empty() + ); + assert!(mistyped(&marker("50f")).is_empty()); + assert_eq!(mistyped(&marker(r#"5.5"#)), ["xmpDM:markers"]); + // A qualified value is checked by its rdf:value. + assert!(mistyped(r#">5q"#).is_empty()); + assert_eq!(mistyped(r#">five"#), ["crs:Contrast"]); + // rdf:value as an attribute of the description. + assert!(mistyped(r#">"#).is_empty()); + assert_eq!(mistyped(r#">"#), ["crs:Contrast"]); + // rdf:value as an attribute of an empty property element. + assert!(mistyped(r#">"#).is_empty()); + assert_eq!(mistyped(r#">"#), ["crs:Contrast"]); + // Qualifiers that are predefined properties keep their types. + assert_eq!( + mistyped(r#">5high"#), + ["crs:Contrast"] + ); + assert_eq!(mistyped(r#">"#), ["crs:Contrast"]); + assert!(mistyped(r#">"#).is_empty()); + // Nesting past any type's depth doesn't have the type. + let nested = |levels: usize| { + let open = r#""#.repeat(levels); + let close = "".repeat(levels); + mistyped(&format!(r#">{open}5{close}"#)) + }; + assert!(nested(3).is_empty()); + assert_eq!(nested(40), ["crs:Contrast"]); +} + +/// PDF/A-2 6.6.2.3.1 through verify: one issue per mistyped property, not fixable. +#[test] +fn xmp_value_types_are_reported_by_verify() { + let metadata = |body: &str| { + let xmp = format!(r#""#); + format!("<< /Type /Metadata /Subtype /XML /Length {} >>\nstream\n{xmp}\nendstream", xmp.len()) + }; + let issues = |doc: &Document| -> Vec { verify(doc, Level::A2b).into_iter().filter(|i| i.clause == "6.6.2.3.1").collect() }; + // Every mistyped property, once per metadata stream, by stream and then by name; not fixable. + let catalog = metadata(r#"crs:Contrast="high" crs:Exposure="bright""#); + let page = metadata(r#"crs:Contrast="higher""#); + let doc = build( + &[ + "<< /Type /Catalog /Pages 2 0 R /Metadata 3 0 R >>", + "<< /Type /Pages /Kids [4 0 R] /Count 1 /MediaBox [0 0 200 200] >>", + &catalog, + "<< /Type /Page /Parent 2 0 R /Metadata 5 0 R >>", + &page, + ], + "", + ); + let found = issues(&doc); + let messages: Vec<&str> = found.iter().map(|i| i.message.as_str()).collect(); + assert_eq!( + messages, + [ + "The XMP property crs:Contrast in metadata stream 3 0 R should have the value type Integer", + "The XMP property crs:Exposure in metadata stream 3 0 R should have the value type Real", + "The XMP property crs:Contrast in metadata stream 5 0 R should have the value type Integer" + ] + ); + let objects: Vec> = found.iter().map(|i| i.object).collect(); + assert_eq!(objects, [Some(ObjRef::new(3, 0)), Some(ObjRef::new(3, 0)), Some(ObjRef::new(5, 0))]); + assert!(found.iter().all(|i| !i.fixable && i.page.is_none())); + // Said in another language, the property and the stream go where that language puts them, and + // what the document wrote goes in as it stands. + let reordered = |w: &str| w.replace("The XMP property {property} in metadata stream {object}", "In {object}, {property}"); + assert_eq!(found[0].said(&in_english), found[0].message); + assert_eq!(found[0].said(&reordered), "In 3 0 R, crs:Contrast should have the value type Integer"); + // A UTF-16 packet, either byte order, is read. + for big in [true, false] { + let xmp = format!( + r#""# + ); + let mut data = if big { vec![0xFE, 0xFF] } else { vec![0xFF, 0xFE] }; + data.extend(xmp.encode_utf16().flat_map(|u| if big { u.to_be_bytes() } else { u.to_le_bytes() })); + let hex: String = data.iter().map(|b| format!("{b:02X}")).collect(); + let stream = format!("<< /Type /Metadata /Subtype /XML /Filter /ASCIIHexDecode /Length {} >>\nstream\n{hex}>\nendstream", hex.len() + 1); + let doc = build(&["<< /Type /Catalog /Pages 2 0 R /Metadata 3 0 R >>", "<< /Type /Pages /Kids [] /Count 0 >>", &stream], ""); + assert_eq!(issues(&doc).len(), 1, "big-endian: {big}"); + } + // A metadata stream of a page counts as well as the catalog's. + let good = metadata(r#"crs:Contrast="5""#); + let doc = build( + &[ + "<< /Type /Catalog /Pages 2 0 R /Metadata 3 0 R >>", + "<< /Type /Pages /Kids [4 0 R] /Count 1 /MediaBox [0 0 200 200] >>", + &good, + "<< /Type /Page /Parent 2 0 R /Metadata 5 0 R >>", + &page, + ], + "", + ); + let found = issues(&doc); + assert_eq!(found.iter().map(|i| i.object).collect::>(), [Some(ObjRef::new(5, 0))]); +} + +/// Holes are filled in one pass: the first `{}` from `value`, named holes from `values`, and what +/// has nothing for it, a value's own braces and a brace left open stay as they are. +#[test] +fn wordings_fill_their_holes_once() { + let values = [("{a}", "{b}".to_owned()), ("{b}", "B".to_owned())]; + assert_eq!(fill("x {a} {b} {c} {} {}", Some("{a}"), &values), "x {b} B {c} {a} {}"); + assert_eq!(fill("{}", None, &[]), "{}"); + assert_eq!(fill("open { and é {a", Some("v"), &values), "open { and é {a"); + assert_eq!(fill("é{}ü", Some("v"), &[]), "évü"); +} + +/// PDF/A-2 6.6.2.3.1: a mistyped property is reported with the type the XMP Specification gives it, +/// as the specification writes it. +#[test] +fn xmp_mistyped_properties_name_their_type() { + let types = |body: &str| -> Vec<(String, String)> { + let xmp = format!( + r#""# + ); + xmp_types::mistyped(&xmp::parse(&xmp).expect("parses")).into_iter().map(|((name, _), ty)| (name, ty)).collect() + }; + let t = |name: &str, ty: &str| (name.to_owned(), ty.to_owned()); + assert_eq!( + types(r#"crs:Contrast="high" crs:Exposure="bright" xmp:CreateDate="never">"#), + [t("crs:Contrast", "Integer"), t("crs:Exposure", "Real"), t("xmp:CreateDate", "Date")] + ); + assert_eq!( + types(">CyanT"), + [t("dc:title", "Lang Alt"), t("xmpTPg:PlateNames", "seq Text")] + ); + assert_eq!( + types(">A4h"), + [t("xmpMM:History", "seq ResourceEvent"), t("xmpTPg:MaxPageSize", "Dimensions")] + ); + assert_eq!(types(">p"), [t("xmpMM:Pantry", "bag struct")]); +} + +/// PDF/A-2 6.6.2.3.1: a property is known by its namespace URI, whatever prefix the packet binds +/// to it, and is named by the namespace's preferred prefix and as the packet writes it. A property +/// a predefined namespace doesn't define, and a namespace that isn't predefined, are left to the +/// clauses about them. +#[test] +fn xmp_properties_are_known_by_namespace_uri_and_named_as_written() { + let crs = "http://ns.adobe.com/camera-raw-settings/1.0/"; + let found = |body: &str| -> Vec<(String, String)> { + let xmp = format!( + r#""# + ); + xmp_types::mistyped(&xmp::parse(&xmp).expect("parses")).into_keys().collect() + }; + let n = |name: &str, written: &str| (name.to_owned(), written.to_owned()); + // Another prefix for the namespace, as an attribute and as an element, and no prefix at all. + assert_eq!(found(&format!(r#"xmlns:c="{crs}" c:Contrast="high">"#)), [n("crs:Contrast", "c:Contrast")]); + assert_eq!(found(&format!(r#">bright"#)), [n("crs:Exposure", "c:Exposure")]); + assert_eq!(found(&format!(r#">bright"#)), [n("crs:Exposure", "Exposure")]); + assert!(found(&format!(r#"xmlns:c="{crs}" c:Contrast="5">"#)).is_empty()); + // The preferred prefix is written as it is. + assert_eq!(found(r#"crs:Contrast="high">"#), [n("crs:Contrast", "crs:Contrast")]); + // A long prefix is repeated only so far. + let long = "p".repeat(100); + let said = found(&format!(r#"xmlns:{long}="{crs}" {long}:Contrast="high">"#)); + assert_eq!(said.len(), 1); + assert_eq!(said[0].1, format!("{}…:Contrast", "p".repeat(64))); + // The preferred prefix bound to another namespace is not that namespace. + assert!(found(r#">high"#).is_empty()); + assert!(found(r#"xmlns:dc="http://example.com/not-dc/" dc:title="T">"#).is_empty()); + // A property a predefined namespace doesn't define, with any value. + assert!(found(r#"crs:NotAProperty="high" dc:notAProperty="T">"#).is_empty()); + assert!(found(">x").is_empty()); + // A namespace that isn't predefined, with a predefined property's local name. + assert!(found(r#"xmlns:ex="http://example.com/ns/" ex:Contrast="high">"#).is_empty()); + assert!(found(r#">T"#).is_empty()); + + // Through verify, the written name is said when it differs. + let xmp = format!( + r#""# + ); + let stream = format!("<< /Type /Metadata /Subtype /XML /Length {} >>\nstream\n{xmp}\nendstream", xmp.len()); + let doc = build(&["<< /Type /Catalog /Pages 2 0 R /Metadata 3 0 R >>", "<< /Type /Pages /Kids [] /Count 0 >>", &stream], ""); + let messages: Vec = verify(&doc, Level::A2b).into_iter().filter(|i| i.clause == "6.6.2.3.1").map(|i| i.message).collect(); + assert_eq!( + messages, + [ + "The XMP property crs:Contrast in metadata stream 3 0 R, written c:Contrast, should have the value type Integer", + "The XMP property crs:Exposure in metadata stream 3 0 R should have the value type Real", + ] + ); +} + +/// PDF/A-2 6.6.2.3.1, the RDF/XML value forms XMP allows (ISO 16684-1 7.9.2.5): one `rdf:value` +/// in a qualified value, no parse type but `Resource`, and an `rdf:resource` element left empty. +#[test] +fn xmp_value_forms_are_those_xmp_allows() { + // A qualified value has exactly one rdf:value. + assert_eq!( + mistyped(r#">5five"#), + ["crs:Contrast"] + ); + assert_eq!( + mistyped(r#">5"#), + ["crs:Contrast"] + ); + assert!(mistyped(r#">5"#).is_empty()); + // Only parseType="Resource" has a place in XMP; an rdf:resource element holds no element. + assert_eq!(mistyped(r#">5"#), ["crs:Contrast"]); + assert_eq!(mistyped(r#">"#), ["xmpMM:DerivedFrom"]); + assert_eq!( + mistyped(r#">"#), + ["xmpMM:DerivedFrom"] + ); +} diff --git a/crates/preflight/src/xmp.rs b/crates/preflight/src/xmp.rs index 7a54d70a1b..b2ea91c2f6 100644 --- a/crates/preflight/src/xmp.rs +++ b/crates/preflight/src/xmp.rs @@ -29,13 +29,18 @@ pub(crate) fn identifier(xmp: &str, namespace: &str, local: &str) -> Option Result, ()> { +/// `xmp` parsed within the depth, attribute and node limits; `None` when it can't be. +pub(crate) fn parse(xmp: &str) -> Option> { if !xml_limits_ok(xmp) { - return Err(()); + return None; } let options = roxmltree::ParsingOptions { allow_dtd: false, nodes_limit: 100_000, ..Default::default() }; - let doc = roxmltree::Document::parse_with_options(xmp, options).map_err(|_| ())?; + roxmltree::Document::parse_with_options(xmp, options).ok() +} + +/// [`identifier`] through an XML parser; `Err` when the packet can't be parsed within the limits. +fn parsed_identifier(xmp: &str, namespace: &str, local: &str) -> Result, ()> { + let doc = parse(xmp).ok_or(())?; for node in doc.descendants().filter(roxmltree::Node::is_element) { if let Some(value) = node.attribute((namespace, local)) { return Ok(Some(value.trim().to_owned())); diff --git a/crates/preflight/src/xmp_types.rs b/crates/preflight/src/xmp_types.rs new file mode 100644 index 0000000000..3563a671db --- /dev/null +++ b/crates/preflight/src/xmp_types.rs @@ -0,0 +1,780 @@ +//! PDF/A-2 6.6.2.3.1: a property of a predefined XMP schema has the value type the XMP +//! Specification gives it. +//! +//! The types come from ISO 16684-1:2011 (XMP Part 1) clause 8 for the Dublin Core, XMP Basic and +//! XMP Rights Management namespaces, and from XMP Specification Part 2 (Additional Properties) for +//! the Media Management, Basic Job Ticket, Paged-Text, Dynamic Media, Adobe PDF, Photoshop and +//! Camera Raw namespaces, with the structures they use. A value is read as RDF/XML serializes it +//! (ISO 16684-1 7.9): a simple value as text or an attribute, an array as `rdf:Seq`, `rdf:Bag` or +//! `rdf:Alt`, a structure as `rdf:parseType="Resource"`, an `rdf:Description` or attributes, and a +//! qualified value through its `rdf:value` (element or attribute). Every metadata stream of the +//! document is read, not only the catalog's, within one decode allowance. +//! +//! The simple types are checked by their lexical form (ISO 16684-1 8.2.1): Boolean is `True` or +//! `False`; Integer an optionally signed run of digits; Real a decimal; Date one of the six forms +//! of 8.2.1.2; Rational two integers around a `/`, as Part 2's `xmpDM:scale` example writes it. +//! The other simple types are text. A field a structure doesn't define, and a property a +//! namespace doesn't define, are left to the clauses about them. + +use std::collections::BTreeMap; + +use pdfcraft_cos::{DecodeBudget, ObjRef, Object}; +use roxmltree::Node; + +/// How deep arrays, structures and qualifiers are followed. No type nests nearly this deep, so a +/// value nested deeper doesn't have its type. +const MAX_DEPTH: usize = 32; +/// The decoded bytes all metadata streams of a document may take together. +const MAX_METADATA_BYTES: usize = 64 << 20; + +const RDF: &str = "http://www.w3.org/1999/02/22-rdf-syntax-ns#"; +const XML: &str = "http://www.w3.org/XML/1998/namespace"; + +/// A value type. +#[derive(Debug, Clone, Copy)] +enum Ty { + /// Any simple value: Text and the types derived from it (AgentName, Choice of Text, Locale, + /// MIMEType, ProperName, RenditionClass, URI, URL, XPath, and the like). + Text, + /// A GUID: any simple value, but not a URI (ISO 16684-1 8.2.2.3), so not an `rdf:resource`. + Guid, + Integer, + Real, + Boolean, + Date, + Rational, + /// An alternative array of text items, each with an `xml:lang` (ISO 16684-1 8.2.2.4). + LangAlt, + Seq(&'static Ty), + Bag(&'static Ty), + Alt(&'static Ty), + /// A structure: the name the specification gives its type, and its fields. + Struct(&'static str, &'static [Field]), + /// A structure whose fields aren't given (`xmpMM:Pantry`). + AnyStruct, + /// Any of these: where editions of the specification differ. + OneOf(&'static [Ty]), +} + +/// A structure field: namespace, local name, type. +type Field = (&'static str, &'static str, Ty); + +/// A predefined namespace: URI, preferred prefix, properties. +struct Schema { + uri: &'static str, + prefix: &'static str, + properties: &'static [(&'static str, Ty)], +} + +use Ty::{Boolean, Date, Integer, LangAlt, Rational, Real, Text}; + +const ST_DIM: &str = "http://ns.adobe.com/xap/1.0/sType/Dimensions#"; +// Part 2 1.2.2.3 prints the Font field namespace as "http:ns.adobe.com/xap/1.0/sType/Font#"; the +// other sType namespaces show the intended form. +const ST_FNT: &str = "http://ns.adobe.com/xap/1.0/sType/Font#"; +const XMP_G: &str = "http://ns.adobe.com/xap/1.0/g/"; +const XMP_G_IMG: &str = "http://ns.adobe.com/xap/1.0/g/img/"; +const ST_EVT: &str = "http://ns.adobe.com/xap/1.0/sType/ResourceEvent#"; +const ST_REF: &str = "http://ns.adobe.com/xap/1.0/sType/ResourceRef#"; +const ST_VER: &str = "http://ns.adobe.com/xap/1.0/sType/Version#"; +const ST_JOB: &str = "http://ns.adobe.com/xap/1.0/sType/Job#"; +const XMP_DM: &str = "http://ns.adobe.com/xmp/1.0/DynamicMedia/"; +const PHOTOSHOP: &str = "http://ns.adobe.com/photoshop/1.0/"; + +/// Part 2 Table 3. +const COLORANT: &[Field] = &[ + (XMP_G, "A", Integer), + (XMP_G, "B", Integer), + (XMP_G, "L", Real), + (XMP_G, "black", Real), + (XMP_G, "cyan", Real), + (XMP_G, "magenta", Real), + (XMP_G, "yellow", Real), + (XMP_G, "blue", Integer), + (XMP_G, "green", Integer), + (XMP_G, "red", Integer), + (XMP_G, "mode", Text), + (XMP_G, "swatchName", Text), + (XMP_G, "type", Text), +]; +/// Part 2 Table 4. +const DIMENSIONS: &[Field] = &[(ST_DIM, "h", Real), (ST_DIM, "w", Real), (ST_DIM, "unit", Text)]; +/// Part 2 Table 5. +const FONT: &[Field] = &[ + (ST_FNT, "childFontFiles", Ty::Seq(&Text)), + (ST_FNT, "composite", Boolean), + (ST_FNT, "fontFace", Text), + (ST_FNT, "fontFamily", Text), + (ST_FNT, "fontFileName", Text), + (ST_FNT, "fontName", Text), + (ST_FNT, "fontType", Text), + (ST_FNT, "versionString", Text), +]; +/// Part 2 Table 6. +const THUMBNAIL: &[Field] = &[(XMP_G_IMG, "format", Text), (XMP_G_IMG, "height", Integer), (XMP_G_IMG, "width", Integer), (XMP_G_IMG, "image", Text)]; +/// Part 2 Table 8. +const RESOURCE_EVENT: &[Field] = &[ + (ST_EVT, "action", Text), + (ST_EVT, "changed", Text), + (ST_EVT, "instanceID", Ty::Guid), + (ST_EVT, "parameters", Text), + (ST_EVT, "softwareAgent", Text), + (ST_EVT, "when", Date), +]; +/// ISO 16684-1 Table 3 and Part 2 Table 9. +const RESOURCE_REF: &[Field] = &[ + (ST_REF, "alternatePaths", Ty::Seq(&Text)), + (ST_REF, "documentID", Ty::Guid), + (ST_REF, "filePath", Text), + (ST_REF, "fromPart", Text), + (ST_REF, "instanceID", Ty::Guid), + (ST_REF, "lastModifyDate", Date), + (ST_REF, "manager", Text), + (ST_REF, "managerVariant", Text), + (ST_REF, "manageTo", Text), + (ST_REF, "manageUI", Text), + (ST_REF, "maskMarkers", Text), + (ST_REF, "partMapping", Text), + (ST_REF, "renditionClass", Text), + (ST_REF, "renditionParams", Text), + (ST_REF, "toPart", Text), + (ST_REF, "versionID", Text), +]; +/// Part 2 Table 10. +const VERSION: &[Field] = &[ + (ST_VER, "comments", Text), + (ST_VER, "event", Ty::Struct("ResourceEvent", RESOURCE_EVENT)), + (ST_VER, "modifier", Text), + (ST_VER, "modifyDate", Date), + (ST_VER, "version", Text), +]; +/// Part 2 Table 11. +const JOB: &[Field] = &[(ST_JOB, "id", Text), (ST_JOB, "name", Text), (ST_JOB, "url", Text)]; +/// Part 2 Table 12. +const BEAT_SPLICE_STRETCH: &[Field] = + &[(XMP_DM, "riseInDecibel", Real), (XMP_DM, "riseInTimeDuration", Ty::Struct("Time", TIME)), (XMP_DM, "useFileBeatsMarker", Boolean)]; +/// Part 2 Table 13. +const CUE_POINT_PARAM: &[Field] = &[(XMP_DM, "key", Text), (XMP_DM, "value", Text)]; +/// A marker's start and duration: FrameCount (text) in Part 2 Table 16, a Time structure in the +/// September 2005 edition that PDF/A-2 cites. Either is taken. +const FRAME_COUNT_OR_TIME: Ty = Ty::OneOf(&[Text, Ty::Struct("Time", TIME)]); +/// Part 2 Table 16. +const MARKER: &[Field] = &[ + (XMP_DM, "comment", Text), + (XMP_DM, "cuePointParams", Ty::Seq(&Ty::Struct("CuePointParam", CUE_POINT_PARAM))), + (XMP_DM, "cuePointType", Text), + (XMP_DM, "duration", FRAME_COUNT_OR_TIME), + (XMP_DM, "location", Text), + (XMP_DM, "name", Text), + (XMP_DM, "probability", Real), + (XMP_DM, "speaker", Text), + (XMP_DM, "startTime", FRAME_COUNT_OR_TIME), + (XMP_DM, "target", Text), + (XMP_DM, "type", Text), +]; +/// Part 2 Table 17. +const MEDIA: &[Field] = &[ + (XMP_DM, "duration", Ty::Struct("Time", TIME)), + (XMP_DM, "managed", Boolean), + (XMP_DM, "path", Text), + (XMP_DM, "startTime", Ty::Struct("Time", TIME)), + (XMP_DM, "track", Text), + (XMP_DM, "webStatement", Text), +]; +/// Part 2 Table 18. +const PROJECT_LINK: &[Field] = &[(XMP_DM, "path", Text), (XMP_DM, "type", Text)]; +/// Part 2 Table 19. +const RESAMPLE_STRETCH: &[Field] = &[(XMP_DM, "quality", Text)]; +/// Part 2 Table 20. +const TIME: &[Field] = &[(XMP_DM, "scale", Rational), (XMP_DM, "value", Integer)]; +/// Part 2 Table 21. +const TIMECODE: &[Field] = &[(XMP_DM, "timeFormat", Text), (XMP_DM, "timeValue", Text)]; +/// Part 2 Table 22. +const TIME_SCALE_STRETCH: &[Field] = &[(XMP_DM, "frameOverlappingPercentage", Real), (XMP_DM, "frameSize", Real), (XMP_DM, "quality", Text)]; +/// Part 2 Table 23. +const TRACK: &[Field] = &[ + (XMP_DM, "frameRate", Text), + (XMP_DM, "markers", Ty::Seq(&Ty::Struct("Marker", MARKER))), + (XMP_DM, "trackName", Text), + (XMP_DM, "trackType", Text), +]; +/// Part 2 Table 31. +const ANCESTOR: &[Field] = &[(PHOTOSHOP, "AncestorID", Text)]; +/// Part 2 Table 32. +const LAYER: &[Field] = &[(PHOTOSHOP, "LayerName", Text), (PHOTOSHOP, "LayerText", Text)]; + +const SCHEMAS: &[Schema] = &[ + // ISO 16684-1 Table 4. + Schema { + uri: "http://purl.org/dc/elements/1.1/", + prefix: "dc", + properties: &[ + ("contributor", Ty::Bag(&Text)), + ("coverage", Text), + ("creator", Ty::Seq(&Text)), + ("date", Ty::Seq(&Date)), + ("description", LangAlt), + ("format", Text), + ("identifier", Text), + ("language", Ty::Bag(&Text)), + ("publisher", Ty::Bag(&Text)), + ("relation", Ty::Bag(&Text)), + ("rights", LangAlt), + ("source", Text), + ("subject", Ty::Bag(&Text)), + ("title", LangAlt), + ("type", Ty::Bag(&Text)), + ], + }, + // ISO 16684-1 Table 5, with Part 2 Table 24. + Schema { + uri: "http://ns.adobe.com/xap/1.0/", + prefix: "xmp", + properties: &[ + ("Advisory", Ty::Bag(&Text)), + ("BaseURL", Text), + ("CreateDate", Date), + ("CreatorTool", Text), + ("Identifier", Ty::Bag(&Text)), + ("Label", Text), + ("MetadataDate", Date), + ("ModifyDate", Date), + ("Nickname", Text), + ("Rating", Real), + ("Thumbnails", Ty::Alt(&Ty::Struct("Thumbnail", THUMBNAIL))), + ], + }, + // ISO 16684-1 Table 6. + Schema { + uri: "http://ns.adobe.com/xap/1.0/rights/", + prefix: "xmpRights", + properties: &[("Certificate", Text), ("Marked", Boolean), ("Owner", Ty::Bag(&Text)), ("UsageTerms", LangAlt), ("WebStatement", Text)], + }, + // ISO 16684-1 Table 7, with Part 2 Tables 25 and 26. + Schema { + uri: "http://ns.adobe.com/xap/1.0/mm/", + prefix: "xmpMM", + properties: &[ + ("DerivedFrom", Ty::Struct("ResourceRef", RESOURCE_REF)), + ("DocumentID", Ty::Guid), + ("History", Ty::Seq(&Ty::Struct("ResourceEvent", RESOURCE_EVENT))), + ("Ingredients", Ty::Bag(&Ty::Struct("ResourceRef", RESOURCE_REF))), + ("InstanceID", Ty::Guid), + ("ManagedFrom", Ty::Struct("ResourceRef", RESOURCE_REF)), + ("Manager", Text), + ("ManageTo", Text), + ("ManageUI", Text), + ("ManagerVariant", Text), + ("OriginalDocumentID", Ty::Guid), + ("Pantry", Ty::Bag(&Ty::AnyStruct)), + ("RenditionClass", Text), + ("RenditionParams", Text), + ("VersionID", Text), + ("Versions", Ty::Seq(&Ty::Struct("Version", VERSION))), + ("LastURL", Text), + ("RenditionOf", Ty::Struct("ResourceRef", RESOURCE_REF)), + ("SaveID", Integer), + ], + }, + // Part 2 Table 27. + Schema { uri: "http://ns.adobe.com/xap/1.0/bj/", prefix: "xmpBJ", properties: &[("JobRef", Ty::Bag(&Ty::Struct("Job", JOB)))] }, + // Part 2 Table 28. + Schema { + uri: "http://ns.adobe.com/xap/1.0/t/pg/", + prefix: "xmpTPg", + properties: &[ + ("Colorants", Ty::Seq(&Ty::Struct("Colorant", COLORANT))), + ("Fonts", Ty::Bag(&Ty::Struct("Font", FONT))), + ("MaxPageSize", Ty::Struct("Dimensions", DIMENSIONS)), + ("NPages", Integer), + ("PlateNames", Ty::Seq(&Text)), + ], + }, + // Part 2 Table 29. + Schema { + uri: XMP_DM, + prefix: "xmpDM", + properties: &[ + ("absPeakAudioFilePath", Text), + ("album", Text), + ("altTapeName", Text), + ("altTimecode", Ty::Struct("Timecode", TIMECODE)), + ("artist", Text), + ("audioModDate", Date), + ("audioChannelType", Text), + ("audioCompressor", Text), + ("audioSampleRate", Integer), + ("audioSampleType", Text), + ("beatSpliceParams", Ty::Struct("BeatSpliceStretch", BEAT_SPLICE_STRETCH)), + ("cameraAngle", Text), + ("cameraLabel", Text), + ("cameraModel", Text), + ("cameraMove", Text), + ("client", Text), + ("comment", Text), + ("composer", Text), + ("contributedMedia", Ty::Bag(&Ty::Struct("Media", MEDIA))), + ("copyright", Text), + ("director", Text), + ("directorPhotography", Text), + ("duration", Ty::Struct("Time", TIME)), + ("engineer", Text), + ("fileDataRate", Rational), + ("genre", Text), + ("good", Boolean), + ("instrument", Text), + ("introTime", Ty::Struct("Time", TIME)), + ("key", Text), + ("logComment", Text), + ("loop", Boolean), + ("numberOfBeats", Real), + ("markers", Ty::Seq(&Ty::Struct("Marker", MARKER))), + ("metadataModDate", Date), + ("outCue", Ty::Struct("Time", TIME)), + ("projectName", Text), + ("projectRef", Ty::Struct("ProjectLink", PROJECT_LINK)), + ("pullDown", Text), + ("relativePeakAudioFilePath", Text), + ("relativeTimestamp", Ty::Struct("Time", TIME)), + ("releaseDate", Date), + ("resampleParams", Ty::Struct("ResampleStretch", RESAMPLE_STRETCH)), + ("scaleType", Text), + ("scene", Text), + ("shotDate", Date), + ("shotDay", Text), + ("shotLocation", Text), + ("shotName", Text), + ("shotNumber", Text), + ("shotSize", Text), + ("speakerPlacement", Text), + ("startTimecode", Ty::Struct("Timecode", TIMECODE)), + ("stretchMode", Text), + ("takeNumber", Integer), + ("tapeName", Text), + ("tempo", Real), + ("timeScaleParams", Ty::Struct("TimeScaleStretch", TIME_SCALE_STRETCH)), + ("timeSignature", Text), + ("trackNumber", Integer), + ("Tracks", Ty::Bag(&Ty::Struct("Track", TRACK))), + ("videoAlphaMode", Text), + ("videoAlphaPremultipleColor", Ty::Struct("Colorant", COLORANT)), + ("videoAlphaUnityIsTransparent", Boolean), + ("videoColorSpace", Text), + ("videoCompressor", Text), + ("videoFieldOrder", Text), + ("videoFrameRate", Text), + ("videoFrameSize", Ty::Struct("Dimensions", DIMENSIONS)), + ("videoModDate", Date), + ("videoPixelDepth", Text), + ("videoPixelAspectRatio", Rational), + ("partOfCompilation", Boolean), + ("lyrics", Text), + ("discNumber", Text), + ], + }, + // Part 2 Table 30. + Schema { + uri: "http://ns.adobe.com/pdf/1.3/", + prefix: "pdf", + properties: &[("Keywords", Text), ("PDFVersion", Text), ("Producer", Text), ("Trapped", Boolean)], + }, + // Part 2 Table 33. + Schema { + uri: PHOTOSHOP, + prefix: "photoshop", + properties: &[ + ("AuthorsPosition", Text), + ("CaptionWriter", Text), + ("Category", Text), + ("City", Text), + ("ColorMode", Integer), + ("Country", Text), + ("Credit", Text), + ("DateCreated", Date), + ("DocumentAncestors", Ty::Bag(&Ty::Struct("Ancestor", ANCESTOR))), + ("Headline", Text), + ("History", Text), + ("ICCProfile", Text), + ("Instructions", Text), + ("Source", Text), + ("State", Text), + ("SupplementalCategories", Ty::Bag(&Text)), + ("TextLayers", Ty::Seq(&Ty::Struct("Layer", LAYER))), + ("TransmissionReference", Text), + ("Urgency", Integer), + ], + }, + // Part 2 Table 34. + Schema { + uri: "http://ns.adobe.com/camera-raw-settings/1.0/", + prefix: "crs", + properties: &[ + ("AutoBrightness", Boolean), + ("AutoContrast", Boolean), + ("AutoExposure", Boolean), + ("AutoShadows", Boolean), + ("BlueHue", Integer), + ("BlueSaturation", Integer), + ("Brightness", Integer), + ("CameraProfile", Text), + ("ChromaticAberrationB", Integer), + ("ChromaticAberrationR", Integer), + ("ColorNoiseReduction", Integer), + ("Contrast", Integer), + ("CropTop", Real), + ("CropLeft", Real), + ("CropBottom", Real), + ("CropRight", Real), + ("CropAngle", Real), + ("CropWidth", Real), + ("CropHeight", Real), + ("CropUnits", Integer), + ("Exposure", Real), + ("GreenHue", Integer), + ("GreenSaturation", Integer), + ("HasCrop", Boolean), + ("HasSettings", Boolean), + ("LuminanceSmoothing", Integer), + ("RawFileName", Text), + ("RedHue", Integer), + ("RedSaturation", Integer), + ("Saturation", Integer), + ("Shadows", Integer), + ("ShadowTint", Integer), + ("Sharpness", Integer), + ("Temperature", Integer), + ("Tint", Integer), + ("ToneCurve", Ty::Seq(&Text)), + ("ToneCurveName", Text), + ("Version", Text), + ("VignetteAmount", Integer), + ("VignetteMidpoint", Integer), + ("WhiteBalance", Text), + ], + }, +]; + +/// Calls `f` with every metadata stream among the indirect objects `objs` (ISO 32000-1 14.3.2) that +/// parses as XML, and the stream's object, each decoded once within one allowance. +pub(crate) fn for_each_packet(objs: &[(ObjRef, std::sync::Arc)], mut f: impl FnMut(ObjRef, &roxmltree::Document<'_>)) { + let mut budget = DecodeBudget::new(MAX_METADATA_BYTES, 0); + for (r, o) in objs { + let Object::Stream(s) = &**o else { continue }; + if s.dict.name(b"Type") != Some(b"Metadata") { + continue; + } + if budget.remaining() == 0 { + break; + } + let Ok(bytes) = budget.decode_within(s, crate::xmp::MAX_XMP_BYTES) else { continue }; + let text = unicode(&bytes); + if let Some(parsed) = crate::xmp::parse(&text) { + f(*r, &parsed); + } + } +} + +/// A packet's text in the Unicode encoding its byte order mark gives: UTF-16 big- or +/// little-endian, or UTF-8 (ISO 16684-1 7.3). +fn unicode(bytes: &[u8]) -> String { + let utf16 = |data: &[u8], big: bool| -> String { + let (units, _) = data.as_chunks::<2>(); + let units = units.iter().map(|&[a, b]| if big { u16::from_be_bytes([a, b]) } else { u16::from_le_bytes([a, b]) }); + char::decode_utf16(units).map(|c| c.unwrap_or(char::REPLACEMENT_CHARACTER)).collect() + }; + match bytes { + [0xFE, 0xFF, rest @ ..] => utf16(rest, true), + [0xFF, 0xFE, rest @ ..] => utf16(rest, false), + [0xEF, 0xBB, 0xBF, rest @ ..] => String::from_utf8_lossy(rest).into_owned(), + _ => String::from_utf8_lossy(bytes).into_owned(), + } +} + +/// The properties of predefined schemas in the packet `doc` whose values don't have their type, in +/// name order: each by its name with the namespace's preferred prefix and by its name as the packet +/// writes it (see [`written`]), with the type it should have as [`named`] gives it. +pub(crate) fn mistyped(doc: &roxmltree::Document<'_>) -> BTreeMap<(String, String), String> { + let mut out = BTreeMap::new(); + let text = doc.input_text(); + let descriptions = doc.descendants().filter(|n| is(n, RDF, "Description") && n.parent_element().is_some_and(|p| is(&p, RDF, "RDF"))); + for d in descriptions { + for a in d.attributes() { + if let Some((schema, ty)) = a.namespace().and_then(|ns| property(ns, a.name())) + && !simple(ty, a.value()) + { + let qname = text.get(a.range_qname()).unwrap_or_default(); + out.insert((format!("{}:{}", schema.prefix, a.name()), written(qname, a.name())), named(ty)); + } + } + for p in d.children().filter(Node::is_element) { + let name = p.tag_name(); + if let Some((schema, ty)) = name.namespace().and_then(|ns| property(ns, name.name())) + && !fits(ty, &p, 0) + { + let qname = text.get(p.range()).unwrap_or_default().strip_prefix('<').unwrap_or_default(); + out.insert((format!("{}:{}", schema.prefix, name.name()), written(qname, name.name())), named(ty)); + } + } + } + out +} + +/// How long a prefix the packet writes is repeated. +const MAX_PREFIX_CHARS: usize = 64; + +/// The property `local` as the packet writes it, from the text its qualified name begins +/// (`qname`): `prefix:local`, or `local` alone in the default namespace. A prefix past +/// [`MAX_PREFIX_CHARS`] is cut, with `…`. +fn written(qname: &str, local: &str) -> String { + let qname = qname.split(|c: char| c.is_whitespace() || matches!(c, '>' | '/' | '=')).next().unwrap_or_default(); + let Some((prefix, _)) = qname.split_once(':') else { return local.to_owned() }; + let mut out: String = prefix.chars().take(MAX_PREFIX_CHARS).collect(); + if out.len() < prefix.len() { + out.push('…'); + } + format!("{out}:{local}") +} + +/// `ty` as the XMP Specification writes value types: `Integer`, `Lang Alt`, `seq Rational`, a +/// structure by its type's name, such as `Dimensions`. +fn named(ty: Ty) -> String { + let array = |kind: &str, item: &Ty| format!("{kind} {}", named(*item)); + match ty { + Text => "Text".into(), + Ty::Guid => "GUID".into(), + Integer => "Integer".into(), + Real => "Real".into(), + Boolean => "Boolean".into(), + Date => "Date".into(), + Rational => "Rational".into(), + LangAlt => "Lang Alt".into(), + Ty::Seq(item) => array("seq", item), + Ty::Bag(item) => array("bag", item), + Ty::Alt(item) => array("alt", item), + Ty::Struct(name, _) => name.into(), + Ty::AnyStruct => "struct".into(), + Ty::OneOf(types) => types.iter().map(|t| named(*t)).collect::>().join(" or "), + } +} + +fn property(ns: &str, name: &str) -> Option<(&'static Schema, Ty)> { + let schema = SCHEMAS.iter().find(|s| s.uri == ns)?; + schema.properties.iter().find(|(n, _)| *n == name).map(|(_, t)| (schema, *t)) +} + +fn is(n: &Node<'_, '_>, ns: &str, name: &str) -> bool { + n.is_element() && n.tag_name().namespace() == Some(ns) && n.tag_name().name() == name +} + +/// The attributes of `n` that are fields or qualifiers, not RDF or XML syntax. +fn field_attributes<'a, 'i>(n: &Node<'a, 'i>) -> impl Iterator> { + n.attributes().filter(|a| !matches!(a.namespace(), Some(RDF) | Some(XML))) +} + +/// Whether the value of the property or array item element `e` has type `ty`. +fn fits(ty: Ty, e: &Node<'_, '_>, depth: usize) -> bool { + if depth > MAX_DEPTH { + return false; + } + let deeper = depth + 1; + let children: Vec> = e.children().filter(Node::is_element).collect(); + // A qualified value: `rdf:value` holds the value itself (ISO 16684-1 7.9.2.5). + let resource = e.attribute((RDF, "parseType")) == Some("Resource"); + // `parseType="Resource"` is the only parse type XMP has (ISO 16684-1 7.9.2.5); an + // `rdf:resource` element is empty. + if e.attribute((RDF, "parseType")).is_some_and(|p| p != "Resource") || e.attribute((RDF, "resource")).is_some() && !children.is_empty() { + return false; + } + let description = match children.as_slice() { + [d] if is(d, RDF, "Description") => Some(*d), + _ => None, + }; + // `rdf:value` as an element of a `parseType="Resource"` node or `rdf:Description`, or as an + // attribute of the description or of an empty property element. Its qualifiers that are + // predefined properties keep their types too. + let node = if resource { Some(*e) } else { description }; + if let Some(n) = node { + // A qualified value has one `rdf:value`: an element, or an attribute in its place. + match n.children().filter(|c| is(c, RDF, "value")).collect::>().as_slice() { + [] => {} + [v] => return n.attribute((RDF, "value")).is_none() && blank(e) && blank(&n) && fits(ty, v, deeper) && predefined_ok(&n, deeper), + _ => return false, + } + } + let attribute_node = description.or_else(|| (children.is_empty() && !resource).then_some(*e)); + if let Some(n) = attribute_node + && let Some(v) = n.attribute((RDF, "value")) + { + return blank(e) && blank(&n) && simple(ty, v) && predefined_ok(&n, deeper); + } + match ty { + Ty::Text | Ty::Guid | Ty::Integer | Ty::Real | Ty::Boolean | Ty::Date | Ty::Rational => { + // `rdf:resource` makes the value a URI, which only a text type takes. + // The element is then empty. + if e.attribute((RDF, "resource")).is_some() { + return matches!(ty, Ty::Text) && children.is_empty() && blank(e); + } + children.is_empty() && !resource && field_attributes(e).next().is_none() && simple(ty, &text(e)) + } + Ty::LangAlt => match array(e, &children, "Alt") { + Some(items) => items.iter().all(|i| i.attribute((XML, "lang")).is_some() && fits(Ty::Text, i, deeper)), + None => false, + }, + Ty::Seq(item) | Ty::Bag(item) | Ty::Alt(item) => { + let kind = match ty { + Ty::Seq(_) => "Seq", + Ty::Bag(_) => "Bag", + _ => "Alt", + }; + array(e, &children, kind).is_some_and(|items| items.iter().all(|i| fits(*item, i, deeper))) + } + Ty::Struct(_, fields) => { + let Some(holder) = structure(e, &children, resource, description) else { return false }; + let attrs_ok = field_attributes(&holder).all(|a| match a.namespace().and_then(|ns| field(fields, ns, a.name())) { + Some(t) => simple(t, a.value()), + None => true, + }); + let elements_ok = holder.children().filter(Node::is_element).all(|f| { + let n = f.tag_name(); + match n.namespace().and_then(|ns| field(fields, ns, n.name())) { + Some(t) => fits(t, &f, deeper), + None => true, + } + }); + attrs_ok && elements_ok + } + // A structure of any fields, such as a Pantry item: predefined properties in it still have + // their types. + Ty::AnyStruct => structure(e, &children, resource, description).is_some_and(|holder| predefined_ok(&holder, deeper)), + Ty::OneOf(types) => types.iter().any(|t| fits(*t, e, depth)), + } +} + +/// Whether the fields of `holder`, as attributes and elements, that are predefined properties have +/// their types. +fn predefined_ok(holder: &Node<'_, '_>, depth: usize) -> bool { + let attrs_ok = field_attributes(holder).all(|a| a.namespace().and_then(|ns| property(ns, a.name())).is_none_or(|(_, t)| simple(t, a.value()))); + let elements_ok = holder.children().filter(Node::is_element).all(|f| { + let n = f.tag_name(); + n.namespace().and_then(|ns| property(ns, n.name())).is_none_or(|(_, t)| fits(t, &f, depth)) + }); + attrs_ok && elements_ok +} + +fn field(fields: &[Field], ns: &str, name: &str) -> Option { + fields.iter().find(|(n, l, _)| *n == ns && *l == name).map(|(_, _, t)| *t) +} + +/// The items of `e`'s value when it is an `rdf:` array. +fn array<'a, 'i>(e: &Node<'a, 'i>, children: &[Node<'a, 'i>], kind: &str) -> Option>> { + match children { + // Items are `rdf:li` (ISO 16684-1 7.9.2.4); any other element, or character data around + // them, breaks the array, and so does `rdf:parseType` or `rdf:resource` on the property, + // which makes its value something else. + [a] if is(a, RDF, kind) && field_attributes(e).next().is_none() && blank(e) && blank(a) && !other_value_form(e) => { + let items: Vec> = a.children().filter(Node::is_element).collect(); + items.iter().all(|i| is(i, RDF, "li")).then_some(items) + } + _ => None, + } +} + +/// The node holding the fields of `e`'s value when it is a structure. +fn structure<'a, 'i>(e: &Node<'a, 'i>, children: &[Node<'a, 'i>], resource: bool, description: Option>) -> Option> { + if resource { + // A structure's fields are elements; character data has no place in it. + return text(e).trim().is_empty().then_some(*e); + } + if let Some(d) = description { + return (blank(e) && blank(&d)).then_some(d); + } + // Fields written as attributes of the property element, with no content. + let empty = text(e).trim().is_empty(); + (children.is_empty() && empty && e.attribute((RDF, "resource")).is_none() && field_attributes(e).next().is_some()).then_some(*e) +} + +/// Whether `e` says its value is a resource or a literal of another kind, not the element it holds. +fn other_value_form(e: &Node<'_, '_>) -> bool { + e.attribute((RDF, "parseType")).is_some() || e.attribute((RDF, "resource")).is_some() +} + +/// Whether `e` holds no character data but white space. +fn blank(e: &Node<'_, '_>) -> bool { + text(e).trim().is_empty() +} + +/// All the character data directly in `e`, however comments or other nodes split it. +fn text(e: &Node<'_, '_>) -> String { + e.children().filter(Node::is_text).filter_map(|t| t.text()).collect() +} + +/// Whether the simple value `v` has the simple type `ty`. +fn simple(ty: Ty, v: &str) -> bool { + match ty { + Ty::Integer => integer(v), + Ty::Real => real(v), + Ty::Boolean => matches!(v, "True" | "False"), + Ty::Date => date(v), + Ty::Rational => v.split_once('/').is_some_and(|(n, d)| integer(n) && integer(d)), + Ty::Text | Ty::Guid => true, + Ty::OneOf(types) => types.iter().any(|t| simple(*t, v)), + _ => false, + } +} + +fn digits(s: &str) -> bool { + !s.is_empty() && s.bytes().all(|b| b.is_ascii_digit()) +} + +/// ISO 16684-1 8.2.1.3. +fn integer(v: &str) -> bool { + digits(v.strip_prefix(['+', '-']).unwrap_or(v)) +} + +/// ISO 16684-1 8.2.1.4: an integer part, a fraction part, or both. +fn real(v: &str) -> bool { + let v = v.strip_prefix(['+', '-']).unwrap_or(v); + match v.split_once('.') { + Some((i, f)) => (i.is_empty() || digits(i)) && digits(f), + None => digits(v), + } +} + +/// ISO 16684-1 8.2.1.2: YYYY, YYYY-MM, YYYY-MM-DD, or a date with hh:mm, hh:mm:ss or hh:mm:ss.s +/// and an optional time zone (Z, +hh:mm, -hh:mm). +fn date(v: &str) -> bool { + let num = |s: &str, len: usize, max: u32| s.len() == len && digits(s) && s.parse::().is_ok_and(|n| n <= max); + let in_range = |s: &str, min: u32, max: u32| num(s, 2, max) && s.parse::().is_ok_and(|n| n >= min); + let (day, time) = match v.split_once('T') { + Some((d, t)) => (d, Some(t)), + None => (v, None), + }; + let parts: Vec<&str> = day.split('-').collect(); + let day_ok = match parts.as_slice() { + [y] => num(y, 4, 9999), + [y, m] => num(y, 4, 9999) && in_range(m, 1, 12), + [y, m, d] => num(y, 4, 9999) && in_range(m, 1, 12) && in_range(d, 1, 31), + _ => false, + }; + let Some(time) = time else { return day_ok }; + if parts.len() != 3 || !day_ok { + return false; + } + let (clock, zone) = match time.find(['Z', '+', '-']) { + Some(i) => (time.get(..i).unwrap_or_default(), time.get(i..).unwrap_or_default()), + None => (time, ""), + }; + let zone_ok = match zone { + "" | "Z" => true, + z => z.strip_prefix(['+', '-']).and_then(|z| z.split_once(':')).is_some_and(|(h, m)| num(h, 2, 23) && num(m, 2, 59)), + }; + let clock_ok = match clock.split(':').collect::>().as_slice() { + [h, m] => num(h, 2, 23) && num(m, 2, 59), + [h, m, s] => { + let (whole, fraction) = s.split_once('.').map_or((*s, None), |(w, f)| (w, Some(f))); + num(h, 2, 23) && num(m, 2, 59) && num(whole, 2, 59) && fraction.is_none_or(digits) + } + _ => false, + }; + zone_ok && clock_ok +} diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index 9b341fcd8b..b2e651ec57 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -8017,8 +8017,8 @@ milestone = "M11" status = "partial" commands = ["standards.pdfa"] tools = ["pdfa_verify"] -evidence = ["crates/preflight/src/tests.rs::verify_reports_what_breaks_pdfa", "crates/preflight/src/tests.rs::standards_identification_uses_namespace_not_prefix", "crates/preflight/src/tests.rs::convert_fixes_all_but_fonts", "crates/preflight/src/tests.rs::verify_reports_an_output_intent_that_does_not_match_device_colour", "crates/automation/tests/it/automation.rs::pdfa_through_tools", "crates/ui-egui/tests/it/compare.rs::pdfa_dialog_verifies_and_converts"] -notes = "PDF/A-2b/3b structural rules: identification, encryption, output intent (present, and RGB/CMYK to match DeviceRGB/DeviceCMYK use), fonts embedded, annotations, actions, images, embedded files. Not yet: content-stream operator checks, transparency details, PDF/A-1 and -4, a veraPDF comparison." +evidence = ["crates/preflight/src/tests.rs::verify_reports_what_breaks_pdfa", "crates/preflight/src/tests.rs::standards_identification_uses_namespace_not_prefix", "crates/preflight/src/tests.rs::convert_fixes_all_but_fonts", "crates/preflight/src/tests.rs::verify_reports_an_output_intent_that_does_not_match_device_colour", "crates/preflight/src/tests.rs::xmp_simple_values_have_their_lexical_forms", "crates/preflight/src/tests.rs::xmp_arrays_have_their_kind_and_item_types", "crates/preflight/src/tests.rs::xmp_structures_have_their_field_types", "crates/preflight/src/tests.rs::xmp_value_types_are_reported_by_verify", "crates/preflight/src/tests.rs::xmp_mistyped_properties_name_their_type", "crates/preflight/src/tests.rs::xmp_properties_are_known_by_namespace_uri_and_named_as_written", "crates/automation/tests/it/automation.rs::pdfa_through_tools", "crates/ui-egui/tests/it/compare.rs::pdfa_dialog_verifies_and_converts"] +notes = "PDF/A-2b/3b structural rules: identification, the value types of the predefined XMP schemas (6.6.2.3.1: dc, xmp, xmpRights, xmpMM, xmpBJ, xmpTPg, xmpDM, pdf, photoshop, crs and their structures; the tiff, exif and aux namespaces, whose tables XMP Part 2 defers to CIPA DC-010, are not yet checked), encryption, output intent (present, and RGB/CMYK to match DeviceRGB/DeviceCMYK use), fonts embedded, annotations, actions, images, embedded files. Not yet: content-stream operator checks, transparency details, PDF/A-1 and -4, a veraPDF comparison." [[feature]] id = "optimize.pdfa-convert" From 6062cd6a4481b15762fbadd01011a00f7d739a75 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:38:25 +0800 Subject: [PATCH 13/75] M5: Replace Text keeps its replacement on the group's primary, the caret (comment.replace-text) (#5689) --- crates/annot/src/modify.rs | 95 ++++++++++-- crates/annot/src/tests.rs | 77 +++++++++- crates/automation/src/comments.rs | 6 +- crates/automation/tests/it/automation.rs | 23 ++- crates/engine/src/edit/run.rs | 4 +- crates/ui-egui/src/panels/comments_panel.rs | 2 +- crates/ui-egui/src/tools/comments.rs | 2 +- crates/ui-egui/src/tools/comments/composer.rs | 5 +- crates/ui-egui/src/tools/comments/geometry.rs | 7 + crates/ui-egui/src/tools/comments/input.rs | 16 +- crates/ui-egui/src/tools/comments/state.rs | 18 ++- crates/ui-egui/tests/it/comments.rs | 48 +++++- crates/xfdf/src/tests.rs | 139 +++++++++++++++++- crates/xfdf/src/xfdf.rs | 28 ++-- parity/acrobat-features.toml | 2 +- 15 files changed, 414 insertions(+), 58 deletions(-) diff --git a/crates/annot/src/modify.rs b/crates/annot/src/modify.rs index 60e97f6a72..6fbb0f35c3 100644 --- a/crates/annot/src/modify.rs +++ b/crates/annot/src/modify.rs @@ -94,7 +94,7 @@ pub fn deletable(doc: &Document, page: usize, index: usize) -> Result<(), AnnotE if let Some(r) = target.as_ref() { unlocked(doc, r)?; } - let doomed = with_replies_and_popups(doc, &list, seed(target)); + let doomed = with_replies_and_popups(doc, &list, seed(doc, target)); // A locked reply can't be deleted on its own, so it doesn't go with the comment or reply it // answers either. Pop-ups follow their comment. for (_, e) in list.iter().enumerate().filter(|(i, e)| *i != index && goes_with(e, &doomed)) { @@ -115,15 +115,28 @@ pub fn lift_annotation(doc: &mut Document, page: usize, index: usize) -> Result< let p = page_ref(doc, page)?; let list = annots(doc, p); let target = list.get(index).ok_or(AnnotError::NoSuchAnnotation { page, index })?; - let doomed = with_replies_and_popups(doc, &list, seed(target)); + let doomed = with_replies_and_popups(doc, &list, seed(doc, target)); let kept: Vec = list.iter().enumerate().filter(|(i, e)| *i != index && !goes_with(e, &doomed)).map(|(_, e)| e.clone()).collect(); Ok(set_annots(doc, p, kept)?) } /// What deleting the `/Annots` entry `target` starts from: its reference, or for an inline -/// comment (which has none to collect) the pop-up it names, which still goes. -fn seed(target: &Object) -> Option { - target.as_ref().or_else(|| target.as_dict().and_then(|d| d.reference(b"Popup"))) +/// comment (which has none to collect) the pop-up it names, which still goes. The strikeout of a +/// Replace Text proposal also takes the caret it is grouped under, so the two go together +/// whichever is deleted. +fn seed(doc: &Document, target: &Object) -> Vec { + let own = target.as_ref().or_else(|| target.as_dict().and_then(|d| d.reference(b"Popup"))); + own.into_iter().chain(replaced_by(doc, target)).collect() +} + +/// The caret a Replace Text strikeout is grouped under (`/IT /StrikeOutTextEdit`, `/RT /Group`, +/// `/IRT` to a caret): the proposal's primary, which holds the replacement. +fn replaced_by(doc: &Document, strike: &Object) -> Option { + let obj = doc.resolve(strike); + let d = obj.as_dict()?; + let edit = d.name(b"Subtype") == Some(b"StrikeOut") && d.name(b"IT") == Some(b"StrikeOutTextEdit") && d.name(b"RT") == Some(b"Group"); + let caret = d.reference(b"IRT").filter(|_| edit)?; + (annot_dict(doc, caret).name(b"Subtype") == Some(b"Caret")).then_some(caret) } /// Whether the `/Annots` entry `entry` goes with `going` (from [`with_replies_and_popups`]): it is @@ -713,9 +726,61 @@ fn ink_rect_for_width(d: &Dict, old_w: f64, w: f64) -> Option<[f64; 4]> { /// Change a comment's colour, opacity (of the whole comment), line width and/or line endings, and /// redraw it. +/// +/// A Replace Text proposal is one look: colour and opacity are the group's (`/C` is read from +/// the primary, ISO 32000-2 §12.5.6.2), so restyling either the caret or its strikeout restyles +/// and redraws the other too, in the same edit. The strikeout takes no line width (a text markup +/// annotation has no border) and neither part has line endings. All or nothing: if one member +/// can't be restyled, none is. pub fn set_style(doc: &mut Document, page: usize, index: usize, change: StyleChange, meta: &Meta) -> Result<(), AnnotError> { - let StyleChange { color, opacity, width, endings } = change; let (_, r) = annot_ref(doc, page, index)?; + let others = replace_text_partners(doc, page, r)?; + // Check every partner before changing anything. + for &o in &others { + unlocked(doc, o)?; + } + let before: Vec<(ObjRef, Dict)> = others.iter().map(|&o| (o, annot_dict(doc, o))).collect(); + let mut done: Vec<(ObjRef, Dict)> = vec![(r, annot_dict(doc, r))]; + style_one(doc, r, change.clone(), meta)?; + for (o, d) in before { + let caret = d.name(b"Subtype") == Some(b"Caret"); + let part = StyleChange { color: change.color, opacity: change.opacity, width: change.width.filter(|_| caret), endings: None }; + let unchanged = part.color.is_none() && part.opacity.is_none() && part.width.is_none(); + if unchanged { + continue; + } + if let Err(e) = style_one(doc, o, part, meta) { + // Put the members already restyled back, the target included. + for (o, d) in done.into_iter().rev() { + doc.set(o, Object::Dict(d)); + } + return Err(e); + } + done.push((o, d)); + } + Ok(()) +} + +/// The other members of the Replace Text proposal `r` belongs to, on `page`: for its caret, the +/// strikeouts grouped under it; for a strikeout, its caret and any other strikeout under it. +/// Empty for anything else. +fn replace_text_partners(doc: &Document, page: usize, r: ObjRef) -> Result, AnnotError> { + let own = Object::Ref(r); + let caret = match replaced_by(doc, &own) { + Some(c) => c, + None if annot_dict(doc, r).name(b"Subtype") == Some(b"Caret") => r, + None => return Ok(Vec::new()), + }; + let list = annots(doc, page_ref(doc, page)?); + let strikes = list.iter().filter(|e| replaced_by(doc, e) == Some(caret)).filter_map(Object::as_ref); + // `/Annots` may list one annotation twice: restyle each once. + let mut seen: HashSet = HashSet::from([r]); + Ok(std::iter::once(caret).chain(strikes).filter(|&o| seen.insert(o)).collect()) +} + +/// [`set_style`] on one annotation. +fn style_one(doc: &mut Document, r: ObjRef, change: StyleChange, meta: &Meta) -> Result<(), AnnotError> { + let StyleChange { color, opacity, width, endings } = change; unlocked(doc, r)?; let d = annot_dict(doc, r); let subtype = String::from_utf8_lossy(d.name(b"Subtype").unwrap_or_default()).into_owned(); @@ -994,8 +1059,13 @@ fn pair_points(d: &Dict, key: &[u8]) -> Option> { } /// Replace Text (Acrobat's proposal): strike out `quads` and add a caret at the end of the -/// struck text holding `replacement`, grouped with the strikeout (`/IRT` + `/RT /Group`) so they -/// move, list and delete as one. Returns the strikeout's index. +/// struck text holding `replacement`, the two grouped so they move, list and delete as one. +/// Returns the strikeout's index. +/// +/// The caret is the group's primary (`/IT /Replace`) and the strikeout its subordinate (`/IRT` to +/// the caret, `/RT /Group`, `/IT /StrikeOutTextEdit`): a group's text is read from its primary, +/// and the subordinates' `Contents` are ignored (ISO 32000-2 §12.5.6.2), so the replacement has +/// to be on the primary. /// /// The two annotations need their own `/NM`: XFDF names a reply by the `/NM` of what it replies /// to, so one id for both writes `inreplyto` equal to the caret's own `name`, and a reader that @@ -1016,7 +1086,6 @@ pub fn add_text_replacement(doc: &mut Document, replacement: TextReplacement<'_> meta, )?; let (_, sr) = annot_ref(doc, page, s)?; - doc.update_dict(sr, |d| d.set(b"IT".to_vec(), Object::name("StrikeOutTextEdit")))?; // The caret sits on the baseline at the end of the last line, its point at the text. let xs = [last[0], last[2], last[4], last[6]]; let ys = [last[1], last[3], last[5], last[7]]; @@ -1031,10 +1100,14 @@ pub fn add_text_replacement(doc: &mut Document, replacement: TextReplacement<'_> )?; let (_, cr) = annot_ref(doc, page, c)?; doc.update_dict(cr, |d| { - d.set(b"IRT".to_vec(), Object::Ref(sr)); - d.set(b"RT".to_vec(), Object::name("Group")); + d.set(b"IT".to_vec(), Object::name("Replace")); d.set(b"Subj".to_vec(), PdfString::text("Replace Text")); })?; + doc.update_dict(sr, |d| { + d.set(b"IT".to_vec(), Object::name("StrikeOutTextEdit")); + d.set(b"IRT".to_vec(), Object::Ref(cr)); + d.set(b"RT".to_vec(), Object::name("Group")); + })?; Ok(s) } diff --git a/crates/annot/src/tests.rs b/crates/annot/src/tests.rs index 9e414078e9..7561938676 100644 --- a/crates/annot/src/tests.rs +++ b/crates/annot/src/tests.rs @@ -1264,22 +1264,36 @@ fn replace_text_groups_a_strikeout_and_a_caret() { let strike = &a[i]; assert_eq!((strike.name(b"Subtype"), strike.name(b"IT")), (Some(&b"StrikeOut"[..]), Some(&b"StrikeOutTextEdit"[..]))); let caret = a.iter().find(|d| d.name(b"Subtype") == Some(b"Caret")).unwrap(); - assert_eq!((text(caret, b"Contents").as_str(), caret.name(b"RT")), ("new words", Some(&b"Group"[..]))); - assert!(caret.reference(b"IRT").is_some()); + // The caret is the group's primary: a group's text is read from its primary and the + // subordinates' Contents are ignored (ISO 32000-2 §12.5.6.2), so the replacement is on it. + assert_eq!((text(caret, b"Contents").as_str(), caret.name(b"IT")), ("new words", Some(&b"Replace"[..]))); + assert!(caret.get(b"IRT").is_none() && caret.get(b"RT").is_none(), "a primary has no /IRT: {caret:?}"); + assert_eq!(strike.name(b"RT"), Some(&b"Group"[..])); + assert!(text(strike, b"Contents").is_empty()); // #2666: the pair needs its own `/NM` each. XFDF names a reply by the `/NM` of what it // replies to, so sharing one id writes `inreplyto` equal to the caret's own `name` and the // pair does not survive an export and import. assert_eq!(text(caret, b"NM").as_str(), "r-caret"); assert_eq!(text(strike, b"NM").as_str(), "r"); assert_ne!(text(caret, b"NM"), text(strike, b"NM"), "the caret must not carry the strikeout's name"); - let irt = caret.reference(b"IRT").unwrap(); - assert_eq!(text(doc.get(irt).as_dict().unwrap(), b"NM").as_str(), "r", "the caret replies to the strikeout"); + let irt = strike.reference(b"IRT").unwrap(); + assert_eq!(text(doc.get(irt).as_dict().unwrap(), b"NM").as_str(), "r-caret", "the strikeout is grouped under the caret"); let r = rect(caret); assert!(r[0] < 200.0 && r[2] > 200.0 && r[1] < 706.0, "at the end of the line, below its baseline: {r:?}"); - // Deleting the strikeout takes the caret with it (the group). + // Deleting either one takes the other with it (the proposal is one comment). + let marks = |doc: &Document| list(doc, 0).iter().filter(|d| matches!(d.name(b"Subtype"), Some(b"Caret" | b"StrikeOut"))).count(); + let mut by_strike = doc.clone(); + delete_annotation(&mut by_strike, 0, i).unwrap(); + assert_eq!(marks(&by_strike), 0, "deleting the strikeout deletes the caret"); let mut doc = doc; - delete_annotation(&mut doc, 0, i).unwrap(); - assert!(!list(&doc, 0).iter().any(|d| d.name(b"Subtype") == Some(b"Caret"))); + let c = list(&doc, 0).iter().position(|d| d.name(b"Subtype") == Some(b"Caret")).unwrap(); + // A locked caret keeps the strikeout too: the pair can't be broken by deleting the other. + let mut locked = doc.clone(); + set_locked(&mut locked, 0, c, true).unwrap(); + assert!(deletable(&locked, 0, i).is_err() && delete_annotation(&mut locked, 0, i).is_err(), "a locked caret refuses its strikeout's deletion"); + assert_eq!(marks(&locked), 2); + delete_annotation(&mut doc, 0, c).unwrap(); + assert_eq!(marks(&doc), 0, "deleting the caret deletes the strikeout"); assert!( add_text_replacement( &mut doc, @@ -1298,6 +1312,55 @@ fn replace_text_groups_a_strikeout_and_a_caret() { ); } +/// Review on #5689 (CodeRabbit): clicking a Replace Text strikeout selects its caret, the +/// group's primary, so a restyle from Comment Properties lands on the caret. Colour and opacity +/// are the proposal's, so the strikeout is restyled and redrawn in the same edit, whichever +/// member is restyled; it takes no border width. A locked member refuses the whole change. +#[test] +fn restyling_a_replace_text_pair_restyles_both_members() { + let mut doc = fixture(); + let quads = [[100.0, 720.0, 200.0, 720.0, 100.0, 706.0, 200.0, 706.0]]; + let red = Style { color: [0.89, 0.13, 0.13], opacity: 1.0, width: 1.0, fill: None, ..Style::default() }; + let tr = TextReplacement { + page: 0, + quads: &quads, + replacement: "new words", + author: "Ada", + strike: &red, + caret: &red, + strike_meta: &meta("r"), + caret_meta: &meta("r-caret"), + }; + let s = add_text_replacement(&mut doc, tr).unwrap(); + let c = list(&doc, 0).iter().position(|d| d.name(b"Subtype") == Some(b"Caret")).unwrap(); + let color = |doc: &Document, i: usize| { + list(doc, 0)[i].get(b"C").and_then(|o| o.as_array().cloned()).unwrap().iter().filter_map(Object::as_f64).collect::>() + }; + let green = [0.0, 0.5, 0.0]; + + // Restyled through the caret (what a click on either part selects). + let mut by_caret = doc.clone(); + set_style(&mut by_caret, 0, c, StyleChange::new(Some(green), Some(0.5), Some(3.0), None), &meta("")).unwrap(); + let by_caret = reopen(&by_caret); + assert_eq!(color(&by_caret, c), green.to_vec()); + assert_eq!(color(&by_caret, s), green.to_vec(), "the strikeout takes the proposal's colour"); + let strike = &list(&by_caret, 0)[s]; + assert_eq!(strike.get(b"CA").and_then(Object::as_f64), Some(0.5), "and its opacity"); + assert!(strike.get(b"BS").is_none(), "a strikeout has no border to widen: {strike:?}"); + assert!(ap_content(&by_caret, strike).contains("0 0.5 0 RG"), "its appearance is redrawn: {}", ap_content(&by_caret, strike)); + + // Restyled through the strikeout: the caret follows. + let mut by_strike = doc.clone(); + set_style(&mut by_strike, 0, s, StyleChange::new(Some(green), None, None, None), &meta("")).unwrap(); + assert_eq!(color(&by_strike, c), green.to_vec(), "the caret takes the proposal's colour"); + + // A locked strikeout refuses the change, and nothing moves. + let mut locked = doc.clone(); + set_locked(&mut locked, 0, s, true).unwrap(); + assert!(set_style(&mut locked, 0, c, StyleChange::new(Some(green), None, None, None), &meta("")).is_err()); + assert_eq!(color(&locked, c), red.color.to_vec(), "the caret keeps its colour"); +} + #[test] fn files_attach_as_comments() { let mut doc = fixture(); diff --git a/crates/automation/src/comments.rs b/crates/automation/src/comments.rs index 78f4cefbe0..390e3b4d1a 100644 --- a/crates/automation/src/comments.rs +++ b/crates/automation/src/comments.rs @@ -490,8 +490,12 @@ impl Automation { if contents.trim().is_empty() { return Err(ToolError::InvalidArgs("replace needs `contents`: the replacement text".into())); } + // `/C` is a group attribute, taken from the primary (ISO 32000-2 §12.5.6.2), so the + // caret carries the colour too. Only the colour: opacity and the rest are the + // caret's own. + let mut caret = Style::default_for(&Shape::Caret { rect: [0.0; 4] }); + caret.color = style.color; let strike = style; - let caret = Style::default_for(&Shape::Caret { rect: [0.0; 4] }); let mut out = self.apply(a, Edit::ReplaceText { page, quads, text: contents, author, strike, caret })?; out["comment"] = json!({ "page": page + 1, "type": "replace" }); return Ok(out); diff --git a/crates/automation/tests/it/automation.rs b/crates/automation/tests/it/automation.rs index 0849f8fa65..da634490e4 100644 --- a/crates/automation/tests/it/automation.rs +++ b/crates/automation/tests/it/automation.rs @@ -6783,11 +6783,26 @@ fn drawing_comments_through_tools() { a.call("comment_add", &json!({ "doc": doc, "page": 1, "type": "polygon", "points": [[1, 1]] })), Err(ToolError::Failed(_) | ToolError::InvalidArgs(_)) )); - // Replace Text: a strikeout over the found text and a grouped caret. - ok(&mut a, "comment_add", json!({ "doc": doc, "page": 2, "type": "replace", "find": "page 2", "contents": "second page" })); + // Replace Text: a strikeout over the found text, grouped under a caret holding the + // replacement (the group's primary, ISO 32000-2 §12.5.6.2). + ok(&mut a, "comment_add", json!({ "doc": doc, "page": 2, "type": "replace", "find": "page 2", "contents": "second page", "color": "green" })); let list = ok(&mut a, "comment_list", json!({ "doc": doc, "page": 2 })); - let strike = list["comments"].as_array().unwrap().iter().find(|c| c["type"] == "StrikeOut").unwrap().clone(); - assert_eq!(strike["replies"][0]["contents"], "second page", "the caret threads under the strikeout"); + let caret = list["comments"].as_array().unwrap().iter().find(|c| c["type"] == "Caret").unwrap().clone(); + assert_eq!(caret["contents"], "second page", "the replacement is on the caret"); + assert_eq!(caret["replies"][0]["type"], "StrikeOut", "the strikeout is grouped under the caret"); + // Review on #5689 (CodeRabbit): the requested colour is the group's, so the primary has it. + assert_eq!(caret["replies"][0]["color"], caret["color"], "one colour for the pair: {caret}"); + assert_eq!(caret["color"], json!("#2E9E5C"), "the requested green, not the caret's default blue: {caret}"); + // Restyling the caret restyles the proposal, its strikeout too, as one undo step. + let pair = |a: &mut Automation| { + let list = ok(a, "comment_list", json!({ "doc": doc, "page": 2 })); + let c = list["comments"].as_array().unwrap().iter().find(|c| c["type"] == "Caret").unwrap().clone(); + (c["color"].clone(), c["replies"][0]["color"].clone()) + }; + ok(&mut a, "comment_edit", json!({ "doc": doc, "page": 2, "index": caret["index"], "color": "purple" })); + assert_eq!(pair(&mut a), (json!("#8A3FD1"), json!("#8A3FD1")), "the strikeout follows its caret"); + ok(&mut a, "edit_undo", json!({ "doc": doc })); + assert_eq!(pair(&mut a), (json!("#2E9E5C"), json!("#2E9E5C")), "one undo puts both back"); assert!(matches!(a.call("comment_add", &json!({ "doc": doc, "page": 2, "type": "replace", "find": "page 2" })), Err(ToolError::InvalidArgs(_)))); // A file attached as a comment. std::fs::write(dir.join("notes.txt"), b"remember").unwrap(); diff --git a/crates/engine/src/edit/run.rs b/crates/engine/src/edit/run.rs index 1d4a799397..265104c657 100644 --- a/crates/engine/src/edit/run.rs +++ b/crates/engine/src/edit/run.rs @@ -561,8 +561,8 @@ pub(crate) fn run_edit(doc: &mut pdfcraft_cos::Document, edit: &Edit, cx: &mut E } Edit::DeleteAttachment { key } => pdfcraft_organize::delete_attachment(doc, key)?, Edit::ReplaceText { page, quads, text, author, strike, caret } => { - // One `/NM` each: the caret replies to the strikeout, and XFDF names that reply by - // the strikeout's `/NM` (#2666). + // One `/NM` each: the strikeout is grouped under the caret, and XFDF names that link by + // the caret's `/NM` (#2666). let (strike_meta, caret_meta) = (cx.meta(), cx.meta()); pdfcraft_annot::add_text_replacement( doc, diff --git a/crates/ui-egui/src/panels/comments_panel.rs b/crates/ui-egui/src/panels/comments_panel.rs index 430bdbec85..f6b0043637 100644 --- a/crates/ui-egui/src/panels/comments_panel.rs +++ b/crates/ui-egui/src/panels/comments_panel.rs @@ -40,7 +40,7 @@ pub fn subtype_label(s: &str) -> &str { /// The comment type as the panel names it, telling callouts, clouds and replaced text apart by `/IT`. pub fn kind_label(a: &Annotation) -> &str { match a.intent.as_deref() { - Some("StrikeOutTextEdit") => "Replace text", + Some("StrikeOutTextEdit" | "Replace") => "Replace text", Some("FreeTextCallout") => "Callout", Some("PolygonCloud") => "Cloud", Some("FreeTextTypeWriter") => "Typewriter", diff --git a/crates/ui-egui/src/tools/comments.rs b/crates/ui-egui/src/tools/comments.rs index 34a9b24a09..56301969a2 100644 --- a/crates/ui-egui/src/tools/comments.rs +++ b/crates/ui-egui/src/tools/comments.rs @@ -441,7 +441,7 @@ pub fn tool_for(a: &Annotation) -> Option { ("FreeText", _) => CommentTool::TextBox, ("Highlight", _) => CommentTool::Highlight, ("Underline", _) => CommentTool::Underline, - ("StrikeOut", Some("StrikeOutTextEdit")) => CommentTool::ReplaceText, + ("StrikeOut", Some("StrikeOutTextEdit")) | ("Caret", Some("Replace")) => CommentTool::ReplaceText, ("StrikeOut", _) => CommentTool::StrikeOut, ("Squiggly", _) => CommentTool::Squiggly, ("Ink", Some("FreehandHighlight")) => CommentTool::FreehandHighlight, diff --git a/crates/ui-egui/src/tools/comments/composer.rs b/crates/ui-egui/src/tools/comments/composer.rs index 8c3dee280c..5ae441a0cd 100644 --- a/crates/ui-egui/src/tools/comments/composer.rs +++ b/crates/ui-egui/src/tools/comments/composer.rs @@ -154,7 +154,10 @@ pub(crate) fn composer(ctx: &egui::Context, view: &mut DocView, info: &DocInfo, text, author: prefs.author.clone(), strike: prefs.style(CommentTool::ReplaceText), - caret: prefs.style(CommentTool::Caret), + // One proposal, one look: the caret, the group's primary, takes the Replace text + // tool's style as the strikeout does, so a default saved from either part (Make + // Current Properties Default) applies to the next proposal as a whole. + caret: prefs.style(CommentTool::ReplaceText), }) } ComposerKind::Caret { size } => { diff --git a/crates/ui-egui/src/tools/comments/geometry.rs b/crates/ui-egui/src/tools/comments/geometry.rs index 7e291be4db..594abae86e 100644 --- a/crates/ui-egui/src/tools/comments/geometry.rs +++ b/crates/ui-egui/src/tools/comments/geometry.rs @@ -5,6 +5,13 @@ pub(super) fn is_markup(subtype: &str) -> bool { matches!(subtype, "Highlight" | "Underline" | "StrikeOut" | "Squiggly") } +/// What lies on the text and doesn't move with a drag: text markup, and the caret of a Replace +/// Text proposal (`/IT /Replace`), which is picked by its strikeout too and would otherwise be +/// dragged away from the text it replaces, leaving the strikeout behind. +pub(super) fn on_text(a: &Annotation) -> bool { + is_markup(&a.subtype) || a.subtype == "Caret" && a.intent.as_deref() == Some("Replace") +} + /// Rectangles, ovals, text boxes, stamps and drawings (drawn signatures among them). A callout's /// `/Rect` also holds its leader line, so it only moves. pub(super) fn resizable(a: &Annotation) -> bool { diff --git a/crates/ui-egui/src/tools/comments/input.rs b/crates/ui-egui/src/tools/comments/input.rs index 40c229e535..59a9b40ca2 100644 --- a/crates/ui-egui/src/tools/comments/input.rs +++ b/crates/ui-egui/src/tools/comments/input.rs @@ -1,5 +1,5 @@ use super::geometry::{ - CARET_SIZE, HANDLES, clamp_to, corner_ratio, drawn_box, handle_pos, held, is_markup, point_handle, reaches_points, reshapable, reshaped, + CARET_SIZE, HANDLES, clamp_to, corner_ratio, drawn_box, handle_pos, held, is_markup, on_text, point_handle, reaches_points, reshapable, reshaped, resizable, resized, snapped_caret, }; use super::paint::{closes_at, drawn_shape, edit_anchor, erase, new_comment}; @@ -137,7 +137,7 @@ pub(crate) fn hit(cx: &PageCx<'_>, view: &DocView, p: Pos2, on_link: bool) -> Op if cx.tool == QuickTool::Select && a.subtype == "Highlight" && on_link { return None; } - let markup = is_markup(&a.subtype); + let markup = on_text(a); let icon = if cx.allowed && !markup { egui::CursorIcon::Move } else { egui::CursorIcon::PointingHand }; // Text markup does not move: a click selects it, a drag across it selects the text. Some(Hit { declines_drag: markup, ..Hit::new(Owner::Comment { index: a.index, part: CommentPart::Body }, layer(a), icon) }) @@ -253,7 +253,7 @@ fn select_input( cv.gesture = Some(Gesture::Resize { page: cx.page, index: a.index, handle: (x, y), from: o, aspect_ratio }); consumed = true; } - CommentPart::Body if !is_markup(&a.subtype) => { + CommentPart::Body if !on_text(a) => { // Dragging one of several selected comments moves them all; any other drag starts // a selection of its own. let grabbed = (cx.page, a.index); @@ -341,14 +341,14 @@ fn finish_select_gesture(ui: &egui::Ui, cx: &PageCx<'_>, view: &mut DocView, p: let (a, b) = (cx.to_user(from), cx.to_user(p)); let (dx, dy) = (b[0] - a[0], b[1] - a[1]); if from.distance(p) >= 2.0 { - // The other selected comments on this page (not text markup, which is where - // its text is, nor a locked one, which would refuse the whole step) move with - // the dragged one, as one undo step. + // The other selected comments on this page (not text markup or a Replace Text + // caret, which are where their text is, nor a locked one, which would refuse the + // whole step) move with the dragged one, as one undo step. let others: Vec = view .comments .also .iter() - .filter(|(p, i)| *p == page && *i != index && cx.get(*i).is_some_and(|a| !is_markup(&a.subtype) && !a.locked)) + .filter(|(p, i)| *p == page && *i != index && cx.get(*i).is_some_and(|a| !on_text(a) && !a.locked)) .map(|(_, i)| *i) .collect(); view.pending_edit = Some(if others.is_empty() { @@ -759,7 +759,7 @@ pub(crate) fn nudge_keys(ctx: &egui::Context, view: &mut DocView, info: &DocInfo return; } let Some((page, index)) = cv.selected else { return }; - let movable = info.annotations.iter().any(|a| a.page == page && a.index == index && !a.is_reply() && !is_markup(&a.subtype) && !a.locked); + let movable = info.annotations.iter().any(|a| a.page == page && a.index == index && !a.is_reply() && !on_text(a) && !a.locked); let (Some(xf), Some(p)) = (view.page_xform(page), info.pages.get(page)) else { return }; if !movable { return; diff --git a/crates/ui-egui/src/tools/comments/state.rs b/crates/ui-egui/src/tools/comments/state.rs index cff7b1d8a6..b686c402b1 100644 --- a/crates/ui-egui/src/tools/comments/state.rs +++ b/crates/ui-egui/src/tools/comments/state.rs @@ -450,9 +450,23 @@ impl PageCx<'_> { } /// The topmost comment under `p`: the one drawn last, in `/Annots` order (inspection lists - /// comments by position, not in paint order). + /// comments by position, not in paint order). A Replace Text strikeout is grouped under its + /// caret (`/RT /Group`), which holds the proposal, so the strikeout is hit as that caret: the + /// proposal is picked by either part, not only by its small caret. pub(super) fn comment_at(&self, p: Pos2) -> Option<&Annotation> { - self.comments().filter(|a| self.screen_rects(a).any(|r| r.expand(3.0).contains(p))).max_by_key(|a| a.index) + let under = |a: &&Annotation| self.screen_rects(a).any(|r| r.expand(3.0).contains(p)); + let strikes = self + .info + .annotations + .iter() + .filter(|a| !self.hidden && a.page == self.page && a.grouped && a.intent.as_deref() == Some("StrikeOutTextEdit")) + .filter(under) + .filter_map(|s| { + let (page, index) = s.parent?; + let caret = self.get(index).filter(|c| page == self.page && c.subtype == "Caret")?; + Some((s.index, caret)) + }); + self.comments().filter(under).map(|a| (a.index, a)).chain(strikes).max_by_key(|(i, _)| *i).map(|(_, a)| a) } pub(crate) fn get(&self, index: usize) -> Option<&Annotation> { diff --git a/crates/ui-egui/tests/it/comments.rs b/crates/ui-egui/tests/it/comments.rs index 2308ba9e0f..56984b6882 100644 --- a/crates/ui-egui/tests/it/comments.rs +++ b/crates/ui-egui/tests/it/comments.rs @@ -2595,9 +2595,45 @@ fn highlighting_an_area_off_the_text() { assert_eq!(comments(&h).len(), 2); } +/// A Replace Text proposal is picked by its strikeout too, not only by its small caret: the +/// strikeout is grouped under the caret (`/RT /Group`), and a click on it selects the caret, the +/// proposal, so it can be deleted or restyled from either part. +#[test] +fn clicking_a_replace_text_strikeout_selects_the_proposal() { + let pdf = b"%PDF-1.7 +1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj +2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj +3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 300 200] /Annots [5 0 R 4 0 R] >> endobj +4 0 obj << /Type /Annot /Subtype /Caret /Rect [135 95 145 105] /C [0 0 1] /T (Bob) /IT /Replace /Contents (slow red) >> endobj +5 0 obj << /Type /Annot /Subtype /StrikeOut /Rect [40 100 130 120] /QuadPoints [40 120 130 120 40 100 130 100] /C [1 0 0] /T (Bob) /IT /StrikeOutTextEdit /IRT 4 0 R /RT /Group >> endobj +trailer << /Root 1 0 R >> +%%EOF"; + let mut h = harness(move |app| { + app.open_bytes("replace.pdf", None, pdf.to_vec()).expect("opens"); + app.close_tab(0); + }); + h.run_steps(6); + let caret = comments(&h).into_iter().find(|a| a.subtype == "Caret").expect("the caret"); + let p = at(&h, 70.0, 110.0); + click(&mut h, p); + assert_eq!(h.state().views[0].comments.selected, Some((caret.page, caret.index)), "the strikeout picks the proposal"); + // Review on #5689 (CodeRabbit): a drag from the strikeout, or from the caret, moves neither: + // the proposal lies on the text it replaces, and the caret alone would leave the strikeout. + let caret_at = at(&h, 140.0, 100.0); + for from in [p, caret_at] { + drag(&mut h, from, from + egui::vec2(60.0, 40.0)); + h.run_steps(3); + let now = comments(&h).into_iter().find(|a| a.subtype == "Caret").expect("the caret"); + assert_eq!(now.rect, caret.rect, "the caret stays on the text"); + } + assert_eq!(h.state().session.get(h.state().views[0].id).unwrap().can_undo(), None, "nothing was moved"); +} + #[test] fn replacing_text_strikes_it_and_adds_a_caret() { let mut h = harness(|app| app.set_option("quick", "replace").unwrap()); + // Review on #5689 (CodeRabbit): both parts take the Replace text tool's style. + h.state_mut().comment_prefs.set_color(pdfcraft_ui_egui::comments::CommentTool::ReplaceText, [0.1, 0.6, 0.2]); drag_pt(&mut h, (20.0, 155.0), (120.0, 155.0)); assert!(h.state().views[0].comments.composer.is_some(), "the composer asks for the replacement"); field(&h, "Replacement text").type_text("slow red"); @@ -2608,15 +2644,19 @@ fn replacing_text_strikes_it_and_adds_a_caret() { let strike = c.iter().find(|a| a.subtype == "StrikeOut").expect("a strikeout"); let caret = c.iter().find(|a| a.subtype == "Caret").expect("a caret"); assert_eq!(caret.contents.as_deref(), Some("slow red")); - assert_eq!(caret.in_reply_to, strike.name, "grouped with the strikeout"); + assert_eq!((strike.color, caret.color), (Some([0.1, 0.6, 0.2]), Some([0.1, 0.6, 0.2])), "one style for the proposal"); + assert_eq!(strike.in_reply_to, caret.name, "the strikeout is grouped under the caret, which holds the text"); assert_eq!(h.state().session.get(h.state().views[0].id).unwrap().can_undo(), Some("Replace text")); assert_eq!(h.state().quick_tool, QuickTool::Select); - // The strikeout is the Replace text tool's, not Strikethrough's (#3947): the panel names it - // so, and Make Current Properties Default updates the Replace text tool. + // Both members are the Replace text tool's, not Strikethrough's or Insert text's (#3947), so + // Make Current Properties Default updates the Replace text tool. assert_eq!(pdfcraft_ui_egui::comments::tool_for(strike), Some(pdfcraft_ui_egui::comments::CommentTool::ReplaceText)); + assert_eq!(pdfcraft_ui_egui::comments::tool_for(caret), Some(pdfcraft_ui_egui::comments::CommentTool::ReplaceText)); + // The panel's card is the caret's, the group's primary: it shows the replacement, and + // nothing is named Strikethrough. h.state_mut().set_option("panel", "comments").unwrap(); h.run_steps(3); - assert!(h.query_by_role_and_label(egui::accesskit::Role::Label, "Replace text").is_some(), "the panel names the comment Replace text"); + assert!(h.query_by_role_and_label(egui::accesskit::Role::Label, "slow red").is_some(), "the panel shows the replacement"); assert!(h.query_by_role_and_label(egui::accesskit::Role::Label, "Strikethrough").is_none(), "not Strikethrough"); } diff --git a/crates/xfdf/src/tests.rs b/crates/xfdf/src/tests.rs index 96e9179faa..52ce9e9508 100644 --- a/crates/xfdf/src/tests.rs +++ b/crates/xfdf/src/tests.rs @@ -143,9 +143,9 @@ fn values(doc: &Document) -> Vec<(String, Vec)> { fields(doc).into_iter().map(|f| (f.name, f.value)).collect() } -/// #2666: Replace Text writes a StrikeOut and a Caret that replies to it. XFDF names a reply by -/// the `/NM` of what it replies to, so when both carried one id the caret's `inreplyto` was its -/// own `name`: importing it found a comment replying to itself and the pair was lost. +/// #2666: Replace Text writes a Caret and a StrikeOut grouped under it. XFDF names the link by +/// the `/NM` of the annotation it points at, so when both carried one id the `inreplyto` was the +/// strikeout's own `name`: importing it found a comment replying to itself and the pair was lost. #[test] fn a_replace_text_pair_survives_an_xfdf_round_trip() { let mut src = blank(); @@ -170,8 +170,8 @@ fn a_replace_text_pair_survives_an_xfdf_round_trip() { let xfdf = export_xfdf(&src, true, false, "a.pdf"); assert!(xfdf.contains("name=\"strike\"") && xfdf.contains("name=\"caret\""), "each gets its own name: {xfdf}"); - assert!(xfdf.contains("inreplyto=\"strike\""), "the caret replies to the strikeout: {xfdf}"); - assert!(!xfdf.contains("inreplyto=\"caret\""), "never a reply to itself"); + assert!(xfdf.contains("inreplyto=\"caret\""), "the strikeout is grouped under the caret: {xfdf}"); + assert!(!xfdf.contains("inreplyto=\"strike\""), "never a reply to itself"); let mut dst = blank(); let r = import(&mut dst, xfdf.as_bytes()).unwrap(); @@ -180,6 +180,135 @@ fn a_replace_text_pair_survives_an_xfdf_round_trip() { let kinds: Vec<&str> = found.iter().map(|s| s.subtype.as_str()).collect(); assert!(kinds.contains(&"StrikeOut") && kinds.contains(&"Caret"), "{kinds:?}"); assert_eq!(comment_view(&dst), comment_view(&src), "the pair and its thread survive"); + // Review on #5689 (CodeRabbit): the intents travel too, so the imported pair is still + // Replace Text, and deleting its strikeout takes the caret and its replacement with it. + assert!(xfdf.contains("intent=\"Replace\"") && xfdf.contains("intent=\"StrikeOutTextEdit\""), "{xfdf}"); + let strike = summaries(&dst).into_iter().find(|s| s.subtype == "StrikeOut").expect("the strikeout"); + pdfcraft_annot::delete_annotation(&mut dst, 0, strike.index).unwrap(); + assert!(summaries(&dst).iter().all(|s| s.subtype != "Caret"), "the caret goes with the strikeout"); +} + +/// Review on #5689 (CodeRabbit): XFDF links a reply to its target by the target's `name`, so a +/// group whose primary has no `/NM` lost its `inreplyto` on export and came back as two loose +/// comments. Both layouts: the caret primary (this PR) and the strikeout primary (older files). +#[test] +fn an_unnamed_group_primary_gets_an_export_only_name() { + for caret_primary in [true, false] { + let mut src = blank(); + let quads = [[100.0, 720.0, 200.0, 720.0, 100.0, 706.0, 200.0, 706.0]]; + let style = Style { color: [0.89, 0.13, 0.13], opacity: 1.0, width: 1.0, fill: None, ..Style::default() }; + let unnamed = Meta { date: None, id: String::new() }; + add_text_replacement( + &mut src, + TextReplacement { + page: 0, + quads: &quads, + replacement: "new words", + author: "Ada", + strike: &style, + caret: &style, + strike_meta: &unnamed, + caret_meta: &unnamed, + }, + ) + .unwrap(); + let page = pdfcraft_model::pages(&src)[0].obj; + let annots = src.get(page).as_dict().unwrap().get(b"Annots").and_then(|a| src.resolve(a).as_array().cloned()).unwrap(); + let refs: Vec = annots.iter().filter_map(Object::as_ref).collect(); + let is = |doc: &Document, r: ObjRef, s: &[u8]| doc.get(r).as_dict().and_then(|d| d.name(b"Subtype")) == Some(s); + let strike = *refs.iter().find(|r| is(&src, **r, b"StrikeOut")).unwrap(); + let caret = *refs.iter().find(|r| is(&src, **r, b"Caret")).unwrap(); + let (primary, sub) = if caret_primary { (caret, strike) } else { (strike, caret) }; + if !caret_primary { + src.update_dict(strike, |d| { + d.remove(b"IRT"); + d.remove(b"RT"); + }) + .unwrap(); + src.update_dict(caret, |d| { + d.set(b"IRT".to_vec(), Object::Ref(strike)); + d.set(b"RT".to_vec(), Object::name("Group")); + }) + .unwrap(); + } + // A comment already named what the made-up name would be: the two must not collide. + let clash = format!("pdfcraft-export-{}-{}", primary.num, primary.generation); + let note = NewAnnotation { + page: 0, + shape: Shape::Note { at: [300.0, 300.0], icon: Default::default() }, + style: style.clone(), + contents: "note".into(), + author: "Ada".into(), + }; + add_annotation(&mut src, ¬e, &Meta { date: None, id: clash.clone() }).unwrap(); + + let xfdf = export_xfdf(&src, true, false, "a.pdf"); + let doc = roxmltree::Document::parse(&xfdf).unwrap(); + let el = |t: &str| doc.descendants().find(|n| n.has_tag_name(t)).unwrap_or_else(|| panic!("no <{t}>: {xfdf}")); + let (tag, sub_tag) = if caret_primary { ("caret", "strikeout") } else { ("strikeout", "caret") }; + let name = el(tag).attribute("name").unwrap_or_else(|| panic!("the unnamed primary gets a name ({caret_primary}): {xfdf}")); + assert_eq!(el(sub_tag).attribute("inreplyto"), Some(name), "{xfdf}"); + assert_ne!(name, clash, "a made-up name never repeats a real /NM"); + assert_eq!(doc.descendants().filter(|n| n.attribute("name") == Some(name)).count(), 1, "{xfdf}"); + assert!(src.get(primary).as_dict().unwrap().get(b"NM").is_none(), "the name is for the export only"); + assert!(src.get(sub).as_dict().unwrap().get(b"NM").is_none()); + + let mut dst = blank(); + import(&mut dst, xfdf.as_bytes()).unwrap(); + // The imported subordinate is grouped under the imported primary again. + let dpage = pdfcraft_model::pages(&dst)[0].obj; + let dannots = dst.get(dpage).as_dict().unwrap().get(b"Annots").and_then(|a| dst.resolve(a).as_array().cloned()).unwrap(); + let drefs: Vec = dannots.iter().filter_map(Object::as_ref).collect(); + let (p_sub, s_sub): (&[u8], &[u8]) = if caret_primary { (b"Caret", b"StrikeOut") } else { (b"StrikeOut", b"Caret") }; + let dprimary = *drefs.iter().find(|r| is(&dst, **r, p_sub)).unwrap(); + let dsub = *drefs.iter().find(|r| is(&dst, **r, s_sub)).unwrap(); + let irt = dst.get(dsub).as_dict().unwrap().get(b"IRT").and_then(Object::as_ref); + assert_eq!(irt, Some(dprimary), "the group survives the round trip ({caret_primary})"); + } +} + +/// An annotation written inline in `/Annots` (a direct dictionary, not a reference) is exported +/// under its `/NM` too, so a made-up name must not repeat it either. +#[test] +fn an_export_only_name_never_repeats_an_inline_annotations_nm() { + let mut src = blank(); + let quads = [[100.0, 720.0, 200.0, 720.0, 100.0, 706.0, 200.0, 706.0]]; + let style = Style { color: [0.89, 0.13, 0.13], opacity: 1.0, width: 1.0, fill: None, ..Style::default() }; + let unnamed = Meta { date: None, id: String::new() }; + add_text_replacement( + &mut src, + TextReplacement { + page: 0, + quads: &quads, + replacement: "new words", + author: "Ada", + strike: &style, + caret: &style, + strike_meta: &unnamed, + caret_meta: &unnamed, + }, + ) + .unwrap(); + let page = pdfcraft_model::pages(&src)[0].obj; + let mut annots = src.get(page).as_dict().unwrap().get(b"Annots").and_then(|a| src.resolve(a).as_array().cloned()).unwrap(); + let primary = + annots.iter().filter_map(Object::as_ref).find(|r| src.get(*r).as_dict().and_then(|d| d.name(b"Subtype")) == Some(&b"Caret"[..])).unwrap(); + // An inline note already named what the made-up name would be. + let clash = format!("pdfcraft-export-{}-{}", primary.num, primary.generation); + let mut note = Dict::new(); + note.set(b"Type".to_vec(), Object::name("Annot")); + note.set(b"Subtype".to_vec(), Object::name("Text")); + note.set(b"NM".to_vec(), PdfString::text(&clash)); + note.set(b"Rect".to_vec(), Object::Array(vec![300.into(), 300.into(), 320.into(), 320.into()])); + annots.push(Object::Dict(note)); + src.update_dict(page, |p| p.set(b"Annots".to_vec(), Object::Array(annots))).unwrap(); + + let xfdf = export_xfdf(&src, true, false, "a.pdf"); + let doc = roxmltree::Document::parse(&xfdf).unwrap(); + let caret = doc.descendants().find(|n| n.has_tag_name("caret")).unwrap_or_else(|| panic!("no : {xfdf}")); + let name = caret.attribute("name").unwrap_or_else(|| panic!("the unnamed primary gets a name: {xfdf}")); + assert_ne!(name, clash, "a made-up name never repeats an inline annotation's /NM: {xfdf}"); + assert_eq!(doc.descendants().filter(|n| n.attribute("name") == Some(&clash[..])).count(), 1, "{xfdf}"); } #[test] diff --git a/crates/xfdf/src/xfdf.rs b/crates/xfdf/src/xfdf.rs index e2137a98f0..3368003483 100644 --- a/crates/xfdf/src/xfdf.rs +++ b/crates/xfdf/src/xfdf.rs @@ -99,31 +99,39 @@ fn read_own_keys(node: roxmltree::Node, d: &mut Dict) { } } -/// The name of every annotation, for `inreplyto`: its /NM, or, for one in `only` without one -/// (other tools leave /NM out), a name new to the document for this export, so a reply copied -/// with it still finds it. +/// The name of every annotation, for `inreplyto`: its /NM, or, for one without one (other tools +/// leave /NM out) that is in `only` or that another annotation replies to or is grouped under +/// (`/IRT`, ISO 32000-2 §12.5.6.2; a Replace Text pair is such a group), a name new to the +/// document for this export, so the link survives the round trip. Made-up names are never written +/// into the PDF. fn annotation_names(doc: &Document, pages: &[pdfcraft_model::Page], only: Option<&HashSet<(usize, usize)>>) -> HashMap { let mut names: HashMap = HashMap::new(); let mut unnamed: Vec<(usize, usize, ObjRef)> = Vec::new(); + let mut targets: HashSet = HashSet::new(); // Every name in use, those of annotations written straight into the page array too, so a // given name is new. let mut taken: HashSet = HashSet::new(); for (pi, p) in pages.iter().enumerate() { for (ai, a) in p.dict.get(b"Annots").map(|a| doc.resolve(a)).and_then(|a| a.as_array().cloned()).unwrap_or_default().into_iter().enumerate() { - let nm = doc.resolve(&a).as_dict().and_then(|d| text_of(doc, d, b"NM")); + let d = doc.resolve(&a).as_dict().cloned(); + if let Some(t) = d.as_ref().and_then(|d| d.get(b"IRT")).and_then(Object::as_ref) { + targets.insert(t); + } + let nm = d.as_ref().and_then(|d| text_of(doc, d, b"NM")).filter(|nm| !nm.is_empty()); if let Some(nm) = &nm { taken.insert(nm.clone()); } + // Only an indirect annotation can be an `/IRT` target, so only it is keyed here. let Some(r) = a.as_ref() else { continue }; match nm { Some(nm) => { names.insert(r, nm); } - None if only.is_some_and(|only| only.contains(&(pi, ai))) => unnamed.push((pi, ai, r)), - None => {} + None => unnamed.push((pi, ai, r)), } } } + unnamed.retain(|(pi, ai, r)| only.is_some_and(|only| only.contains(&(*pi, *ai))) || targets.contains(r)); for (pi, ai, r) in unnamed { let mut name = format!("pdfcraft-{pi}-{ai}"); while taken.contains(&name) { @@ -180,8 +188,8 @@ fn xfdf_annot_children(doc: &Document, out: &mut String, d: &Dict, pi: usize) { /// `own`: write each comment's PdfCraft keys too (see [`own_keys`]), for a copy and paste. fn xfdf_annots(doc: &Document, out: &mut String, only: Option<&HashSet<(usize, usize)>>, own: bool) { let pages = pdfcraft_model::pages(doc); - // NM of every annotation, for inreplyto. A selected annotation without one (other tools - // leave /NM out) gets a name for this export, so a reply copied with it still finds it. + // NM of every annotation, for inreplyto. A selected or replied-to annotation without one + // (other tools leave /NM out) gets a name for this export, so its replies still find it. let names = annotation_names(doc, &pages, only); for (pi, p) in pages.iter().enumerate() { for (ai, a) in p.dict.get(b"Annots").map(|a| doc.resolve(a)).and_then(|a| a.as_array().cloned()).unwrap_or_default().into_iter().enumerate() { @@ -203,8 +211,8 @@ fn xfdf_annots(doc: &Document, out: &mut String, only: Option<&HashSet<(usize, u n(rect[1].max(rect[3])) ); } - // The name: /NM, or the one given above for this export. - if let Some(nm) = text_of(doc, &d, b"NM").or_else(|| a.as_ref().and_then(|r| names.get(&r).cloned())) { + // Its `/NM`, or the name given above for this export. + if let Some(nm) = a.as_ref().and_then(|r| names.get(&r)).cloned().or_else(|| text_of(doc, &d, b"NM")) { let _ = write!(attrs, " name=\"{}\"", esc(&nm)); } for (key, attr) in [ diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index b2e651ec57..6405daf739 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -3735,7 +3735,7 @@ status = "shipped" commands = ["comment.replace"] tools = ["comment_add"] evidence = ["crates/annot/src/tests.rs::replace_text_groups_a_strikeout_and_a_caret", "crates/ui-egui/tests/it/comments.rs::replacing_text_strikes_it_and_adds_a_caret", "crates/automation/tests/it/automation.rs::drawing_comments_through_tools"] -notes = "strikeout (/IT StrikeOutTextEdit) and a grouped caret (/RT /Group) with the replacement" +notes = "a caret holding the replacement, the group's primary (/IT /Replace), with the strikeout (/IT /StrikeOutTextEdit) grouped under it (/IRT, /RT /Group), as ISO 32000-2 §12.5.6.2 reads a group's text from its primary; deleting either deletes both. Missing: /RT /Group is not read yet, so the Comments panel and comment_list show the strikeout as a reply of the caret (#4357)." [[feature]] id = "comment.insert-text" From bf409db55c22f125284ea376c9bbf9cb2e47ad69 Mon Sep 17 00:00:00 2001 From: "Hasan N." <42965049+hasannadeem@users.noreply.github.com> Date: Sun, 8 Nov 2026 19:38:40 -0500 Subject: [PATCH 14/75] M3: pages shorter than the window sit in its middle (#5708) (#5759) --- ROADMAP.md | 1 + crates/ui-egui/src/canvas.rs | 35 ++++++++----- crates/ui-egui/src/canvas/tests.rs | 20 ++++++++ crates/ui-egui/tests/it/dialog_chrome.rs | 16 +++--- crates/ui-egui/tests/it/view.rs | 62 ++++++++++++++++++++++++ docs/roadmap.md | 1 + parity/acrobat-features.toml | 4 +- 7 files changed, 118 insertions(+), 21 deletions(-) diff --git a/ROADMAP.md b/ROADMAP.md index 834e61cfd7..2df1bd26ca 100644 --- a/ROADMAP.md +++ b/ROADMAP.md @@ -256,6 +256,7 @@ Ranked; detail and the full milestone table in [docs/roadmap.md](docs/roadmap.md - **2026-11-02 (M10, #5743):** macOS print jobs name their title and owner (`com.apple.print.JobInfo.PMJobName`, `PMJobOwner`), as the system print dialog does, because a Kyocera filter crashes on a job without them. Checked by a test and through `doc_print` against a stand-in `lp`; not yet on a physical printer. +- **2026-11-02 (M3, #5708):** Pages shorter than the window sit in its middle, one page or row at a time and when scrolling continuously with every page in view. - **2026-11-02 (M14, #5737, #5460):** Preferences can turn off the update check at start. It stays on by default, and Help ▸ Check for updates still asks. - **2026-11-01 (M7, #5657):** Header/footer, watermark and background additions and replacements stage the complete call before committing. Synthetic later-page failures preserve exact incremental bytes, object count, dirty state and allocation cursor; success controls retain other marks and shared/inherited resources. Estimates unchanged. diff --git a/crates/ui-egui/src/canvas.rs b/crates/ui-egui/src/canvas.rs index bae59fc980..3e51dac3bf 100644 --- a/crates/ui-egui/src/canvas.rs +++ b/crates/ui-egui/src/canvas.rs @@ -2136,8 +2136,7 @@ fn single_page_paging(ui: &egui::Ui, view: &mut DocView, cx: &PagingCx<'_>) -> ( if !view.continuous { // The row's bounding box on screen (one page, or a two-page spread). let row_rect = row.iter().filter_map(|&i| rects.get(i)).copied().reduce(Rect::union); - // Within a point, so layout rounding can't stop a fitting page from turning. - let fits = row_rect.is_some_and(|r| r.height() + 2.0 * MARGIN <= avail.height() + 1.0); + let fits = row_rect.is_some_and(|r| fits_height(r, avail.height(), ui.ctx().pixels_per_point())); let can_turn = !middle_gesture && unobstructed && ui.rect_contains_pointer(avail); single_paging = can_turn && fits; // Every wheel event goes to the pager, so it follows each trackpad touch to its end @@ -2936,6 +2935,28 @@ fn refine_goto_point(mut scroll: egui::ScrollArea, view: &mut DocView, geometry: scroll } +/// Where the `shown` pages sit in the scroll content, as how far the content starts into the +/// layout and how tall it is. Scrolling continuously every page scrolls; otherwise only the +/// current row does, the content being its bounding box plus margins. Pages that fit a view +/// `view_height` tall sit in its middle, the spare height shared above and below (#5708): the +/// content fills the view, so there is nothing to scroll, and starts on a whole physical pixel +/// as the layout's pages do, or their edges would blur and the seams between them open. +fn vertical_placement(continuous: bool, rects: &[Rect], shown: &[usize], view_height: f32, pixels_per_point: f32) -> (f32, f32) { + let Some(r) = shown.iter().filter_map(|&i| rects.get(i)).copied().reduce(Rect::union) else { return (0.0, 0.0) }; + if fits_height(r, view_height, pixels_per_point) { + return (((r.center().y - view_height / 2.0) * pixels_per_point).round() / pixels_per_point, view_height); + } + if continuous { (0.0, r.bottom() + MARGIN) } else { (r.top() - MARGIN, r.height() + 2.0 * MARGIN) } +} + +/// Whether the pages in `r` show whole in a view `view_height` tall, without their margins. +/// Within half a physical pixel, which is all that snapping the pages to pixels can add: a +/// fitting page is centred and has nothing to scroll, so any more would be cut off unseen. +fn fits_height(r: Rect, view_height: f32, pixels_per_point: f32) -> bool { + let slack = if pixels_per_point.is_finite() && pixels_per_point > 0.0 { 0.5 / pixels_per_point } else { 0.0 }; + r.height() <= view_height + slack +} + /// Draws the document tab at `index`: its notice bars, then its pages (or the Organize grid) in a /// scroll area with this frame's zoom, scrolling, tool input, overlays and render requests. pub fn document_area(app: &mut PdfCraftApp, index: usize, ui: &mut egui::Ui) { @@ -3057,15 +3078,7 @@ pub fn document_area(app: &mut PdfCraftApp, index: usize, ui: &mut egui::Ui) { // row, the jump to the new page scrolled both out of view, a blank frame on every turn forward. let row = shown_row(view, &rects); let visible_pages: Vec = if view.continuous { (0..rects.len()).collect() } else { row.clone() }; - let (y_shift, content_h) = if view.continuous { - (0.0, rects.iter().map(|r| r.bottom()).reduce(f32::max).map(|bottom| bottom + MARGIN).unwrap_or(0.0)) - } else { - // Only the current row scrolls into view; the content is its bounding box plus margins. - match row.iter().filter_map(|&i| rects.get(i)).copied().reduce(Rect::union) { - Some(r) => (r.top() - MARGIN, r.height() + 2.0 * MARGIN), - None => (0.0, 0.0), - } - }; + let (y_shift, content_h) = vertical_placement(view.continuous, &rects, &visible_pages, avail.height(), ui.ctx().pixels_per_point()); let mut scroll = page_scroll_area( ui, diff --git a/crates/ui-egui/src/canvas/tests.rs b/crates/ui-egui/src/canvas/tests.rs index 013d643990..62c00d6b7a 100644 --- a/crates/ui-egui/src/canvas/tests.rs +++ b/crates/ui-egui/src/canvas/tests.rs @@ -2560,3 +2560,23 @@ fn enhancing_a_scan_re_renders_the_page_shown() { assert!(app.scan_clean_run.is_none()); assert_eq!(app.views[i].pages.get(&0).unwrap().tag, STALE_TAG, "the old raster is re-rendered"); } + +/// #5708 review: a page that fits is centred with nothing to scroll, so only pixel snapping may +/// pass for fitting. A page half a pixel or more too tall scrolls, or its ends would be cut off. +#[test] +fn only_pixel_snapping_passes_for_fitting_the_view() { + let page = |height: f32| [Rect::from_min_size(pos2(0.0, MARGIN), vec2(300.0, height))]; + for ppp in [1.0, 2.0] { + let slack = 0.5 / ppp; + let (_, content) = vertical_placement(false, &page(600.0 + slack), &[0], 600.0, ppp); + assert_eq!(content, 600.0, "at {ppp}x, snapping's half pixel still fits"); + for over in [slack + 0.1, 1.0] { + let (shift, content) = vertical_placement(false, &page(600.0 + over), &[0], 600.0, ppp); + assert_eq!((shift, content), (0.0, 600.0 + over + 2.0 * MARGIN), "at {ppp}x, {over} pt too tall scrolls"); + } + } + // A scale that isn't one has no slack and never panics. + for ppp in [0.0, f32::NAN, f32::INFINITY] { + assert!(!fits_height(Rect::from_min_size(pos2(0.0, 0.0), vec2(1.0, 600.5)), 600.0, ppp)); + } +} diff --git a/crates/ui-egui/tests/it/dialog_chrome.rs b/crates/ui-egui/tests/it/dialog_chrome.rs index 47462fcf2b..c495e608ff 100644 --- a/crates/ui-egui/tests/it/dialog_chrome.rs +++ b/crates/ui-egui/tests/it/dialog_chrome.rs @@ -236,25 +236,25 @@ fn page_hover_cursors_stay_off_while_a_dialog_is_open() { let mut app = PdfCraftApp::new(); app.set_option("language", "en").unwrap(); app.open_bytes("form.pdf", None, include_bytes!("../data/form.pdf").to_vec()).unwrap(); - // Small enough that the check box sits above the dialog card. - app.set_option("zoom", "100").unwrap(); + // Large enough that the drop-down sits below the dialog card. + app.set_option("zoom", "300").unwrap(); app }); for _ in 0..30 { h.run_steps(2); } - // The middle of the "agree" check box (/Rect [20 310 36 326] on a 300 × 400 page). + // The middle of the "country" drop-down (/Rect [20 220 200 240] on a 300 × 400 page). let page = h.state().views[0].page_screen_rect(0).expect("on screen"); let k = page.width() / 300.0; - let check = page.min + egui::vec2(28.0 * k, (400.0 - 318.0) * k); - h.hover_at(check); + let field = page.min + egui::vec2(110.0 * k, (400.0 - 230.0) * k); + h.hover_at(field); h.run_steps(2); - assert_eq!(h.output().platform_output.cursor_icon, egui::CursorIcon::PointingHand, "the check box takes clicks"); + assert_eq!(h.output().platform_output.cursor_icon, egui::CursorIcon::PointingHand, "the drop-down takes clicks"); h.state_mut().set_option("dialog", "security").unwrap(); h.run_steps(4); let card = card(&h, "dialog"); - assert!(!card.contains(check), "the check box {check:?} is beside the dialog {card:?}, under its backdrop"); - h.hover_at(check + egui::vec2(1.0, 0.0)); + assert!(!card.contains(field), "the drop-down {field:?} is beside the dialog {card:?}, under its backdrop"); + h.hover_at(field + egui::vec2(1.0, 0.0)); h.run_steps(2); assert_eq!(h.output().platform_output.cursor_icon, egui::CursorIcon::Default, "nothing under the dialog's backdrop can be clicked"); } diff --git a/crates/ui-egui/tests/it/view.rs b/crates/ui-egui/tests/it/view.rs index 59503bba30..f6384127e8 100644 --- a/crates/ui-egui/tests/it/view.rs +++ b/crates/ui-egui/tests/it/view.rs @@ -1345,6 +1345,68 @@ fn single_page_fit_page_fits_the_shown_page() { assert_ne!(h.state().views[0].zoom, z1, "the wider page refits instead of keeping page 1's zoom"); } +#[test] +fn pages_shorter_than_the_window_sit_in_its_middle() { + let centred = |h: &Harness<'static, PdfCraftApp>, r: egui::Rect| { + let vp = h.state().views[0].viewport_rect(); + assert!(r.height() < vp.height(), "{r:?} is shorter than {vp:?}"); + assert!((r.center().y - vp.center().y).abs() < 1.0, "{r:?} is in the middle of {vp:?}"); + }; + let mut h = harness(&[("layout", "single"), ("continuous", "off"), ("zoom", "50")]); + h.run_steps(4); + centred(&h, rect(&h, 0).expect("page 1")); + assert_eq!(h.state().views[0].current, 0); + + // The page drawn there is the page a click finds: its link still goes to page 3. + let r = rect(&h, 0).expect("page 1"); + let at = egui::pos2(r.min.x + r.width() * 0.5, r.min.y + r.height() * (1.0 - 325.0 / 400.0)); + h.hover_at(at); + h.run_steps(2); + h.drag_at(at); // press + h.drop_at(at); // release + h.run_steps(8); + assert_eq!(h.state().views[0].current, 2, "the link is where the page is drawn"); + + // Scrolling continuously, the column of pages as a whole. + let mut h = harness(&[("layout", "single"), ("continuous", "on"), ("zoom", "25")]); + h.run_steps(4); + let column = rect(&h, 0).expect("page 1").union(rect(&h, 2).expect("page 3")); + centred(&h, column); + + // A page taller than the view starts at its top, as before. + let mut h = harness(&[("layout", "single"), ("continuous", "off"), ("zoom", "250")]); + h.run_steps(4); + let (r, vp) = (rect(&h, 0).expect("page 1"), h.state().views[0].viewport_rect()); + assert!(r.height() > vp.height() && (r.min.y - vp.min.y).abs() < 40.0, "{r:?} in {vp:?}"); +} + +/// A page too tall for its margins still fits the window: it shows whole, in the middle. +#[test] +fn a_page_that_fits_without_its_margins_shows_whole() { + let mut h = harness(&[("layout", "single"), ("continuous", "off"), ("zoom", "50")]); + h.run_steps(4); + let (half, vp) = (rect(&h, 0).expect("page 1"), h.state().views[0].viewport_rect()); + let spare = 20.0; + let zoom = (50.0 * (vp.height() - spare) / half.height()).floor(); + h.state_mut().set_option("zoom", &zoom.to_string()).unwrap(); + h.run_steps(4); + let r = rect(&h, 0).expect("page 1"); + assert!(r.height() < vp.height() && r.height() + 2.0 * 28.0 > vp.height(), "{r:?} fits {vp:?} only without its margins"); + assert!(r.min.y >= vp.min.y && r.max.y <= vp.max.y, "{r:?} shows whole in {vp:?}"); + assert!((r.center().y - vp.center().y).abs() < 1.0, "{r:?} is in the middle of {vp:?}"); +} + +/// Two pages side by side are placed as the row they make: the taller decides, not the first. +#[test] +fn a_row_too_tall_for_the_window_starts_at_its_top_whichever_page_is_shorter() { + let h = mixed_harness(&[("layout", "two-page"), ("zoom", "180")]); + let vp = h.state().views[0].viewport_rect(); + let (first, second) = (rect(&h, 0).expect("page 1"), rect(&h, 1).expect("page 2")); + assert!(first.height() > vp.height() && second.height() < vp.height(), "{first:?}, {second:?} in {vp:?}"); + let row = first.union(second); + assert!((row.min.y - vp.min.y).abs() < 40.0, "{row:?} starts in {vp:?}"); +} + #[test] fn clicking_a_link_goes_to_its_destination() { // Default fit-width zoom: pages are taller than the window, so the target can scroll to the top. diff --git a/docs/roadmap.md b/docs/roadmap.md index 7ce026b3a2..ac9bcbc9e1 100644 --- a/docs/roadmap.md +++ b/docs/roadmap.md @@ -324,6 +324,7 @@ Estimates unchanged. | 2026-11-03 | patch | M5 #5850 review: indirect typewriter intent and composer edits preserve outer whitespace; movement regression commits by outside click; estimates unchanged. | | 2026-11-03 | patch | M5 #5850: Enter adds Fill & Sign line breaks; outside clicks commit, Escape cancels, and saved multiline appearances retain their text and height; estimates unchanged. | | 2026-11-02 | patch | M10 #5743: macOS print jobs carry the job name and owner some vendors' filters need; not yet checked on a physical printer | +| 2026-11-02 | patch | M3 #5708: pages shorter than the window sit in its middle, on a whole physical pixel | | 2026-11-02 | patch | M14 #5737, #5460: Preferences turns off the update check at start, on by default | | 2026-11-01 | patch | M7 #5657: header/footer, watermark and background operations commit all selected pages together; later-page errors preserve marks, content, shared/inherited resources and pending edits; estimates unchanged. | | 2026-10-31 | patch | #1577: reconcile optional page shadows with current viewer/organizer and settings modules; CI validation pending. | diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index 6405daf739..e0c7adf36a 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -878,8 +878,8 @@ milestone = "M3" status = "shipped" commands = ["view.layout.single"] tools = [] -evidence = ["crates/ui-egui/tests/it/view.rs::single_page_layout_shows_one_page_at_a_time", "crates/ui-egui/tests/it/view.rs::single_page_wheel_pans_instead_when_zoomed_in", "crates/ui-egui/tests/it/view.rs::wheel_turns_one_page_per_notch_and_per_trackpad_swipe", "crates/ui-egui/tests/it/view.rs::single_page_wheel_at_the_bottom_edge_turns_to_the_next_page_top_and_back", "crates/ui-egui/tests/it/view.rs::single_page_swipe_that_reaches_the_edge_does_not_turn_but_a_fresh_one_does", "crates/ui-egui/tests/it/view.rs::rail_page_display_button_switches_layouts", "crates/ui-egui/tests/it/view.rs::view_menu_page_display_uses_the_same_commands"] -notes = "when the whole page fits, each wheel notch turns a page and each trackpad swipe turns one, momentum included; zoomed in so the page is taller than the window, the wheel pans, and at the bottom (top) edge a wheel notch, or a fresh trackpad swipe, shows the next page at its top (the previous page at its bottom); a swipe that panned to the edge, momentum included, does not turn until the fingers lift and swipe again (#1924). Not verified against Acrobat's own thresholds. Switched from the View menu, the rail's Page display button or the palette" +evidence = ["crates/ui-egui/tests/it/view.rs::pages_shorter_than_the_window_sit_in_its_middle", "crates/ui-egui/tests/it/view.rs::a_page_that_fits_without_its_margins_shows_whole", "crates/ui-egui/tests/it/view.rs::a_row_too_tall_for_the_window_starts_at_its_top_whichever_page_is_shorter", "crates/ui-egui/tests/it/view.rs::single_page_layout_shows_one_page_at_a_time", "crates/ui-egui/tests/it/view.rs::single_page_wheel_pans_instead_when_zoomed_in", "crates/ui-egui/tests/it/view.rs::wheel_turns_one_page_per_notch_and_per_trackpad_swipe", "crates/ui-egui/tests/it/view.rs::single_page_wheel_at_the_bottom_edge_turns_to_the_next_page_top_and_back", "crates/ui-egui/tests/it/view.rs::single_page_swipe_that_reaches_the_edge_does_not_turn_but_a_fresh_one_does", "crates/ui-egui/tests/it/view.rs::rail_page_display_button_switches_layouts", "crates/ui-egui/tests/it/view.rs::view_menu_page_display_uses_the_same_commands"] +notes = "a page or row shorter than the window sits in its middle, the spare height shared above and below, continuous scrolling included when every page fits (#5708); when the whole page fits, each wheel notch turns a page and each trackpad swipe turns one, momentum included; zoomed in so the page is taller than the window, the wheel pans, and at the bottom (top) edge a wheel notch, or a fresh trackpad swipe, shows the next page at its top (the previous page at its bottom); a swipe that panned to the edge, momentum included, does not turn until the fingers lift and swipe again (#1924). Not verified against Acrobat's own thresholds. Switched from the View menu, the rail's Page display button or the palette" [[feature]] id = "view.layout-two-up" From 607f79e5e078f533c814871072179ada0fb5dccb Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:39:01 +0800 Subject: [PATCH 15/75] M3: a protected file named before others at launch keeps its password prompt (#5756) (#5762) --- crates/ui-egui/src/app.rs | 6 +- crates/ui-egui/src/app/files.rs | 107 +++++++++++-- crates/ui-egui/src/control/host.rs | 11 +- crates/ui-egui/src/lib.rs | 18 +++ crates/ui-egui/src/platform/open_worker.rs | 2 +- .../ui-egui/src/platform/open_worker/tests.rs | 17 ++ crates/ui-egui/src/platform/pickers.rs | 1 + crates/ui-egui/tests/it/editing.rs | 147 ++++++++++++++++++ 8 files changed, 291 insertions(+), 18 deletions(-) diff --git a/crates/ui-egui/src/app.rs b/crates/ui-egui/src/app.rs index 91bad575c4..e37b199df8 100644 --- a/crates/ui-egui/src/app.rs +++ b/crates/ui-egui/src/app.rs @@ -144,6 +144,10 @@ impl PdfCraftApp { #[cfg(not(target_arch = "wasm32"))] pending_opens: Vec::new(), #[cfg(not(target_arch = "wasm32"))] + deferred_opens: Default::default(), + #[cfg(not(target_arch = "wasm32"))] + password_retry: None, + #[cfg(not(target_arch = "wasm32"))] next_open_token: 0, #[cfg(not(target_arch = "wasm32"))] launch_layout: None, @@ -537,7 +541,7 @@ impl eframe::App for PdfCraftApp { // Don't replace a pending prompt or import. A startup file may need its own password // prompt, so its bytes stay queued until the current interaction finishes. #[cfg(not(target_arch = "wasm32"))] - let opening = !self.pending_opens.is_empty(); + let opening = !self.pending_opens.is_empty() || !self.deferred_opens.is_empty(); #[cfg(target_arch = "wasm32")] let opening = false; if !opening diff --git a/crates/ui-egui/src/app/files.rs b/crates/ui-egui/src/app/files.rs index 76dbfc3921..c8ac133a00 100644 --- a/crates/ui-egui/src/app/files.rs +++ b/crates/ui-egui/src/app/files.rs @@ -1,6 +1,17 @@ use super::*; use crate::file_filter::AddFilterAnyCase as _; +/// Whether an open's reply asks for a password: the file is encrypted and none, or a wrong one, +/// was given. +#[cfg(not(target_arch = "wasm32"))] +fn asks_for_password(kind: &open_worker::ResultKind) -> bool { + matches!( + kind, + open_worker::ResultKind::Prepared { result, .. } + if matches!(**result, Err(pdfcraft_render::OpenError::NeedsPassword | pdfcraft_render::OpenError::WrongPassword)) + ) +} + /// Whether `a` and `b` name the same file: equal, or equal once made absolute against the /// current directory (links are not resolved, as in `last_session`). pub(crate) fn same_file(a: &str, b: &str) -> bool { @@ -122,16 +133,36 @@ impl PdfCraftApp { recovered: request.recovered, then: request.then, display_override: None, + ready: None, }); Ok(token) } + /// Hand waiting files to the worker, oldest first, while its queue has room. Every queued + /// job has a `pending_opens` entry, so fewer entries than the bound means room. + #[cfg(not(target_arch = "wasm32"))] + fn submit_deferred_opens(&mut self) { + let mut queue = std::mem::take(&mut self.deferred_opens); + while self.deferred_opens.is_empty() && self.pending_opens.len() < open_worker::MAX_QUEUED_OPENS { + let Some(next) = queue.pop_front() else { break }; + self.open_path_request(&next.path, next.then, next.display_override); + } + // A file put back (the queue filled again) stays ahead of the ones behind it. + self.deferred_opens.append(&mut queue); + } + #[cfg(not(target_arch = "wasm32"))] pub(crate) fn poll_pending_opens(&mut self) { let pending_opens = std::mem::take(&mut self.pending_opens); + // A retry that is no longer waiting holds nothing back. + if let Some(token) = self.password_retry + && !pending_opens.iter().any(|pending| pending.token == token) + { + self.password_retry = None; + } let mut remaining = pending_opens.into_iter(); let mut waiting = Vec::new(); - while let Some(pending) = remaining.next() { + while let Some(mut pending) = remaining.next() { // A tool, password prompt or file picker belongs to the document just admitted. // Keep later replies queued so another open cannot change its recipient. if self.modal_open() || self.pickers.pending() { @@ -139,19 +170,41 @@ impl PdfCraftApp { waiting.extend(remaining); break; } - match pending.receiver.try_recv() { - Ok(result) if result.token == pending.token => self.finish_pending_open(pending, result.kind), - Ok(_) => waiting.push(pending), - Err(std::sync::mpsc::TryRecvError::Empty) => waiting.push(pending), - Err(std::sync::mpsc::TryRecvError::Disconnected) => { - self.notify_fmt("Couldn't open {name}: the background worker stopped", &[("name", &pending.name)]); - } + let retry = self.password_retry == Some(pending.token); + let kind = match pending.ready.take() { + Some(kind) => kind, + None => match pending.receiver.try_recv() { + Ok(result) if result.token == pending.token => result.kind, + Ok(_) | Err(std::sync::mpsc::TryRecvError::Empty) => { + waiting.push(pending); + continue; + } + Err(std::sync::mpsc::TryRecvError::Disconnected) => { + if retry { + self.password_retry = None; + } + self.notify_fmt("Couldn't open {name}: the background worker stopped", &[("name", &pending.name)]); + continue; + } + }, + }; + // While a closed prompt's password is being tried, a later file that asks for its own + // password keeps its reply: its prompt would hide the retry's outcome. Others admit. + if !retry && self.password_retry.is_some() && asks_for_password(&kind) { + pending.ready = Some(kind); + waiting.push(pending); + continue; } + if retry { + self.password_retry = None; + } + self.finish_pending_open(pending, kind); } // A completion may schedule another open; preserve those requests as well as the // earlier waiting replies, without consuming a blocked receiver's result. waiting.append(&mut self.pending_opens); self.pending_opens = waiting; + self.submit_deferred_opens(); self.settle_launch_layout(); } @@ -452,6 +505,9 @@ impl PdfCraftApp { && let Some(pending) = self.pending_opens.iter_mut().find(|pending| pending.token == *token) { pending.display_override = p.display_override; + // Replies behind it that ask for a password wait for its outcome + // (`poll_pending_opens`). + self.password_retry = Some(*token); } if let Err(error) = result { let mut prompt = p; @@ -594,10 +650,24 @@ impl PdfCraftApp { } return OpenOutcome::Pending; } - if display_override.is_some() - && let Some(prompt) = self.password_prompt.as_mut().filter(|p| p.path.as_deref().is_some_and(|p| same_file(p, path))) - { - prompt.display_override = display_override; + // The same file named again while its password is being asked for joins that prompt + // instead of queuing a second one; a display override or a tool it brings is kept. + if let Some(prompt) = self.password_prompt.as_mut().filter(|p| p.path.as_deref().is_some_and(|p| same_file(p, path))) { + if display_override.is_some() { + prompt.display_override = display_override; + } + if then.is_some() { + prompt.then = then; + } + return OpenOutcome::Pending; + } + if let Some(deferred) = self.deferred_opens.iter_mut().find(|d| same_file(&d.path, path)) { + if display_override.is_some() { + deferred.display_override = display_override; + } + if then.is_some() { + deferred.then = then; + } return OpenOutcome::Pending; } let name = std::path::Path::new(path).file_name().map(|n| n.to_string_lossy().into_owned()).unwrap_or_else(|| path.to_string()); @@ -606,8 +676,10 @@ impl PdfCraftApp { // Before the first frame there is no interaction to keep responsive; this also preserves // deterministic inline startup and test flows. Once live, reading and preparation use // the bounded worker and commit on the UI thread. - // Keep flagged launch files behind an existing password prompt or picker. - if (self.ctx.is_none() || self.run_inline) && !(display_override.is_some() && (self.modal_open() || self.pickers.pending())) { + // Keep later files behind an open password prompt: it belongs to an earlier file, and an + // inline open would clear it. Keep flagged launch files behind any dialog or picker too. + let held = self.password_prompt.is_some() || (display_override.is_some() && (self.modal_open() || self.pickers.pending())); + if (self.ctx.is_none() || self.run_inline) && !held { match std::fs::read(path) { Ok(bytes) => { let before = self.views.len(); @@ -649,6 +721,12 @@ impl PdfCraftApp { } } else { self.startup_superseded = true; + // The worker queue is bounded. A long file list (a launch behind a password prompt, + // or many files at once) waits here, in order, and is submitted as opens finish. + if self.pending_opens.len() >= open_worker::MAX_QUEUED_OPENS || !self.deferred_opens.is_empty() { + self.deferred_opens.push_back(DeferredOpen { path: path.to_string(), then, display_override }); + return OpenOutcome::Pending; + } match self.schedule_open(OpenRequest { name: name.clone(), path: Some(path.to_string()), @@ -686,6 +764,7 @@ impl PdfCraftApp { #[cfg(not(target_arch = "wasm32"))] fn settle_launch_layout(&mut self) { if self.pending_opens.iter().any(|open| open.display_override.is_some()) + || self.deferred_opens.iter().any(|open| open.display_override.is_some()) || self.password_prompt.as_ref().is_some_and(|prompt| prompt.display_override.is_some()) { return; diff --git a/crates/ui-egui/src/control/host.rs b/crates/ui-egui/src/control/host.rs index c2e82f98e5..12c98007cc 100644 --- a/crates/ui-egui/src/control/host.rs +++ b/crates/ui-egui/src/control/host.rs @@ -67,8 +67,15 @@ impl Host for crate::PdfCraftApp { }) .collect(); #[cfg(not(target_arch = "wasm32"))] - let pending_opens: Vec = - self.pending_opens.iter().map(|pending| json!({ "token": pending.token, "name": pending.name, "path": pending.path })).collect(); + let pending_opens: Vec = self + .pending_opens + .iter() + .map(|pending| json!({ "token": pending.token, "name": pending.name, "path": pending.path })) + .chain(self.deferred_opens.iter().map(|deferred| { + let name = std::path::Path::new(&deferred.path).file_name().map(|n| n.to_string_lossy().into_owned()); + json!({ "token": null, "name": name, "path": deferred.path }) + })) + .collect(); #[cfg(target_arch = "wasm32")] let pending_opens: Vec = Vec::new(); json!({ diff --git a/crates/ui-egui/src/lib.rs b/crates/ui-egui/src/lib.rs index 1663c3be73..23625cb97b 100644 --- a/crates/ui-egui/src/lib.rs +++ b/crates/ui-egui/src/lib.rs @@ -262,6 +262,17 @@ struct PendingOpen { recovered: Option, then: Option<&'static str>, display_override: Option<(canvas::PageLayout, bool)>, + /// A reply taken from `receiver` and kept back for later (a password prompt waiting for an + /// earlier file's password retry). + ready: Option, +} + +/// A file named while the worker's queue was full: submitted once it has room. +#[cfg(not(target_arch = "wasm32"))] +struct DeferredOpen { + path: String, + then: Option<&'static str>, + display_override: Option<(canvas::PageLayout, bool)>, } /// Puts `(html, plain_text)` on a clipboard; `false` when it couldn't, and plain-text Copy is @@ -419,6 +430,13 @@ pub struct PdfCraftApp { open_worker: Option, #[cfg(not(target_arch = "wasm32"))] pending_opens: Vec, + /// Files waiting for room in the worker's bounded queue, in the order they were named. + #[cfg(not(target_arch = "wasm32"))] + deferred_opens: std::collections::VecDeque, + /// The open retrying a closed password prompt's file with its password: replies behind it + /// wait for it. + #[cfg(not(target_arch = "wasm32"))] + password_retry: Option, #[cfg(not(target_arch = "wasm32"))] next_open_token: u64, /// Explicit CLI layout waits for flagged launch opens, never targeting restored tabs. diff --git a/crates/ui-egui/src/platform/open_worker.rs b/crates/ui-egui/src/platform/open_worker.rs index fe564db398..274071ea78 100644 --- a/crates/ui-egui/src/platform/open_worker.rs +++ b/crates/ui-egui/src/platform/open_worker.rs @@ -9,7 +9,7 @@ use crate::RecoveryStore; use pdfcraft_engine::{OpenContext, PreparedDocument, SourceKind}; use pdfcraft_render::OpenError; -const MAX_QUEUED_OPENS: usize = 16; +pub(crate) const MAX_QUEUED_OPENS: usize = 16; pub(crate) enum Input { Path(PathBuf), diff --git a/crates/ui-egui/src/platform/open_worker/tests.rs b/crates/ui-egui/src/platform/open_worker/tests.rs index c0fe2f03a2..efe911bff6 100644 --- a/crates/ui-egui/src/platform/open_worker/tests.rs +++ b/crates/ui-egui/src/platform/open_worker/tests.rs @@ -763,6 +763,7 @@ fn launch_layout_waits_for_cli_opens_and_preserves_restored_tabs() { recovered: None, then: None, display_override: Some((PageLayout::TwoUp, false)), + ready: None, }); replies.push(( sender, @@ -820,6 +821,7 @@ fn launch_layout_failed_or_cancelled_opens_do_not_target_existing_documents() { recovered: None, then: None, display_override: Some((PageLayout::TwoUp, false)), + ready: None, }); app.set_launch_layout("single").unwrap(); assert!(app.set_launch_layout("invalid").is_err()); @@ -904,3 +906,18 @@ fn a_new_file_from_the_os_brings_the_window_forward_before_it_is_prepared() { assert_eq!(doc.info.pages.len(), 2); let _ = std::fs::remove_file(path); } + +/// A file waiting for room in the open queue, named again with a tool (Open and run), keeps +/// that tool; a later repeat without one doesn't drop it. +#[test] +fn a_repeated_waiting_open_keeps_the_tool_it_brings() { + let file = tool_open_file("deferred-tool", "pdf", &fixture(1)); + let path = file.0.to_string_lossy().into_owned(); + let mut app = PdfCraftApp::new(); + app.deferred_opens.push_back(crate::DeferredOpen { path: path.clone(), then: None, display_override: None }); + assert_eq!(app.open_path_then(&path, Some("export.text")), crate::OpenOutcome::Pending); + assert_eq!(app.deferred_opens.len(), 1); + assert_eq!(app.deferred_opens.front().and_then(|d| d.then), Some("export.text")); + assert_eq!(app.open_path(&path), crate::OpenOutcome::Pending); + assert_eq!(app.deferred_opens.front().and_then(|d| d.then), Some("export.text")); +} diff --git a/crates/ui-egui/src/platform/pickers.rs b/crates/ui-egui/src/platform/pickers.rs index 80d44d975b..c680eaeaba 100644 --- a/crates/ui-egui/src/platform/pickers.rs +++ b/crates/ui-egui/src/platform/pickers.rs @@ -377,6 +377,7 @@ mod tests { recovered: None, then: None, display_override: None, + ready: None, }); let got = Arc::new(Mutex::new(Vec::new())); app.pickers.showing.store(true, Ordering::SeqCst); diff --git a/crates/ui-egui/tests/it/editing.rs b/crates/ui-egui/tests/it/editing.rs index 4f3e15bd5d..500d504403 100644 --- a/crates/ui-egui/tests/it/editing.rs +++ b/crates/ui-egui/tests/it/editing.rs @@ -2727,6 +2727,153 @@ fn canceling_launch_display_override_does_not_affect_later_documents() { std::fs::remove_file(path).unwrap(); } +/// `pdfcraft locked.pdf other.pdf`: the files open before the first frame, and opening the +/// second one inline used to clear the first one's password prompt, so `locked.pdf` was never +/// opened or asked for. The later file now waits behind the prompt. +#[test] +fn a_startup_file_after_a_protected_one_keeps_its_password_prompt() { + for unlock in [true, false] { + let nonce = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH).unwrap().as_nanos(); + let locked = temp_path(&format!("startup-locked-{unlock}-{nonce}.pdf")); + let other = temp_path(&format!("startup-other-{unlock}-{nonce}.pdf")); + std::fs::write(&locked, protected("reader", "owner", -4)).unwrap(); + std::fs::write(&other, fixture(2)).unwrap(); + let (locked_path, other_path) = (locked.to_str().unwrap().to_string(), other.to_str().unwrap().to_string()); + let (first, second) = (locked_path.clone(), other_path.clone()); + let mut h = Harness::builder().with_size(egui::vec2(1400.0, 900.0)).build_eframe(move |_| { + let mut app = PdfCraftApp::new(); + app.set_option("language", "en").unwrap(); + app.open_path(&first); + app.open_path(&second); + app + }); + h.run_steps(4); + let prompt = h.state().password_prompt.as_ref().map(|p| p.path.clone()); + assert_eq!(prompt, Some(Some(locked_path.clone())), "the protected file is asked for"); + assert!(h.state().views.is_empty(), "the later file waits behind the prompt"); + if unlock { + h.state_mut().submit_password(Some("reader".into())); + support::wait_until(&mut h, "both files", |app| app.views.len() == 2); + // The other file opened while the password was being typed; the unlocked one comes + // after it and is the one in view. + assert_eq!(open_paths(h.state()), vec![other_path, locked_path]); + assert_eq!(h.state().active_ids().map(|(i, _)| i), Some(1)); + } else { + h.state_mut().submit_password(None); + support::wait_until(&mut h, "the other file", |app| app.views.len() == 1); + assert_eq!(open_paths(h.state()), vec![other_path]); + } + std::fs::remove_file(locked).unwrap(); + std::fs::remove_file(other).unwrap(); + } +} + +/// `pdfcraft locked.pdf locked.pdf other.pdf`: the repeated file joins its open password +/// prompt. It used to be queued again behind the prompt, so the unlocked file asked for its +/// password a second time. +#[test] +fn a_protected_file_named_twice_at_startup_asks_for_its_password_once() { + let nonce = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH).unwrap().as_nanos(); + let locked = temp_path(&format!("startup-twice-locked-{nonce}.pdf")); + let other = temp_path(&format!("startup-twice-other-{nonce}.pdf")); + std::fs::write(&locked, protected("reader", "owner", -4)).unwrap(); + std::fs::write(&other, fixture(2)).unwrap(); + let (locked_path, other_path) = (locked.to_str().unwrap().to_string(), other.to_str().unwrap().to_string()); + let (first, second) = (locked_path.clone(), other_path.clone()); + let mut h = Harness::builder().with_size(egui::vec2(1400.0, 900.0)).build_eframe(move |_| { + let mut app = PdfCraftApp::new(); + app.set_option("language", "en").unwrap(); + app.open_path(&first); + app.open_path(&first); + app.open_path(&second); + app + }); + h.run_steps(4); + assert!(h.state().password_prompt.is_some(), "the protected file is asked for"); + h.state_mut().submit_password(Some("reader".into())); + support::wait_until(&mut h, "both files, or a second prompt", |app| app.views.len() == 2 || app.password_prompt.is_some()); + assert!(h.state().password_prompt.is_none(), "the password is asked for once"); + let mut paths = open_paths(h.state()); + paths.sort(); + let mut expected = vec![locked_path, other_path]; + expected.sort(); + assert_eq!(paths, expected); + std::fs::remove_file(locked).unwrap(); + std::fs::remove_file(other).unwrap(); +} + +/// `pdfcraft locked-a.pdf locked-b.pdf`: B's reply is ready while A's password is asked for. +/// Once A's password is given, A's retry comes first: A opens (or asks again) before B's prompt +/// can take the dialog. B's prompt used to replace A's retry, so A waited for B's answer. +#[test] +fn a_password_retry_finishes_before_the_next_protected_files_prompt() { + let nonce = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH).unwrap().as_nanos(); + let a = temp_path(&format!("startup-retry-a-{nonce}.pdf")); + let b = temp_path(&format!("startup-retry-b-{nonce}.pdf")); + std::fs::write(&a, protected("reader", "owner", -4)).unwrap(); + std::fs::write(&b, protected("reader", "owner", -4)).unwrap(); + let (a_path, b_path) = (a.to_str().unwrap().to_string(), b.to_str().unwrap().to_string()); + let (first, second) = (a_path.clone(), b_path.clone()); + let mut h = Harness::builder().with_size(egui::vec2(1400.0, 900.0)).build_eframe(move |_| { + let mut app = PdfCraftApp::new(); + app.set_option("language", "en").unwrap(); + app.open_path(&first); + app.open_path(&second); + app + }); + h.run_steps(4); + // B is prepared in the background meanwhile; its reply waits behind A's prompt. + std::thread::sleep(std::time::Duration::from_millis(300)); + h.run_steps(2); + assert_eq!(h.state().password_prompt.as_ref().and_then(|p| p.path.clone()), Some(a_path.clone())); + h.state_mut().submit_password(Some("reader".into())); + support::wait_until(&mut h, "the next prompt", |app| app.password_prompt.is_some()); + assert_eq!(open_paths(h.state()), vec![a_path], "A's retry finished first"); + assert_eq!(h.state().password_prompt.as_ref().and_then(|p| p.path.clone()), Some(b_path)); + std::fs::remove_file(a).unwrap(); + std::fs::remove_file(b).unwrap(); +} + +/// `pdfcraft locked.pdf` followed by more files than the open worker queues at once: the files +/// past its bound wait in order and open as it drains, instead of failing to open. +#[test] +fn a_long_startup_list_behind_a_password_prompt_opens_every_file() { + const OTHERS: usize = 24; + let nonce = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH).unwrap().as_nanos(); + let locked = temp_path(&format!("startup-long-locked-{nonce}.pdf")); + std::fs::write(&locked, protected("reader", "owner", -4)).unwrap(); + let others: Vec = (0..OTHERS).map(|i| temp_path(&format!("startup-long-{i:02}-{nonce}.pdf"))).collect(); + for path in &others { + std::fs::write(path, fixture(1)).unwrap(); + } + let locked_path = locked.to_str().unwrap().to_string(); + let other_paths: Vec = others.iter().map(|p| p.to_str().unwrap().to_string()).collect(); + let (first, rest) = (locked_path.clone(), other_paths.clone()); + let mut h = Harness::builder().with_size(egui::vec2(1400.0, 900.0)).build_eframe(move |_| { + let mut app = PdfCraftApp::new(); + app.set_option("language", "en").unwrap(); + app.open_path(&first); + for path in &rest { + app.open_path(path); + } + app + }); + h.run_steps(4); + assert!(h.state().password_prompt.is_some(), "the protected file is asked for"); + h.state_mut().submit_password(Some("reader".into())); + support::wait_until(&mut h, "every file", |app| app.views.len() == OTHERS + 1); + let mut paths = open_paths(h.state()); + paths.sort(); + let mut expected = other_paths; + expected.push(locked_path); + expected.sort(); + assert_eq!(paths, expected); + std::fs::remove_file(locked).unwrap(); + for path in others { + std::fs::remove_file(path).unwrap(); + } +} + #[test] fn password_prompt_opens_and_security_tab_reports_the_details() { let mut h = Harness::builder().with_size(egui::vec2(1400.0, 900.0)).build_eframe(|_cc| { From 67a6037be1319a68569ac74de1d829e281fefa9d Mon Sep 17 00:00:00 2001 From: NetterSilver <133535268+Silvertree2010@users.noreply.github.com> Date: Mon, 9 Nov 2026 01:39:16 +0100 Subject: [PATCH 16/75] Keep the font size when making text box properties default (#5784) --- crates/ui-egui/src/app/prefs.rs | 16 +++++ crates/ui-egui/src/dialogs/comment_props.rs | 5 +- crates/ui-egui/src/tools/comments.rs | 34 ++++++++++ crates/ui-egui/src/tools/comments/composer.rs | 12 ++-- crates/ui-egui/tests/it/comments.rs | 64 +++++++++++++++++++ 5 files changed, 125 insertions(+), 6 deletions(-) diff --git a/crates/ui-egui/src/app/prefs.rs b/crates/ui-egui/src/app/prefs.rs index 1a8c2d269e..aa40ea0d24 100644 --- a/crates/ui-egui/src/app/prefs.rs +++ b/crates/ui-egui/src/app/prefs.rs @@ -127,6 +127,12 @@ impl PdfCraftApp { "border_shape": s.border_shape.name(), })) .collect::>(), + // Default font sizes of new text boxes and callouts (#5767). + "comment_font_sizes": self + .comment_prefs + .font_sizes() + .map(|(t, size)| serde_json::json!({ "tool": t.command(), "font_size": size })) + .collect::>(), // Drawn signatures keep their original form (older settings read the same). "signature": match &self.signature { Some(fill_sign::SavedSig::Drawn(s)) => Some(s), _ => None }, "signature_text": match &self.signature { Some(fill_sign::SavedSig::Typed(t)) => Some(t), _ => None }, @@ -420,6 +426,16 @@ impl PdfCraftApp { } } } + // Default font sizes (#5767), clamped by the setter; a missing or malformed size keeps + // the tool's own. + if let Some(sizes) = v["comment_font_sizes"].as_array() { + for e in sizes { + let Some(tool) = e["tool"].as_str().and_then(crate::comments::CommentTool::from_command) else { continue }; + if let Some(size) = e["font_size"].as_f64() { + self.comment_prefs.set_font_size(tool, size); + } + } + } } /// Saved signatures, stamps, actions and trusted certificates. diff --git a/crates/ui-egui/src/dialogs/comment_props.rs b/crates/ui-egui/src/dialogs/comment_props.rs index 12752dde3c..7b404c2f76 100644 --- a/crates/ui-egui/src/dialogs/comment_props.rs +++ b/crates/ui-egui/src/dialogs/comment_props.rs @@ -97,7 +97,7 @@ impl PdfCraftApp { } /// New comments of the kind at `(page, index)` take the look in `p`: its colour, opacity, - /// line width and (for shapes) fill. `p` may be a Comment Properties draft not applied yet. + /// line width, (for shapes) fill and (for text boxes and callouts) font size. `p` may be a Comment Properties draft not applied yet. pub fn make_props_default(&mut self, page: usize, index: usize, p: &CommentProps) { let Some((_, id)) = self.active_ids() else { return }; let Some(doc) = self.session.get(id) else { return }; @@ -143,6 +143,9 @@ impl PdfCraftApp { if let Some(w) = p.width { self.comment_prefs.set_width(tool, w); } + if let Some(size) = p.font_size { + self.comment_prefs.set_font_size(tool, size); + } self.notify_fmt("New {tool} comments will look like this one", &[("tool", &crate::i18n::in_sentence(tl!(tool.label())))]); } diff --git a/crates/ui-egui/src/tools/comments.rs b/crates/ui-egui/src/tools/comments.rs index 56301969a2..bfd0592a08 100644 --- a/crates/ui-egui/src/tools/comments.rs +++ b/crates/ui-egui/src/tools/comments.rs @@ -56,6 +56,9 @@ const SELECT_BLUE: Color32 = Color32::from_rgb(0x14, 0x73, 0xE6); /// Maximum stroke thickness offered by commenting controls, in PDF points. pub const MAX_LINE_WIDTH: f64 = 100.0; +/// The font sizes a text box or callout default may take (Comment Properties' range). +const MIN_FONT_SIZE: f64 = 1.0; +const MAX_FONT_SIZE: f64 = 400.0; #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)] pub enum CommentTool { @@ -258,6 +261,14 @@ impl CommentTool { matches!(self, Self::Rectangle | Self::Oval | Self::Polygon | Self::Cloud | Self::TextBox | Self::Callout) } + /// The font size new comments of this kind start with (`None`: no text). + fn default_font_size(self) -> Option { + match self.sample() { + Shape::TextBox { font_size, .. } | Shape::Callout { font_size, .. } => Some(font_size), + _ => None, + } + } + /// A placeholder shape of this kind (for per-tool default styles). fn sample(self) -> Shape { match self { @@ -309,6 +320,8 @@ pub struct CommentPrefs { /// `/T` of new comments (Acrobat: Preferences ▸ Identity, default the login name). pub author: String, styles: Vec<(CommentTool, Style)>, + /// The font size of new text boxes and callouts (Make Properties Default). + font_sizes: Vec<(CommentTool, f64)>, /// Keep the tool selected after use (the pin). pub pinned: bool, /// The tool each flyout group shows (the last one used). @@ -322,6 +335,7 @@ impl Default for CommentPrefs { Self { author: login_name(), styles: ALL.iter().map(|t| (*t, Style::default_for(&t.sample()))).collect(), + font_sizes: ALL.iter().filter_map(|t| t.default_font_size().map(|s| (*t, s))).collect(), pinned: false, group_tool: [CommentTool::Note, CommentTool::Highlight, CommentTool::Ink], hide_empty_highlight_hover: false, @@ -390,6 +404,26 @@ impl CommentPrefs { } } + /// The font size of new comments of `tool` (`None`: the tool has no text). + pub fn font_size(&self, tool: CommentTool) -> Option { + self.font_sizes.iter().find(|(t, _)| *t == tool).map(|(_, s)| *s) + } + + /// Every text tool with its default font size (the persisted settings read this). + pub fn font_sizes(&self) -> impl Iterator + '_ { + self.font_sizes.iter().copied() + } + + /// Clamp and store the font size (1–400 pt, as Comment Properties allows) for a text tool; + /// a non-finite size keeps the current one. + pub fn set_font_size(&mut self, tool: CommentTool, size: f64) { + if let Some((_, s)) = self.font_sizes.iter_mut().find(|(t, _)| *t == tool) + && size.is_finite() + { + *s = size.clamp(MIN_FONT_SIZE, MAX_FONT_SIZE); + } + } + pub fn set_style(&mut self, tool: CommentTool, style: Style) { if let Some((_, s)) = self.styles.iter_mut().find(|(t, _)| *t == tool) { *s = style; diff --git a/crates/ui-egui/src/tools/comments/composer.rs b/crates/ui-egui/src/tools/comments/composer.rs index 5ae441a0cd..41c3a1e6c5 100644 --- a/crates/ui-egui/src/tools/comments/composer.rs +++ b/crates/ui-egui/src/tools/comments/composer.rs @@ -125,16 +125,18 @@ pub(crate) fn composer(ctx: &egui::Context, view: &mut DocView, info: &DocInfo, view.comments.tool_done = true; let rotation = page_rotation(info, page); let border = prefs.style(CommentTool::TextBox).width; + let font_size = prefs.font_size(CommentTool::TextBox).unwrap_or(12.0); let rect = match (size, info.pages.get(page)) { - (Some(size), Some(p)) => drawn_text_box_rect(p, c.at, size, &text, 12.0, border), - _ => text_box_rect(c.at, &text, 12.0, rotation), + (Some(size), Some(p)) => drawn_text_box_rect(p, c.at, size, &text, font_size, border), + _ => text_box_rect(c.at, &text, font_size, rotation), }; - Some(new_comment(&cx, CommentTool::TextBox, Shape::TextBox { rect, font_size: 12.0 }, text)) + Some(new_comment(&cx, CommentTool::TextBox, Shape::TextBox { rect, font_size }, text)) } ComposerKind::Callout { point } => { view.comments.tool_done = true; // A callout's leader and box stay in user-space axes (its appearance isn't turned). - let rect = text_box_rect(c.at, &text, 10.0, 0); + let font_size = prefs.font_size(CommentTool::Callout).unwrap_or(10.0); + let rect = text_box_rect(c.at, &text, font_size, 0); let side = if point[0] < rect[0] { rect[0] } else if point[0] > rect[2] { @@ -143,7 +145,7 @@ pub(crate) fn composer(ctx: &egui::Context, view: &mut DocView, info: &DocInfo, (rect[0] + rect[2]) / 2.0 }; let knee = [(point[0] + side) / 2.0, (rect[1] + rect[3]) / 2.0]; - Some(new_comment(&cx, CommentTool::Callout, Shape::Callout { rect, knee, point, font_size: 10.0, ending: LineEnding::OpenArrow }, text)) + Some(new_comment(&cx, CommentTool::Callout, Shape::Callout { rect, knee, point, font_size, ending: LineEnding::OpenArrow }, text)) } ComposerKind::Replace => { view.comments.tool_done = true; diff --git a/crates/ui-egui/tests/it/comments.rs b/crates/ui-egui/tests/it/comments.rs index 56984b6882..532c45de91 100644 --- a/crates/ui-egui/tests/it/comments.rs +++ b/crates/ui-egui/tests/it/comments.rs @@ -4355,3 +4355,67 @@ fn escape_with_a_fill_tool_deselects_every_comment_then_drops_the_tool() { h.run_steps(2); assert_eq!(h.state().quick_tool, QuickTool::Select); } + +/// #5767: Make Properties Default on a text box or callout keeps its font size too, so the next +/// one (also after a restart) takes it, along with the colour. +#[test] +fn properties_dialog_makes_the_text_box_font_size_default() { + let blue = [0.0, 0.47, 0.84]; + let mut h = harness(|_| {}); + let font_size_of_last = |h: &Harness<'static, PdfCraftApp>| { + let s = h.state(); + let doc = s.session.get(s.views[0].id).unwrap(); + let last = doc.info.annotations.len() - 1; + doc.comment_props(0, last).unwrap() + }; + for (quick, size, first, second, third) in + [("freetext", 24.0, (40.0, 60.0), (40.0, 120.0), (40.0, 100.0)), ("callout", 18.0, (150.0, 40.0), (150.0, 100.0), (150.0, 60.0))] + { + h.state_mut().set_option("quick", quick).unwrap(); + if quick == "callout" { + drag_pt(&mut h, first, (first.0 + 60.0, first.1 + 20.0)); + } else { + click_pt(&mut h, first); + } + post_text(&mut h, "First"); + let index = comments(&h).len() - 1; + h.state_mut().open_comment_props(0, index); + h.run_steps(2); + { + let d = h.state_mut().comment_props.as_mut().expect("open"); + d.edited.font_size = Some(size); + d.edited.color = Some(blue); + } + h.get_by_label("Make Properties Default").click(); + h.run_steps(1); + h.get_by_label("OK").click(); + h.run_steps(3); + assert_eq!(font_size_of_last(&h).font_size, Some(size)); + + // The next one takes the font size and the colour. + h.state_mut().set_option("quick", quick).unwrap(); + if quick == "callout" { + drag_pt(&mut h, second, (second.0 + 60.0, second.1 + 20.0)); + } else { + click_pt(&mut h, second); + } + post_text(&mut h, "Second"); + let p = font_size_of_last(&h); + assert_eq!(p.font_size, Some(size), "{quick}: the next one takes the default font size"); + assert_eq!(p.color, Some(blue), "{quick}: and the default colour"); + + // The default survives a restart. + let json = h.state().persist(); + let mut fresh = harness(move |app| { + app.restore(&json); + app.set_option("quick", quick).unwrap(); + }); + if quick == "callout" { + drag_pt(&mut fresh, third, (third.0 + 60.0, third.1 + 20.0)); + } else { + click_pt(&mut fresh, third); + } + post_text(&mut fresh, "Third"); + assert_eq!(font_size_of_last(&fresh).font_size, Some(size), "{quick}: the default font size persisted"); + } +} From 359cef8b0d8c9dc7ba300fdf1ffa57ba1b6c5fc5 Mon Sep 17 00:00:00 2001 From: mtlcl410 <2368091+mtlcl410@users.noreply.github.com> Date: Mon, 9 Nov 2026 08:39:31 +0800 Subject: [PATCH 17/75] M12: an XFA page area keeps at most 128 content areas, so a subform split across them lays out within its budget (#5843) (#5844) --- crates/xfa/src/layout/flow.rs | 10 +++++- crates/xfa/src/parse.rs | 16 +++++++++ crates/xfa/src/tests.rs | 63 +++++++++++++++++++++++++++++++++++ 3 files changed, 88 insertions(+), 1 deletion(-) diff --git a/crates/xfa/src/layout/flow.rs b/crates/xfa/src/layout/flow.rs index 8f656d7eaf..b65e72477a 100644 --- a/crates/xfa/src/layout/flow.rs +++ b/crates/xfa/src/layout/flow.rs @@ -1,6 +1,6 @@ use super::common::{clamp_opt, divide, is_row, moved, solid_extent, splittable}; use super::items::Item; -use super::{Ctx, Frame, Layout, Layouter, MAX_DEPTH, Node, Rect, SomIndices, Subform, XfaError, data_path, instance_count}; +use super::{Ctx, Frame, Layout, Layouter, MAX_DEPTH, MAX_MEASURES, Node, Rect, SomIndices, Subform, XfaError, data_path, instance_count}; impl Layouter<'_> { /// Place `node` as one block of size (`w`, `h`) at the cursor: in this content area when it fits, @@ -75,6 +75,14 @@ impl Layouter<'_> { let mut seg_top = top; let mut first_segment = true; loop { + // Each pass rescans every item still to be placed. A block split across very many + // content areas, and so across very many pages, would do this once per page, costing + // time quadratic in its items; the still-unplaced count spends the work budget so a + // crafted template stops rather than run for seconds before the page limit catches it. + self.measures.set(self.measures.get().saturating_add(rest.len())); + if self.measures.get() > MAX_MEASURES { + return Err(XfaError::TooLarge("a subform is split across too many content areas to lay out".into())); + } let (rect, _, placed) = self.cursor(); let limit = rect.bottom() + shift; let mut cut = limit; diff --git a/crates/xfa/src/parse.rs b/crates/xfa/src/parse.rs index 76409dc697..a627c70e5f 100644 --- a/crates/xfa/src/parse.rs +++ b/crates/xfa/src/parse.rs @@ -21,6 +21,10 @@ const MAX_IMAGE: usize = 32 << 20; const MAX_ITEMS: usize = 1000; /// Longest text run kept. const MAX_TEXT: usize = 1 << 20; +/// Most content areas kept on one page area. A master page has a handful (the most in the +/// 500-form corpus is nine); laying a splittable subform out across tens of thousands of them, +/// which `next_area` walks before starting a new page, costs time quadratic in the paragraphs. +const MAX_CONTENT_AREAS: usize = 128; /// Parse a measurement such as `12.7mm`, `72pt`, `1in`, `2.54cm` into points. Bare numbers are /// points. `None` for anything else (including `em`). @@ -609,7 +613,19 @@ fn parse_page_set(n: XmlNode, warnings: &mut Vec, depth: usize) -> PageS attr_measure(c, "h").unwrap_or(height).clamp(1.0, height), ) }) + .take(MAX_CONTENT_AREAS + 1) .collect(); + if content.len() > MAX_CONTENT_AREAS { + content.truncate(MAX_CONTENT_AREAS); + // Said once however many warnings came before: other warnings fill the shared cap of + // 100 without limit of their own, and this one must not go silent behind them. It is + // kept first, so whether it was said is one comparison, not a scan of every warning + // for every page area (nothing else is put first). + let said = format!("a page area has more than {MAX_CONTENT_AREAS} content areas; the rest are left out"); + if warnings.first() != Some(&said) { + warnings.insert(0, said); + } + } if content.is_empty() { content.push(Rect::new(0.0, 0.0, width, height)); } diff --git a/crates/xfa/src/tests.rs b/crates/xfa/src/tests.rs index 2e70a81472..b4c59df5d0 100644 --- a/crates/xfa/src/tests.rs +++ b/crates/xfa/src/tests.rs @@ -188,6 +188,46 @@ fn the_template_parses_with_fonts_captions_and_rich_text() { assert!(last.break_before_page); } +/// A page area that declares far more content areas than any real master keeps only the first +/// 128, with a warning. `split_block` walks every content area of a page before starting a new +/// one, re-scanning the unplaced items each time, so tens of thousands of them cost time +/// quadratic in a splittable subform's paragraphs. Capping the count bounds that walk. +#[test] +fn a_page_area_keeps_at_most_a_hundred_and_twenty_eight_content_areas() { + let areas = (0..200).map(|i| format!("", i)).collect::(); + let xml = format!( + r##" + + + +"## + ); + let (tpl, warnings) = parse(&xml).unwrap(); + let ps = tpl.root.page_set.as_ref().unwrap(); + assert_eq!(ps.areas[0].content.len(), 128, "the content areas are capped"); + assert!(warnings.iter().any(|w| w.contains("content areas")), "a warning records the cap: {warnings:?}"); + // Still said after a hundred other warnings (CodeRabbit review on #5844): images that can't + // be decoded, parsed before the page set, fill the warnings' shared cap first. + let bad = "!!!".repeat(120); + let crowded = xml.replace("", &format!("{bad}")); + let (_, warnings) = parse(&crowded).unwrap(); + assert!(warnings.len() > 100, "the other warnings came first: {}", warnings.len()); + assert!(warnings.iter().any(|w| w.contains("content areas")), "the cap is still recorded"); + // Many capped page areas say it once, found without scanning every warning (it comes first). + let page_area = |n: usize| format!(r#"{areas}"#); + let many: String = (0..50).map(page_area).collect(); + let crowded = crowded.replacen(&page_area(0).replace("p0", "front"), &many, 1); + let (_, warnings) = parse(&crowded).unwrap(); + assert_eq!(warnings.iter().filter(|w| w.contains("content areas")).count(), 1, "said once"); + assert!(warnings[0].contains("content areas"), "kept first: {:?}", warnings.first()); +} + #[test] fn rich_text_max_length_reaches_the_field_and_layout() { for (attribute, expected) in [ @@ -678,6 +718,29 @@ fn empty_repeating_subforms_spend_the_layout_budget() { assert!(layout_xml(&tpl).is_ok()); } +#[test] +fn a_block_split_across_very_many_pages_spends_the_layout_budget() { + // One-point content areas slice a flowed subform's paragraphs one line to a page, so the + // block is split across hundreds of pages; split_block rescans the still-unplaced + // paragraphs once per page, which is quadratic in their number. Fewer paragraphs than this + // lay out (188 pages here), so the page limit does not catch it; the rescans spend the + // work budget instead. + let areas: String = (0..128).map(|i| format!("")).collect(); + let paras: String = (0..2500).map(|i| format!("

Paragraph number {i} with a few words to fill a line.

")).collect(); + let tpl = format!( + "" + ); + match layout_xml(&tpl) { + // It says what ran out, not that the form is nested too deeply (CodeRabbit review on #5844). + Err(XfaError::TooLarge(m)) => assert!(m.contains("split across"), "{m}"), + other => panic!("a block split across very many pages must be refused: {other:?}"), + } +} + #[test] fn a_packet_stream_listed_many_times_is_read_once() { let mut doc = Document::open(Arc::new(shell(""))).unwrap(); From 87139acf2c0f472da86cc9049ee61e3fb2f85a16 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:39:52 -0500 Subject: [PATCH 18/75] fix(create): image export checks a JPEG's Extended XMP chunks before decoding it (#5860) --- crates/create/src/extract.rs | 25 ++++++++++- crates/create/src/tests.rs | 64 +++++++++++++++++++++++++++ crates/create/src/upright_image.rs | 71 ++++++++++++++++++++++-------- 3 files changed, 141 insertions(+), 19 deletions(-) diff --git a/crates/create/src/extract.rs b/crates/create/src/extract.rs index 863ad2da0a..46b203a071 100644 --- a/crates/create/src/extract.rs +++ b/crates/create/src/extract.rs @@ -302,9 +302,14 @@ thread_local! { } /// The size a JPEG's header gives, which is what decoding it goes by, whatever its dictionary -/// says: an image of one pixel can hold a header of 500 megapixels. +/// says: an image of one pixel can hold a header of 500 megapixels. `None`, as for a JPEG that +/// can't be read, when its metadata would take the decoder quadratic time +/// ([`decodable_metadata`](crate::upright_image::decodable_metadata)). fn jpeg_size(data: &[u8]) -> Option<(usize, usize)> { use zune_jpeg::zune_core::{bytestream::ZCursor, options::DecoderOptions}; + if !crate::upright_image::decodable_metadata(data) { + return None; + } let options = DecoderOptions::default().set_strict_mode(false).set_max_width(usize::from(u16::MAX)).set_max_height(usize::from(u16::MAX)); let mut dec = zune_jpeg::JpegDecoder::new_with_options(ZCursor::new(data), options); dec.decode_headers().ok()?; @@ -329,6 +334,15 @@ fn jpeg_samples(data: &[u8], space: &Space, w: usize, h: usize) -> Option<(Vec return cmyk_jpeg_samples(data, w, h).map(|px| (px, Space::Cmyk)), _ => {} } + // Past the decoder's metadata bound, drop the Extended XMP chunks rather than the mask: + // decoded as it stands, an over-chunked JPEG is written as is, without its /SMask. + let stripped; + let data = if crate::upright_image::decodable_metadata(data) { + data + } else { + stripped = crate::upright_image::without_extended_xmp(data); + &stripped[..] + }; let img = jpeg_of_size(data, w, h)?; Some(match space { Space::Gray => (img.into_luma8().into_raw(), Space::Gray), @@ -340,6 +354,15 @@ fn jpeg_samples(data: &[u8], space: &Space, w: usize, h: usize) -> Option<(Vec Option> { use zune_jpeg::zune_core::{bytestream::ZCursor, colorspace::ColorSpace, options::DecoderOptions}; + // Past the decoder's metadata bound, decode the image without its Extended XMP chunks + // rather than not at all: written out as stored, a CMYK JPEG shows inverted colours. + let stripped; + let data = if crate::upright_image::decodable_metadata(data) { + data + } else { + stripped = crate::upright_image::without_extended_xmp(data); + &stripped[..] + }; let options = DecoderOptions::default().set_strict_mode(false).set_max_width(usize::from(u16::MAX)).set_max_height(usize::from(u16::MAX)); let mut dec = zune_jpeg::JpegDecoder::new_with_options(ZCursor::new(data), options); dec.decode_headers().ok()?; diff --git a/crates/create/src/tests.rs b/crates/create/src/tests.rs index 88a6e05ac9..9d5a70ffcd 100644 --- a/crates/create/src/tests.rs +++ b/crates/create/src/tests.rs @@ -941,6 +941,43 @@ fn damaged_masks_and_sizes_never_panic() { } } +#[test] +fn a_jpeg_with_thousands_of_extended_xmp_chunks_keeps_its_mask() { + // Export decodes a JPEG with a soft mask, or a JPEG soft mask, to make a PNG with alpha. A JPEG + // whose Extended XMP chunks would take the decoder quadratic time is decoded without them, + // which is linear in what is left, so the mask survives; a JPEG mask past the same bound is + // still left out, without reaching the decoder. + let flood = crate::upright_image::tests::with_extended_xmp_chunks(&real_jpeg(), 5_000); + let mut doc = from_images(&[("flat.png".into(), png_bytes(false))]).unwrap(); + with_image(&mut doc, |doc| { + let mask = doc.add(Object::Stream(Stream::from_raw(image_dict(4, 2, "DeviceGray", 8), vec![0; 8]))); + let mut d = image_dict(4, 2, "DeviceGray", 8); + d.set(b"Filter".to_vec(), Object::name("DCTDecode")); + d.set(b"SMask".to_vec(), Object::Ref(mask)); + Stream::from_raw(d, flood.clone()) + }); + crate::extract::JPEGS_DECODED.with(|n| n.set(0)); + let out = extract_images(&doc, &[0], 0); + assert_eq!(out.images[0].extension, "png"); + assert_eq!(crate::extract::JPEGS_DECODED.with(std::cell::Cell::get), 1); + // The chunks themselves are gone, so the decoder never saw them. + assert!(!out.images[0].data.windows(5).any(|w| w == b"http:")); + + crate::extract::JPEGS_DECODED.with(|n| n.set(0)); + + let mut doc = from_images(&[("flat.png".into(), png_bytes(false))]).unwrap(); + with_image(&mut doc, |doc| { + let mut md = image_dict(4, 2, "DeviceGray", 8); + md.set(b"Filter".to_vec(), Object::name("DCTDecode")); + let mask = doc.add(Object::Stream(Stream::from_raw(md, flood.clone()))); + let mut d = image_dict(4, 2, "DeviceRGB", 8); + d.set(b"SMask".to_vec(), Object::Ref(mask)); + Stream::from_raw(d, vec![10; 24]) + }); + assert_eq!(alphas(&exported_rgba(&doc).1), [255; 8]); + assert_eq!(crate::extract::JPEGS_DECODED.with(std::cell::Cell::get), 0); +} + #[test] fn indexed_image_export_clamps_samples_to_hival() { for palette in [vec![255, 0, 0, 0, 0, 255], vec![255, 0, 0, 0, 0, 255, 0, 255, 0, 255, 255, 255]] { @@ -2788,3 +2825,30 @@ fn proportions_up_to_the_limit_scale_as_they_did() { let mask = noise(20, 10); assert_eq!(crate::extract::scaled(&mask, 20, 10), mask); } + +/// A CMYK JPEG past the decoder's Extended XMP bound is still decoded (its chunks left out) and +/// exported with the colours the page shows, not written as stored, which inverts them. +#[test] +fn a_cmyk_jpeg_flooded_with_extended_xmp_is_exported_in_its_colours() { + let jpeg = crate::upright_image::tests::with_extended_xmp_chunks(&flat_four_component_jpeg([255, 0, 0, 0], 0), 5_000); + let mut doc = from_images(&[("flat.png".into(), png_bytes(false))]).unwrap(); + with_image(&mut doc, |doc| { + let mut d = image_dict(8, 8, "DeviceCMYK", 8); + d.set(b"Filter".to_vec(), Object::name("DCTDecode")); + let m = doc.add(Object::Stream(Stream::from_raw(image_dict(8, 8, "DeviceGray", 8), vec![255; 64]))); + d.set(b"SMask".to_vec(), Object::Ref(m)); + Stream::from_raw(d, jpeg.clone()) + }); + let (_, rgba) = exported_rgba(&doc); + assert!(rgba.as_chunks::<4>().0.iter().all(|p| p[0] <= 6 && p[1] >= 249 && p[2] >= 249), "{:?}", &rgba[..4]); +} + +/// Stripping leaves everything but the Extended XMP chunks, byte for byte. +#[test] +fn stripping_extended_xmp_leaves_the_rest_of_the_jpeg() { + let plain = flat_four_component_jpeg([255, 0, 0, 0], 0); + let flooded = crate::upright_image::tests::with_extended_xmp_chunks(&plain, 3); + assert_ne!(flooded, plain); + // The helper puts a stray byte before the chunks, which the decoder skips: it stays. + assert_eq!(crate::upright_image::without_extended_xmp(&flooded), [&plain[..2], b"\0".as_slice(), &plain[2..]].concat()); +} diff --git a/crates/create/src/upright_image.rs b/crates/create/src/upright_image.rs index b89c517190..d193a38bfb 100644 --- a/crates/create/src/upright_image.rs +++ b/crates/create/src/upright_image.rs @@ -30,7 +30,7 @@ pub(crate) fn upright(bytes: &[u8]) -> Cow<'_, [u8]> { } fn turned(bytes: &[u8]) -> Option> { - if !bytes.starts_with(&[0xFF, 0xD8]) || extended_xmp_chunks(bytes) > MAX_EXTENDED_XMP_CHUNKS { + if !bytes.starts_with(&[0xFF, 0xD8]) || !decodable_metadata(bytes) { return None; } #[cfg(test)] @@ -65,10 +65,41 @@ fn turned(bytes: &[u8]) -> Option> { /// decoder re-sorts and rescans every chunk so far after each marker, quadratic in their number. const MAX_EXTENDED_XMP_CHUNKS: usize = 1024; +/// Whether JPEG `bytes` may be handed to a decoder (`image`'s or `zune-jpeg`'s, which `image` uses), +/// headers included: not when it carries more than [`MAX_EXTENDED_XMP_CHUNKS`]. Every path that +/// decodes a JPEG from a file checks this first: image import here, and image export. +pub(crate) fn decodable_metadata(bytes: &[u8]) -> bool { + extended_xmp_chunks(bytes) <= MAX_EXTENDED_XMP_CHUNKS +} + /// How many Extended XMP chunks (APP1 segments) come before the image data. fn extended_xmp_chunks(bytes: &[u8]) -> usize { + let mut chunks = 0usize; + each_extended_xmp(bytes, |_| chunks += 1); + chunks +} + +/// JPEG `bytes` without their Extended XMP chunks, which nothing here reads: what is left decodes +/// in time linear in its size however many chunks the file carried. +pub(crate) fn without_extended_xmp(bytes: &[u8]) -> Vec { + // Only the bytes that are kept: a CMYK JPEG can be mostly Extended XMP, and reserving the + // whole file for the little that remains aborts the export under a tight address-space limit. + let mut removed = 0usize; + each_extended_xmp(bytes, |r| removed = removed.saturating_add(r.len())); + let mut out = Vec::with_capacity(bytes.len().saturating_sub(removed)); + let mut at = 0; + each_extended_xmp(bytes, |r| { + out.extend_from_slice(bytes.get(at..r.start).unwrap_or_default()); + at = r.end; + }); + out.extend_from_slice(bytes.get(at..).unwrap_or_default()); + out +} + +/// Calls `f` with the byte range of each Extended XMP chunk (APP1 segment) before the image data. +fn each_extended_xmp(bytes: &[u8], mut f: impl FnMut(std::ops::Range)) { const PREFIX: &[u8] = b"http://ns.adobe.com/xmp/extension/\0"; - let (mut i, mut chunks) = (2usize, 0usize); + let mut i = 2usize; while let Some(&byte) = bytes.get(i) { // Stray bytes between segments, which the decoder skips too. if byte != 0xFF { @@ -86,14 +117,16 @@ fn extended_xmp_chunks(bytes: &[u8]) -> usize { _ => { let Some(&[hi, lo]) = bytes.get(i.saturating_add(2)..i.saturating_add(4)) else { break }; let len = usize::from(u16::from_be_bytes([hi, lo])); - if marker == 0xE1 && bytes.get(i.saturating_add(4)..).is_some_and(|seg| seg.starts_with(PREFIX)) { - chunks += 1; + let end = i.saturating_add(2).saturating_add(len.max(2)); + // The prefix must start the APP1 segment's own payload: a zero-length header + // followed by prefix bytes is not a chunk. + if marker == 0xE1 && bytes.get(i.saturating_add(4)..end.min(bytes.len())).is_some_and(|seg| seg.starts_with(PREFIX)) { + f(i..end.min(bytes.len())); } - i = i.saturating_add(2).saturating_add(len.max(2)); + i = end; } } } - chunks } /// The resolution in a JFIF header, if the file has one in inches or centimetres. @@ -120,7 +153,7 @@ fn jfif_density(bytes: &[u8]) -> Option { } #[cfg(test)] -mod tests { +pub(crate) mod tests { use super::*; use crate::tests::phone_jpeg; @@ -138,22 +171,24 @@ mod tests { assert!(matches!(upright(b"\xff\xd8 not a jpeg"), Cow::Borrowed(_))); } + /// `jpeg` with `n` empty chunks of one never-complete Extended XMP packet after its SOI. + pub(crate) fn with_extended_xmp_chunks(jpeg: &[u8], n: usize) -> Vec { + let mut seg = b"http://ns.adobe.com/xmp/extension/\0".to_vec(); + seg.extend_from_slice(&[b'0'; 32]); // the packet's GUID + seg.extend_from_slice(&1000u32.to_be_bytes()); // its full length + seg.extend_from_slice(&0u32.to_be_bytes()); // this chunk's offset + let mut chunk = vec![0xFF, 0xE1]; + chunk.extend_from_slice(&u16::try_from(seg.len() + 2).unwrap().to_be_bytes()); + chunk.extend_from_slice(&seg); + [&jpeg[..2], b"\0".as_slice(), &chunk.repeat(n), &jpeg[2..]].concat() + } + #[test] fn a_jpeg_with_thousands_of_extended_xmp_chunks_is_not_decoded() { // Empty chunks of one never-complete extended packet: the decoder re-sorts and rescans every // chunk so far after each marker, quadratic in their number. - let chunk = |n: usize| { - let mut seg = b"http://ns.adobe.com/xmp/extension/\0".to_vec(); - seg.extend_from_slice(&[b'0'; 32]); // the packet's GUID - seg.extend_from_slice(&1000u32.to_be_bytes()); // its full length - seg.extend_from_slice(&0u32.to_be_bytes()); // this chunk's offset - let mut out = vec![0xFF, 0xE1]; - out.extend_from_slice(&u16::try_from(seg.len() + 2).unwrap().to_be_bytes()); - out.extend_from_slice(&seg); - out.repeat(n) - }; let photo = phone_jpeg(6, 144); - let with = |n: usize| [&photo[..2], b"\0".as_slice(), &chunk(n), &photo[2..]].concat(); + let with = |n: usize| with_extended_xmp_chunks(&photo, n); DECODES.with(|c| c.set(0)); assert!(matches!(upright(&with(4)), Cow::Owned(_)), "a few chunks are an ordinary photo"); assert!(matches!(upright(&with(5_000)), Cow::Borrowed(_))); From ccbfa6d24a9e450066dad5192e2f4404bec778d3 Mon Sep 17 00:00:00 2001 From: "Hasan N." <42965049+hasannadeem@users.noreply.github.com> Date: Sun, 8 Nov 2026 19:40:07 -0500 Subject: [PATCH 19/75] M14: localization parity counts every catalog with the i18n report (#5805) (#5894) --- ROADMAP.md | 33 +++++----- docs/localization-parity.md | 119 ++++++++++++++++++++---------------- 2 files changed, 84 insertions(+), 68 deletions(-) diff --git a/ROADMAP.md b/ROADMAP.md index 2df1bd26ca..ee91096427 100644 --- a/ROADMAP.md +++ b/ROADMAP.md @@ -198,27 +198,28 @@ What's strong and missing in each: [target-app-parity.md](docs/target-app-parity ## Languages -Twelve key languages (historical % snapshot vs 2,143 keys, except Vietnamese, Brazilian -Portuguese and Spanish vs the current union). Detail and script notes: [localization-parity.md](docs/localization-parity.md). +Twelve key languages: the share of the 2,897 strings in the i18n report's denominator (the interface's strings plus every catalog key), from +`cargo xtask i18n-report` on 2026-11-03. Detail and script notes: [localization-parity.md](docs/localization-parity.md). | Language | Code | Strings | Status | |---|---|---|---| | English | `en` | 100% | full | -| Simplified Chinese | `zh-hans` | 97.4% | partial (native LXGW WenKai Lite; no Hans face in the web build, #826) | -| Spanish | `es` | 100% of current union | partial (dialogs / tooltips; no help) | -| Hindi | `hi` | 0% | none | -| Arabic | `ar` | 100% | partial (logical catalog, explicit RTL labels, mirrored stock chrome; native/browser/IME workflows unqualified) | -| French | `fr` | 99.3% | partial | -| Portuguese (Brazil) | `pt-br` | 100% of current union | partial | -| Indonesian | `id` | 0% | none | -| Japanese | `ja` | 99.1% | partial | -| German | `de` | 99.3% | partial | -| Korean | `ko` | 0% | none | -| Vietnamese | `vi` | 100% of current union | partial (dialogs / tooltips; no help) | +| Simplified Chinese | `zh-hans` | 88.4% | partial (native LXGW WenKai Lite; no Hans face in the web build, #826) | +| Spanish | `es` | 98.4% | partial (dialogs / tooltips; no help) | +| Hindi | `hi` | 97.5% | partial (Devanagari rendering unqualified) | +| Arabic | `ar` | 97.9% | partial (logical catalog, explicit RTL labels, mirrored stock chrome; native/browser/IME workflows unqualified) | +| French | `fr` | 98.4% | partial | +| Portuguese (Brazil) | `pt-br` | 98.4% | partial | +| Indonesian | `id` | 97.5% | partial | +| Japanese | `ja` | 89.6% | partial | +| German | `de` | 89.7% | partial | +| Korean | `ko` | 97.5% | partial (no Hangul UI face) | +| Vietnamese | `vi` | 98.4% | partial (dialogs / tooltips; no help) | Also shipped: Traditional Chinese, Russian, Bulgarian, Telugu, Hungarian, Ukrainian, Italian, -Kazakh, Polish, Slovenian, Greek, Turkish, Lithuanian, Romanian and Czech: **15 more catalogs** -(24 interface languages including English). +Kazakh, Polish, Slovenian, Greek, Turkish, Lithuanian, Finnish, Albanian, Catalan, Romanian, +Norwegian Bokmål, Persian, Central Kurdish, Thai, Lao and Czech: **23 more catalogs** +(35 interface languages including English). ## Upcoming @@ -252,7 +253,7 @@ Ranked; detail and the full milestone table in [docs/roadmap.md](docs/roadmap.md - **2026-11-03 (community, M7, Thai word breaking):** Thai rows end between words, found with ICU4X's dictionary line segmenter (`icu_segmenter` 2.3, compiled dictionary data, no LSTM model; Unicode-3.0), instead of only at spaces when rewrapping and at any cluster in Add text. A row never starts with a following vowel, tone or other mark, Mai Yamok or Paiyannoi, nor ends with a leading vowel; a word wider than the row still breaks between clusters, and the spaces where an Add text row ends are dropped as in Latin text. Rewrapping existing text works out the breaks once per run of non-space characters. Rewriting a paragraph into Thai still needs a source font with the Thai glyphs (there is no Thai fallback face for it yet). The compiled dictionaries (Thai, Lao, Khmer, Burmese) add about 1.8 MB to a stripped release binary. Estimates unchanged. - **2026-11-03 (M7, #5835):** In image crop mode, a right-click on the image, its crop box or its handles keeps the crop and opens no menu, so the menu no longer switches as the pointer moves. A right-click elsewhere clears the selection and opens the page menu. UI tests cover all four page rotations, overlapping images, a 45° image, and a check that the Done/Cancel/Remove Crop pills at a page edge still keep the crop. -- **2026-11-02 (M12, #1297):** `model_add` embeds a U3D or PRC file, recognised by its header, as a 3D annotation with a chosen activation and a poster (an image or PDF page fitted in the box, or a labelled frame); an older file's catalog `/Version` is raised to what the model needs. Views, a rendered poster and a UI command are later steps. Estimates unchanged. +- **2026-11-03 (M14 docs, #5805):** `docs/localization-parity.md` and the Languages table here count every catalog with `cargo xtask i18n-report`: the share of the 2,897 strings in its denominator, on `14ae839a`. The catalog-union measure it replaces had grown to 2,895 keys and left four catalogs at "100%" that were 44 keys short. Hindi, Indonesian and Korean are listed with their catalogs (97.5% each) instead of 0, and Thai, Lao and Central Kurdish join the table. The ≈ 68% localization score is not re-scored. - **2026-11-02 (M10, #5743):** macOS print jobs name their title and owner (`com.apple.print.JobInfo.PMJobName`, `PMJobOwner`), as the system print dialog does, because a Kyocera filter crashes on a job without them. Checked by a test and through `doc_print` against a stand-in `lp`; not yet on a physical printer. diff --git a/docs/localization-parity.md b/docs/localization-parity.md index 1944575ce3..4e30fcdf0a 100644 --- a/docs/localization-parity.md +++ b/docs/localization-parity.md @@ -1,24 +1,32 @@ # Localization parity -> **Last reviewed:** 2026-11-03 · **Last updated:** 2026-11-03 · **Change:** minor (Turkish: 2,867 entries, 0 missing in `cargo xtask i18n-report` on 2026-11-03; AI-assisted) -> **Last reviewed:** 2026-10-25 · **Last updated:** 2026-10-25 · **Change:** minor (Persian (`fa`, 2,785 entries) added right-to-left like Arabic and Kurdish; Arabic current state corrected: catalogs stay logical, RTL labels take an explicit direction and stock chrome mirrors; native/browser RTL input, IME and complex workflows remain unqualified. Native release Hans face via LXGW WenKai Lite; release-pinned craft-fonts still has no Telugu face. Current union 2,798 keys. Scores and estimates unchanged) · **Target:** Adobe Acrobat Pro (Acrobat DC, continuous track 26.002.21931, macOS) +> **Last reviewed:** 2026-11-03 · **Last updated:** 2026-11-03 · **Change:** minor (every catalog restated from `cargo xtask i18n-report` on `main` at `14ae839a`: 2,897 strings in the report's denominator; Hindi, Indonesian, Korean, Thai, Lao and Central Kurdish listed with their catalogs. Scores and estimates unchanged) · **Target:** Adobe Acrobat Pro (Acrobat DC, continuous track 26.002.21931, macOS) Per-language status of PdfCraft's interface against Acrobat Pro's. How translations work and how to add a language: [localization.md](localization.md). Summary in [ROADMAP.md](../ROADMAP.md). ## How this was measured -- **Strings translated** is measured: the catalogs in `crates/ui-egui/src/i18n/.tsv` (since moved to craft-i18n's `locales//`), counted - as distinct `(context, English)` keys with a non-empty translation. The denominator is the union of - keys across all catalogs, counting language-specific context rows such as `theme`/`Appearance`. - **That union is a moving target**: every change that adds a string widens it, so a count is only - comparable with others measured on the same date, and a catalog that reached 100% falls behind - without losing a single translation. Every figure here therefore carries its date. The current - union is **2,798** keys as of 2026-10-25, against the **2,143-key snapshot** of 2026-10-10; earlier - revisions quote the union of their own date (see the revision history), not this one. The - twelve-language percentage table below still quotes the 2026-10-10 snapshot for every language - except Spanish, Vietnamese, Lithuanian and Brazilian Portuguese, which are measured against the current union. English is +- **Strings translated** is measured with `cargo xtask i18n-report`, which CI also runs: the + English strings in its denominator, less the ones a catalog has no translation for. The + denominator is the UI source's literals, the command, All tools and engine wordings, and every + key of every bundled catalog, which brings in the contextual, plural and dynamically looked-up + strings the scan can't see. On `main` at `14ae839a` (2026-11-03) that is **2,897** strings, and + every count and percentage in the tables below is of that number. A key only some catalogs have + counts as missing in the others; a contextual one (such as `theme`/`Appearance`) is met by a + plain translation of the same English, but a `@plural` one only by its own plural forms. The + denominator still grows whenever the interface gains a string, so a catalog that was complete + falls behind without losing a translation, and every figure here carries its date. English is built in. +- **Catalogs:** the figures above were taken from the in-repo catalogs + (`crates/ui-egui/src/i18n/.tsv`), which have since moved to craft-i18n's `locales//`. + The report reads the catalogs the build embeds, so measuring again needs `CRAFT_I18N_DIR` set to a + craft-i18n checkout at the commit in `craft-i18n.rev`; without it the report says it was + skipped. +- **Before 2026-11-03** this page counted distinct `(context, English)` keys against the union of + keys across all catalogs: 2,143 keys on 2026-10-10, 2,798 on 2026-10-25, and 2,895 on + `14ae839a`. That union also grew whenever one catalog added a key no other had, so it moved every + row at once. Earlier revisions in the history quote it, at the size of their own date. - **Not counted:** error details from the engine and the operating system, which are shown in English inside a translated frame; there is no translated user documentation or help yet (Acrobat ships localized help). @@ -32,64 +40,71 @@ add a language: [localization.md](localization.md). Summary in [ROADMAP.md](../R ## The twelve key languages -| Language | Code | UI strings translated | Dialogs / tooltips / help | Script support | Native review | Status | Acrobat ships it | To `full` (h) | +| Language | Code | UI strings translated (of 2,897, 2026-11-03) | Dialogs / tooltips / help | Script support | Native review | Status | Acrobat ships it | To `full` (h) | |---|---|---|---|---|---|---|---|---| | English | `en` | built in (100%) | all; README and docs | Latin | yes | **full** | yes | 0 | -| Simplified Chinese | `zh-hans` | 2,088 (97.4%) | dialogs and tooltips through `tl!`; no help | CJK text and IME work; native craft-fonts builds embed the **LXGW WenKai Lite** Hans face, while the web build has no Hans face beyond BIZ UDPGothic, and missing-glyph reports (#826, #688, #728, #758; Flatpak #689) predate the native face | no | partial | yes | 6–12 | -| Spanish | `es` | 2,798 (100% of current key union) | dialogs and tooltips; no help | Latin | no | partial | yes | 2–4 | -| Hindi | `hi` | 0 | none | Devanagari not tested (egui shapes Telugu conjuncts, so shaping is likely workable); the pinned craft-fonts manifest has Noto Sans Devanagari Regular; rendering remains unqualified | no | none | no | 10–18 | -| Arabic | `ar` | 2,143 (100%) | dialogs and tooltips; no help | Logical-order catalog; egui 0.37's epaint reorders with UAX #9, Arabic, Kurdish and Persian labels use an explicit right-to-left paragraph direction, and stock chrome is mirrored. Native/browser RTL input and IME, and complex bidirectional workflows, are not yet qualified (`localization.md` §Arabic) | no | partial | yes (Middle East edition) | 25–45 | -| French | `fr` | 2,128 (99.3%) | dialogs and tooltips; no help | Latin | no | partial | yes | 1–2 | -| Portuguese (Brazil) | `pt-br` | 2,798 (100% of current union) | dialogs and tooltips; no help | Latin | community (#288) | partial | yes | 1–2 | -| Indonesian | `id` | 0 | none | Latin | no | none | no | 3–5 | -| Japanese | `ja` | 2,123 (99.1%) | dialogs and tooltips; no help | BIZ UDPGothic from craft-fonts; IME patched on macOS, Windows, X11, Wayland; vertical text renders but can't be edited | no | partial | yes | 4–8 | -| German | `de` | 2,128 (99.3%) | dialogs and tooltips; no help | Latin | community, in progress (#828: terminology, clipping) | partial | yes | 2–4 | -| Korean | `ko` | 0 | none | Hangul IME patched (Microsoft and Apple Korean, `vendor/README.md` winit 3, 6, 9, 10); no Hangul UI face in craft-fonts | no | none | yes | 6–10 | -| Vietnamese | `vi` | 2,798 (100% of current key union) | dialogs and tooltips; no help | Latin with stacked diacritics; bundled Inter covers them (fonts test) | community | partial | no | 2–4 | +| Simplified Chinese | `zh-hans` | 2,560 (88.4%) | dialogs and tooltips through `tl!`; no help | CJK text and IME work; native craft-fonts builds embed the **LXGW WenKai Lite** Hans face, while the web build has no Hans face beyond BIZ UDPGothic, and missing-glyph reports (#826, #688, #728, #758; Flatpak #689) predate the native face | no | partial | yes | 6–12 | +| Spanish | `es` | 2,851 (98.4%) | dialogs and tooltips; no help | Latin | no | partial | yes | 2–4 | +| Hindi | `hi` | 2,825 (97.5%) | dialogs and tooltips; no help | Devanagari not tested (egui shapes Telugu conjuncts, so shaping is likely workable); the pinned craft-fonts manifest has Noto Sans Devanagari Regular; rendering remains unqualified | no | partial | no | 10–18 | +| Arabic | `ar` | 2,836 (97.9%) | dialogs and tooltips; no help | Logical-order catalog; egui 0.37's epaint reorders with UAX #9, Arabic, Kurdish and Persian labels use an explicit right-to-left paragraph direction, and stock chrome is mirrored. Native/browser RTL input and IME, and complex bidirectional workflows, are not yet qualified (`localization.md` §Arabic) | no | partial | yes (Middle East edition) | 25–45 | +| French | `fr` | 2,851 (98.4%) | dialogs and tooltips; no help | Latin | no | partial | yes | 1–2 | +| Portuguese (Brazil) | `pt-br` | 2,851 (98.4%) | dialogs and tooltips; no help | Latin | community (#288) | partial | yes | 1–2 | +| Indonesian | `id` | 2,825 (97.5%) | dialogs and tooltips; no help | Latin | no | partial | no | 3–5 | +| Japanese | `ja` | 2,597 (89.6%) | dialogs and tooltips; no help | BIZ UDPGothic from craft-fonts; IME patched on macOS, Windows, X11, Wayland; vertical text renders but can't be edited | no | partial | yes | 4–8 | +| German | `de` | 2,600 (89.7%) | dialogs and tooltips; no help | Latin | community, in progress (#828: terminology, clipping) | partial | yes | 2–4 | +| Korean | `ko` | 2,825 (97.5%) | dialogs and tooltips; no help | Hangul IME patched (Microsoft and Apple Korean, `vendor/README.md` winit 3, 6, 9, 10); no Hangul UI face in craft-fonts | no | partial | yes | 6–10 | +| Vietnamese | `vi` | 2,851 (98.4%) | dialogs and tooltips; no help | Latin with stacked diacritics; bundled Inter covers them (fonts test) | community | partial | no | 2–4 | -**Localization dimension ≈ 68%**: the mean over the twelve, counting a catalog's percentage but only -60% for Chinese and Arabic until their script problems are fixed, and 0 for Hindi, Indonesian and -Korean (still missing). Vietnamese moved from 0% to 100% in that mean (59% → 68%). **≈ 55–110 h** +**Localization dimension ≈ 68%** (scored 2026-10-13, before the counts above): the mean over the +twelve, counting a catalog's percentage of the 2,143-key snapshot but only 60% for Chinese and +Arabic until their script problems are fixed, and 0 for Hindi, Indonesian and Korean, which had no +catalog yet. Vietnamese moved from 0% to 100% in that mean (59% → 68%). Hindi, Indonesian and +Korean now have catalogs, so the score understates them. It is not re-scored here, because that +decides how their script caveats count (Devanagari rendering unqualified, no Hangul UI face). **≈ 55–110 h** to bring all twelve to `full`, of which ≈ 35–60 h is engineering (remaining RTL qualification, a web Hans face, bidi text) and the rest catalogs and fixes; native-speaker review needs humans. No language is `full` except English: every catalog leaves engine error details in English, and -none has translated help. The **To `full`** hours of Vietnamese and Spanish are therefore residual -help/docs polish (catalog and script work are already done), not missing UI strings. +none has translated help. The **To `full`** hours of Vietnamese and Spanish are therefore mostly +help/docs polish: their script work is done, and each catalog lacks 46 strings on 2026-11-03. ## Other shipped languages -| Language | Code | UI strings translated | Status | Acrobat ships it | +| Language | Code | UI strings translated (of 2,897, 2026-11-03) | Status | Acrobat ships it | |---|---|---|---|---| -| Traditional Chinese | `zh-hant` | 2,016 (94.1%) | partial (same Hans/Hant face caveat) | yes | -| Russian | `ru` | 2,043 (95.3%) | partial | yes | -| Bulgarian | `bg` | 2,094 (97.7%) | partial | no | -| Czech | `cs` | 2,794 entries (all but four keys of the current union; the engine's sentences, weekdays and certificate fields) | partial | yes | -| Telugu | `te` | 2,058 (96.0%) | partial (the release-pinned craft-fonts has no Telugu face yet) | no | -| Hungarian | `hu` | 2,126 (99.2%) | partial | yes | -| Ukrainian | `uk` | 2,142 (100%) | partial | yes | -| Italian | `it` | 2,736 entries (all but 62 keys of the current union) | partial | yes | -| Kazakh | `kk` | 2,539 entries | partial | no | -| Polish | `pl` | 2,611 entries | partial | yes | -| Slovenian | `sl` | 2,536 entries | partial | no | -| Greek | `el` | 2,545 entries | partial | no | -| Turkish | `tr` | 2,867 entries (0 missing in `cargo xtask i18n-report`, 2026-11-03) | partial; AI-assisted, judgement calls reviewed by a native speaker, full proofreading pending | yes | -| Lithuanian | `lt` | 2,798 (100% of current key union) | partial; community native-speaker review ([#1401](https://github.com/storytold/pdfcraft/pull/1401)) | no | -| Finnish | `fi` | 2,798 entries (every current key) | partial; AI-assisted, native proofreading pending | yes | -| Albanian | `sq` | 2,524 entries | partial; not yet reviewed by a native speaker | no | -| Catalan | `ca` | 2,798 entries | partial: no translated help, engine error details in English; native-speaker review by the contributor | no | -| Romanian | `ro` | 2,798 entries (every current key) | partial; AI-assisted, not yet reviewed line by line by a native speaker | no | -| Norwegian Bokmål | `nb` | 2,707 entries | partial; native-speaker terminology review by the contributor ([#1836](https://github.com/storytold/pdfcraft/pull/1836)), not yet line by line | yes | -| Persian | `fa` | 2,785 entries (all but 13 keys of the current 2,798-key union) | partial; AI-assisted, not yet reviewed line by line by a native speaker | no | +| Traditional Chinese | `zh-hant` | 2,559 (88.3%) | partial (same Hans/Hant face caveat) | yes | +| Russian | `ru` | 2,515 (86.8%) | partial | yes | +| Bulgarian | `bg` | 2,564 (88.5%) | partial | no | +| Czech | `cs` | 2,847 (98.3%) | partial | yes | +| Telugu | `te` | 2,513 (86.7%) | partial (the release-pinned craft-fonts has no Telugu face yet) | no | +| Hungarian | `hu` | 2,599 (89.7%) | partial | yes | +| Ukrainian | `uk` | 2,600 (89.7%) | partial | yes | +| Italian | `it` | 2,790 (96.3%) | partial | yes | +| Kazakh | `kk` | 2,609 (90.1%) | partial | no | +| Polish | `pl` | 2,681 (92.5%) | partial | yes | +| Slovenian | `sl` | 2,618 (90.4%) | partial | no | +| Greek | `el` | 2,615 (90.3%) | partial | no | +| Turkish | `tr` | 2,895 (99.9%) | partial; AI-assisted, judgement calls reviewed by a native speaker, full proofreading pending | yes | +| Lithuanian | `lt` | 2,851 (98.4%) | partial; community native-speaker review ([#1401](https://github.com/storytold/pdfcraft/pull/1401)) | no | +| Finnish | `fi` | 2,849 (98.3%) | partial; AI-assisted, native proofreading pending | yes | +| Albanian | `sq` | 2,604 (89.9%) | partial; not yet reviewed by a native speaker | no | +| Catalan | `ca` | 2,849 (98.3%) | partial: no translated help, engine error details in English; native-speaker review by the contributor | no | +| Romanian | `ro` | 2,849 (98.3%) | partial; AI-assisted, not yet reviewed line by line by a native speaker | no | +| Norwegian Bokmål | `nb` | 2,758 (95.2%) | partial; native-speaker terminology review by the contributor ([#1836](https://github.com/storytold/pdfcraft/pull/1836)), not yet line by line | yes | +| Persian | `fa` | 2,801 (96.7%) | partial; AI-assisted, not yet reviewed line by line by a native speaker | no | +| Central Kurdish (Sorani) | `ckb` | 2,754 (95.1%) | partial; right-to-left like Arabic | no | +| Thai | `th` | 2,751 (95.0%) | partial | no | +| Lao | `lo` | 2,504 (86.4%) | partial | no | -**20 other languages** ship (34 catalogs plus English = **35** interface languages in Preferences). -The entry counts without percentages are as of 2026-10-25; they are not percentages of the older 2,143-key snapshot above. Their presence does not imply a detailed linguistic or runtime review. +**23 other languages** ship (34 catalogs plus English = **35** interface languages in Preferences). +Their presence does not imply a detailed linguistic or runtime review. Acrobat languages PdfCraft lacks: Danish, Dutch, Swedish, Hebrew. Vietnamese is shipped here but not by Acrobat. ## Revision history | Date | Change | Summary | |---|---|---| +| 2026-11-03 | minor | Every catalog restated from `cargo xtask i18n-report` on `14ae839a` (of the 2,897 strings in the report's denominator), replacing the catalog-union measure, which had reached 2,895 keys and left rows "100%" that were 44 short (#5805). Hindi, Indonesian and Korean restated from 0 to their catalogs, and Thai, Lao and Central Kurdish added to the table. The ≈ 68% score is not re-scored | | 2026-10-25 | minor | Arabic current state corrected: catalogs stay logical, Arabic/Kurdish labels take an explicit RTL direction, and stock chrome mirrors; native/browser RTL input, IME and complex workflows remain unqualified. Native Simplified Chinese uses LXGW WenKai Lite; the web build still has no Hans face and the release-pinned craft-fonts still has no Telugu face. Scores and estimates unchanged | | 2026-10-25 | minor | Added Persian (`fa`, 2,785 entries, all but 13 of the current 2,798-key union); right-to-left like Arabic and Kurdish, with logical entries reordered by egui 0.37's epaint; AI-assisted, not yet native-reviewed | | 2026-10-21 | minor | Brazilian Portuguese (`pt-br`) now covers all 2,674 distinct keys in the current catalog union; no help or engine/OS error-detail localization; twelve-language mean still rounds to ≈ 68% | From eaeffa748b2ce5aa798352cb85690e2d44838ebf Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:40:21 -0500 Subject: [PATCH 20/75] fix(measure): charge snapping per filter stage and bound listing work (#5920) --- crates/measure/src/lib.rs | 47 +++++++++++++++++++++-- crates/measure/src/snap.rs | 39 +++++++++++-------- crates/measure/src/tests.rs | 74 +++++++++++++++++++++++++++++++++++++ 3 files changed, 141 insertions(+), 19 deletions(-) diff --git a/crates/measure/src/lib.rs b/crates/measure/src/lib.rs index d52d31dcc6..7ff8bfdcf1 100644 --- a/crates/measure/src/lib.rs +++ b/crates/measure/src/lib.rs @@ -452,6 +452,13 @@ pub struct Listing { pub truncated: bool, } const MAX_ANNOTATIONS: usize = 100_000; +/// Vertex pairs one listing compares for self-intersection in all: an area's boundary check +/// is quadratic in its vertices, and many areas (or many annotations sharing one indirect vertex +/// array) must not multiply that into a stall. +const MAX_LISTED_WORK: usize = 50_000_000; +/// Scales already read by one listing, by the indirect `/Measure` object: one scale shared by +/// many annotations is read (and its ratio decoded) once, failures included. +type Scales = std::collections::HashMap>; /// Points and text one listing keeps, as counted by [`retained`]: about a thousand /// measurements of [`MAX_POINTS`] points each. const MAX_LISTED_BYTES: usize = 32 << 20; @@ -462,8 +469,18 @@ fn retained(m: &Measurement) -> usize { .into_iter() .fold(m.points.len().saturating_mul(std::mem::size_of::()), usize::saturating_add) } -fn read_measurement(doc: &Document, d: &Dict, kind: Kind, page: usize, index: usize) -> Result { - let scale = Scale::read(doc, d.get(b"Measure").ok_or_else(|| invalid("measurement has no scale"))?)?; +/// The scale a measurement annotation reads with, through the listing's cache: a scale given by +/// reference is read once, and its error kept, however many annotations name it. +fn scale_of(doc: &Document, d: &Dict, scales: &mut Scales) -> Result { + let measure = d.get(b"Measure").ok_or_else(|| invalid("measurement has no scale"))?; + match measure.as_ref() { + Some(r) => scales.entry(r).or_insert_with(|| Scale::read(doc, measure).map_err(|e| e.to_string())).clone().map_err(|e| invalid(&e)), + None => Scale::read(doc, measure), + } +} + +fn read_measurement(doc: &Document, d: &Dict, kind: Kind, page: usize, index: usize, scales: &mut Scales) -> Result { + let scale = scale_of(doc, d, scales)?; let key = if kind == Kind::Distance { &b"L"[..] } else { b"Vertices" }; let o = doc.resolve(d.get(key).ok_or_else(|| invalid("measurement has no vertices"))?); let a = o.as_array().ok_or_else(|| invalid("invalid measurement vertices"))?; @@ -482,7 +499,9 @@ pub fn list(doc: &Document) -> Listing { let mut out = Listing::default(); let mut count = 0usize; let mut kept = 0usize; + let mut work = 0usize; let mut seen = std::collections::HashSet::new(); + let mut scales = Scales::new(); for (page, p) in pdfcraft_model::pages(doc).iter().enumerate() { let Some(o) = p.dict.get(b"Annots") else { continue }; let o = doc.resolve(o); @@ -507,7 +526,29 @@ pub fn list(doc: &Document) -> Listing { Some(b"PolygonDimension") => Kind::Area, _ => continue, }; - match read_measurement(doc, d, kind, page, index) { + // An area's boundary check compares every pair of edges: charge it before reading the + // vertices, so a listing stops at its work bound rather than at a stall. Other kinds + // read their vertices once. + let vertices = d.get(if kind == Kind::Distance { b"L".as_slice() } else { b"Vertices" }).map(|v| doc.resolve(v)); + let n = vertices.as_deref().and_then(Object::as_array).map_or(0, |a| a.len() / 2).min(MAX_POINTS); + // An area's boundary check compares every pair of edges: charge it before reading the + // vertices, so a listing stops at its work bound rather than at a stall. The scale is + // resolved first, as `read_measurement` reads it: one that is rejected reads nothing at + // all, so it must not spend the work that later, valid measurements need. + if kind == Kind::Area { + if let Err(e) = scale_of(doc, d, &mut scales) { + out.unsupported.push(Unsupported { page, index, reason: e.to_string() }); + continue; + } + work = work.saturating_add(n.saturating_mul(n)); + } else { + work = work.saturating_add(n.max(1)); + } + if work > MAX_LISTED_WORK { + out.truncated = true; + return out; + } + match read_measurement(doc, d, kind, page, index, &mut scales) { Ok(m) => { kept = kept.saturating_add(retained(&m)); if kept > MAX_LISTED_BYTES { diff --git a/crates/measure/src/snap.rs b/crates/measure/src/snap.rs index b26c389e96..c39d811d9f 100644 --- a/crates/measure/src/snap.rs +++ b/crates/measure/src/snap.rs @@ -1,7 +1,7 @@ //! Bounded path extraction and snapping in PDF user space, including nested Form XObjects. use crate::{Point, Result, check_points, distance, invalid}; -use pdfcraft_content::{Matrix, parse}; -use pdfcraft_cos::{Dict, Document, ObjRef, Object}; +use pdfcraft_content::{Matrix, ops}; +use pdfcraft_cos::{DecodeBudget, Dict, Document, ObjRef, Object}; use serde::Serialize; use std::collections::HashSet; @@ -199,7 +199,9 @@ struct Walker<'a> { doc: &'a Document, geometry: Geometry, seen: HashSet, - bytes: usize, + /// Decoded bytes left for the page: every filter stage's output counts, so a stream that + /// expands and then shrinks pays for the expansion each time it is drawn. + budget: DecodeBudget, streams: usize, } impl Walker<'_> { @@ -215,46 +217,44 @@ impl Walker<'_> { for part in arr { let s = self.doc.resolve(part); if let Object::Stream(s) = s.as_ref() { - let remaining = MAX_BYTES.saturating_sub(self.bytes).saturating_sub(joined.len()); + let remaining = self.budget.remaining(); if remaining == 0 { self.geometry.truncated = true; break; } // An undecodable part is skipped; the parts before it still snap. - match s.decoded_within(remaining) { + match self.budget.decode_within(s, remaining) { Ok(data) => { joined.extend(data); // A space, as the renderer joins pieces: a trailing comment runs on. joined.push(b' '); } - Err(_) => self.decode_failed(remaining), + Err(_) => self.decode_failed(), } } } return self.ops(&joined, resources, initial, depth); } let Object::Stream(stream) = o.as_ref() else { return Ok(()) }; - let remaining = MAX_BYTES.saturating_sub(self.bytes); + let remaining = self.budget.remaining(); if remaining == 0 { self.geometry.truncated = true; return Ok(()); } - match stream.decoded_within(remaining) { + match self.budget.decode_within(stream, remaining) { Ok(data) => self.ops(&data, resources, initial, depth), Err(_) => { - self.decode_failed(remaining); + self.decode_failed(); Ok(()) } } } - /// A tolerant decode fails on an unsupported filter or once it has produced `remaining` bytes, - /// so a failure uses up the byte budget: repeating a failing stream cannot repeat that work. - fn decode_failed(&mut self, remaining: usize) { + /// A failed decode has paid the budget for what it decoded before failing (all of it, past + /// the bound): repeating a failing stream cannot repeat that work for free. + fn decode_failed(&mut self) { self.geometry.unreadable = self.geometry.unreadable.saturating_add(1); - self.bytes = self.bytes.saturating_add(remaining); } fn ops(&mut self, data: &[u8], resources: &Dict, initial: Matrix, depth: usize) -> Result<()> { - self.bytes = self.bytes.saturating_add(data.len()); self.streams = self.streams.saturating_add(1); let mut matrix = initial; let mut stack = Vec::new(); @@ -268,7 +268,8 @@ impl Walker<'_> { let (x, y) = m.apply(x, y); [x, y] }; - for op in parse(data).ops { + // One operator at a time: a stream of millions of short operators is never held whole. + for op in ops(data) { if self.geometry.segments.len().saturating_add(path.segments.len()) >= MAX_SEGMENTS || self.geometry.targets().saturating_add(path.targets()) >= MAX_TARGETS { @@ -425,6 +426,11 @@ impl Walker<'_> { self.seen.insert(r); } self.walk(o, res, m, depth + 1)?; + // A form that hit a bound ends this walk too: more `Do`s would + // only repeat the work of copying its resources. + if self.geometry.truncated { + break; + } if let Some(r) = r { self.seen.remove(&r); } @@ -449,7 +455,8 @@ pub fn geometry(doc: &Document, page_index: usize) -> Result { let resources = p.dict.get(b"Resources").map(|o| doc.resolve(o)); let empty = Dict::new(); let resources = resources.as_ref().and_then(|o| o.as_dict()).unwrap_or(&empty); - let mut walker = Walker { doc, geometry: Geometry::default(), seen: HashSet::new(), bytes: 0, streams: 0 }; + let mut walker = + Walker { doc, geometry: Geometry::default(), seen: HashSet::new(), budget: DecodeBudget::new(MAX_BYTES, usize::MAX), streams: 0 }; // Extraction is lenient: limits and unreadable streams are reported in the geometry. if let Some(contents) = p.dict.get(b"Contents") { walker.walk(contents, resources, Matrix::IDENTITY, 0)?; diff --git a/crates/measure/src/tests.rs b/crates/measure/src/tests.rs index 88906910c7..f265f086ec 100644 --- a/crates/measure/src/tests.rs +++ b/crates/measure/src/tests.rs @@ -471,3 +471,77 @@ fn a_stream_over_the_byte_budget_uses_it_up_once() { assert!(g.truncated); assert_eq!(g.segments.len(), 1, "the parts before it still snap"); } + +/// A content stream drawn again and again pays for every filter stage's output each time, not +/// only for what the last stage leaves: `/ASCIIHexDecode /RunLengthDecode` turns 4 MiB of hex +/// into 2 MiB and then 1 MiB, and eight draws of it are 24 MiB of decoding against a 16 MiB bound. +#[test] +fn repeated_content_pays_for_every_filter_stage() { + let hex = "0058".repeat(1 << 20); + let pdf = Pdf::new() + .obj("<< /Type /Catalog /Pages 2 0 R >>") + .obj("<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 612 792] >>") + .obj("<< /Type /Page /Parent 2 0 R /Contents [4 0 R 4 0 R 4 0 R 4 0 R 4 0 R 4 0 R 4 0 R 4 0 R] >>") + .stream("/Filter [/ASCIIHexDecode /RunLengthDecode]", &hex); + let d = Document::open(Arc::new(pdf.build())).unwrap(); + let g = snap::geometry(&d, 0).unwrap(); + assert!(g.truncated, "eight draws decode past the page's bound"); +} + +/// One scale shared by many measurements is read once, and the self-intersection check of many +/// areas is charged against one work bound rather than run in full for each. +#[test] +fn listing_reads_a_shared_scale_once_and_bounds_boundary_checks() { + let measure = "<< /Type /Measure /Subtype /RL /R (1 pt = 1 pt) /X [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] /D [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] /A [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] >>"; + // A convex 2,048-gon: every one of its edges is checked against every other. + let n = MAX_POINTS; + let verts: Vec = (0..n) + .map(|i| { + let a = i as f64 / n as f64 * std::f64::consts::TAU; + format!("{:.3} {:.3}", 300.0 + 100.0 * a.cos(), 400.0 + 100.0 * a.sin()) + }) + .collect(); + let verts = verts.join(" "); + let areas = 40; + let annots: Vec = (0..areas).map(|i| format!("{} 0 R", 6 + i)).collect(); + let mut pdf = Pdf::new() + .obj("<< /Type /Catalog /Pages 2 0 R >>") + .obj("<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 612 792] >>") + .obj(format!("<< /Type /Page /Parent 2 0 R /Contents 4 0 R /Annots [{}] >>", annots.join(" "))) + .stream("", "") + .obj(measure); + for _ in 0..areas { + pdf = pdf.obj(format!("<< /Type /Annot /Subtype /Polygon /IT /PolygonDimension /Rect [0 0 612 792] /Vertices [{verts}] /Measure 5 0 R >>")); + } + let d = Document::open(Arc::new(pdf.build())).unwrap(); + let l = list(&d); + // 40 areas of 2,048 vertices are 168 million pairs: the listing stops at its bound, with the + // areas before it listed. + assert!(l.truncated, "{} listed", l.measurements.len()); + assert!((5..areas).contains(&l.measurements.len()), "{} listed", l.measurements.len()); + assert!(l.unsupported.is_empty(), "{:?}", l.unsupported); +} + +/// Perimeters and distances read their vertices once: only areas pay for the pairwise boundary +/// check, so many maximal perimeters list in full. +#[test] +fn perimeters_are_charged_linear_work() { + let measure = "<< /Type /Measure /Subtype /RL /R (1 pt = 1 pt) /X [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] /D [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] /A [<< /Type /NumberFormat /U (pt) /C 1 /D 100 >>] >>"; + let verts: Vec = (0..MAX_POINTS).map(|i| format!("{} {}", 10 + i % 500, 10 + i / 500 * 3 + (i % 2))).collect(); + let verts = verts.join(" "); + let lines = 40; + let annots: Vec = (0..lines).map(|i| format!("{} 0 R", 6 + i)).collect(); + let mut pdf = Pdf::new() + .obj("<< /Type /Catalog /Pages 2 0 R >>") + .obj("<< /Type /Pages /Kids [3 0 R] /Count 1 /MediaBox [0 0 612 792] >>") + .obj(format!("<< /Type /Page /Parent 2 0 R /Contents 4 0 R /Annots [{}] >>", annots.join(" "))) + .stream("", "") + .obj(measure); + for _ in 0..lines { + pdf = pdf.obj(format!("<< /Type /Annot /Subtype /PolyLine /IT /PolyLineDimension /Rect [0 0 612 792] /Vertices [{verts}] /Measure 5 0 R >>")); + } + let d = Document::open(Arc::new(pdf.build())).unwrap(); + let l = list(&d); + assert!(!l.truncated, "{} listed", l.measurements.len()); + assert_eq!(l.measurements.len() + l.unsupported.len(), lines, "{:?}", l.unsupported.first()); +} From a55eaa90e3f2f4aee400f4b95ced62cbfd24a762 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:40:35 -0500 Subject: [PATCH 21/75] fix(redact): prune the resources of Forms a surviving Type 3 glyph draws (#5924) --- .../redact/src/tests/surviving_references.rs | 44 ++++++++++++++ crates/redact/src/type3.rs | 57 ++++++++++++++++++- 2 files changed, 100 insertions(+), 1 deletion(-) diff --git a/crates/redact/src/tests/surviving_references.rs b/crates/redact/src/tests/surviving_references.rs index 477f03f9a5..c561b7f8d1 100644 --- a/crates/redact/src/tests/surviving_references.rs +++ b/crates/redact/src/tests/surviving_references.rs @@ -161,6 +161,30 @@ fn redacted_type3_glyph_and_exclusive_resources_do_not_survive_full_save() { assert!(t3_xo.contains(b"ImB"), "Type 3 font resources lost surviving image ImB"); } +/// The surviving glyph B draws a Form whose own resources also name A's image, which the Form +/// never draws. Pruning only the font's resources left that entry, so the image of the removed +/// glyph stayed in the file through the Form. +#[test] +fn a_redacted_type3_glyphs_image_does_not_survive_in_a_surviving_glyphs_form() { + let mut doc = pdf(vec![ + b"<< /Type /Catalog /Pages 2 0 R >>".to_vec(), + b"<< /Type /Pages /Kids [3 0 R] /Count 1 >>".to_vec(), + b"<< /Type /Page /Parent 2 0 R /MediaBox [0 0 300 300] /Contents 4 0 R /Resources 5 0 R >>".to_vec(), + stream("", b"BT /T3 10 Tf 10 100 Td (AB) Tj ET"), + b"<< /Font << /T3 6 0 R >> >>".to_vec(), + b"<< /Type /Font /Subtype /Type3 /FontBBox [0 -200 1000 800] /FontMatrix [0.001 0 0 0.001 0 0] /FirstChar 65 /LastChar 66 /Widths [500 500] /Encoding << /Differences [65 /A /B] >> /CharProcs << /A 7 0 R /B 8 0 R >> /Resources << /XObject << /ImA 9 0 R /Fm 10 0 R >> >> >>".to_vec(), + stream("", b"/ImA Do"), + stream("", b"/Fm Do"), + stream("/Type /XObject /Subtype /Image /Width 1 /Height 1 /ColorSpace /DeviceGray /BitsPerComponent 8", b"CONFIDENTIAL_IMAGE_A"), + stream("/Type /XObject /Subtype /Form /BBox [0 0 1000 1000] /Resources << /XObject << /ImA 9 0 R /ImB 11 0 R >> >>", b"/ImB Do"), + stream("/Type /XObject /Subtype /Image /Width 1 /Height 1 /ColorSpace /DeviceGray /BitsPerComponent 8", b"PUBLIC_IMAGE_B"), + ]); + mark(&mut doc, 0, &[[9.0, 95.0, 14.9, 110.0]], ""); + assert_eq!(apply(&mut doc, None).unwrap().glyphs, 1); + assert!(!full_save_holds(&doc, b"CONFIDENTIAL_IMAGE_A"), "the removed glyph's image survives through the surviving glyph's form"); + assert!(full_save_holds(&doc, b"PUBLIC_IMAGE_B"), "the image the surviving glyph's form draws was lost"); +} + /// When all glyphs of a Type 3 font are redacted, the entire font is removed from Resources (F-011). #[test] fn all_redacted_type3_glyphs_cause_font_removal_from_resources() { @@ -298,3 +322,23 @@ fn type3_font_used_in_a_form_xobject_has_its_redacted_glyph_removed() { assert_eq!(report.glyphs, 1); assert!(!full_save_holds(&doc, b"SECRET_FORM_GLYPH_PAYLOAD"), "redacted glyph program survived"); } + +/// A Form the surviving glyph draws holds `/Resources` as an array that references the removed +/// glyph's image: it can't be pruned, so the redaction refuses rather than keep the image. +#[test] +fn a_type3_glyphs_form_with_non_dictionary_resources_fails_closed() { + let mut doc = pdf(vec![ + b"<< /Type /Catalog /Pages 2 0 R >>".to_vec(), + b"<< /Type /Pages /Kids [3 0 R] /Count 1 >>".to_vec(), + b"<< /Type /Page /Parent 2 0 R /MediaBox [0 0 300 300] /Contents 4 0 R /Resources 5 0 R >>".to_vec(), + stream("", b"BT /T3 10 Tf 10 100 Td (AB) Tj ET"), + b"<< /Font << /T3 6 0 R >> >>".to_vec(), + b"<< /Type /Font /Subtype /Type3 /FontBBox [0 -200 1000 800] /FontMatrix [0.001 0 0 0.001 0 0] /FirstChar 65 /LastChar 66 /Widths [500 500] /Encoding << /Differences [65 /A /B] >> /CharProcs << /A 7 0 R /B 8 0 R >> /Resources << /XObject << /ImA 9 0 R /Fm 10 0 R >> >> >>".to_vec(), + stream("", b"/ImA Do"), + stream("", b"/Fm Do"), + stream("/Type /XObject /Subtype /Image /Width 1 /Height 1 /ColorSpace /DeviceGray /BitsPerComponent 8", b"CONFIDENTIAL_IMAGE_A"), + stream("/Type /XObject /Subtype /Form /BBox [0 0 1000 1000] /Resources [9 0 R]", b""), + ]); + mark(&mut doc, 0, &[[9.0, 95.0, 14.9, 110.0]], ""); + assert!(apply(&mut doc, None).is_err()); +} diff --git a/crates/redact/src/type3.rs b/crates/redact/src/type3.rs index 57971571af..3bba33a645 100644 --- a/crates/redact/src/type3.rs +++ b/crates/redact/src/type3.rs @@ -135,7 +135,13 @@ fn collect_used_resources_in_stream(data: &[u8], used: &mut UsedInGlyphs) -> Res if parsed.skipped > 0 { return Err(CosError::Syntax { offset: 0, detail: "redaction could not fully inspect glyph stream syntax".into() }); } - for op in parsed.ops { + parsed.ops.iter().for_each(|op| record_used(op, used)); + Ok(()) +} + +/// Record the resource name `op` uses, if any. +fn record_used(op: &pdfcraft_content::Op, used: &mut UsedInGlyphs) { + { match op.op.as_slice() { b"Do" => { if let Some(n) = op.last_name() { @@ -175,6 +181,54 @@ fn collect_used_resources_in_stream(data: &[u8], used: &mut UsedInGlyphs) -> Res _ => {} } } +} + +/// Prune the own `/Resources` of every Form a Type 3 font's (already pruned) resources still +/// list, and of the Forms those list in turn, to the names each Form's content uses. A Form a +/// surviving glyph draws may name, unused, an image only a removed glyph drew; that entry would +/// keep the image in the file. Pruning only unused entries is safe for a shared Form: no drawer +/// of it can use a name its content doesn't. Fails closed when a Form can't be read. +fn prune_font_forms(doc: &mut Document, font_res: &Dict) -> Result<(), CosError> { + let mut pending: Vec<(ObjRef, usize)> = res_dict(doc, font_res, b"XObject").iter().filter_map(|(_, v)| v.as_ref()).map(|r| (r, 0)).collect(); + let mut seen = HashSet::new(); + while let Some((form_ref, depth)) = pending.pop() { + if !seen.insert(form_ref) { + continue; + } + if depth > MAX_DEPTH || seen.len() > MAX_HOLDERS { + return Err(CosError::Syntax { offset: 0, detail: "redaction could not establish safe Type 3 font cleanup: forms nest too deep".into() }); + } + let Object::Stream(form) = &*doc.get(form_ref) else { continue }; + if form.dict.name(b"Subtype") != Some(b"Form") { + continue; + } + // A Form without resources of its own (no key, or `null`) draws through its caller's, + // which were pruned already. Any other value that isn't a dictionary may still reference + // what a removed glyph drew, and can't be pruned: fail closed. + let mut res = match form.dict.get(b"Resources").map(|r| doc.resolve(r)) { + None => continue, + Some(r) if matches!(*r, Object::Null) => continue, + Some(r) => match r.as_dict() { + Some(d) => d.clone(), + None => { + return Err(CosError::Syntax { + offset: 0, + detail: "redaction could not establish safe Type 3 font cleanup: a form's /Resources is not a dictionary".into(), + }); + } + }, + }; + let data = form.decoded()?; + let mut used = UsedInGlyphs::default(); + crate::interp::resource_ops(doc, &res, std::slice::from_ref(&data))?.iter().for_each(|op| record_used(op, &mut used)); + // `rg`, `g` and `k` draw through these entries when the Form keeps them as its defaults, + // and nothing cites them with `cs`/`CS`: pruning them would change the Form's colours. + used.colorspaces.extend([b"DefaultRGB".to_vec(), b"DefaultGray".to_vec(), b"DefaultCMYK".to_vec()]); + if prune_type3_font_resources(doc, &mut res, &used) { + doc.update_dict(form_ref, |d| d.set(b"Resources".to_vec(), Object::Dict(res.clone())))?; + } + pending.extend(res_dict(doc, &res, b"XObject").iter().filter_map(|(_, v)| v.as_ref()).map(|r| (r, depth + 1))); + } Ok(()) } @@ -631,6 +685,7 @@ fn clean_type3_font( } })?; } + prune_font_forms(doc, &f_res)?; Ok(()) } From ae21773fefa919340e1b910e4eca24264007ba6f Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:40:50 -0500 Subject: [PATCH 22/75] fix(edit): stop rescanning what a loop already holds in wrapping, grouping and flattening (#5963) --- crates/edit/src/added/layout.rs | 17 +++++++++++------ crates/edit/src/added/store.rs | 11 +++++++++-- crates/edit/src/flatten.rs | 17 +++++++++++++++-- crates/edit/src/tests.rs | 14 ++++++++++++++ crates/edit/src/text_regions/grouping.rs | 6 +++++- 5 files changed, 54 insertions(+), 11 deletions(-) diff --git a/crates/edit/src/added/layout.rs b/crates/edit/src/added/layout.rs index d1be9ef521..a6a55224d9 100644 --- a/crates/edit/src/added/layout.rs +++ b/crates/edit/src/added/layout.rs @@ -54,16 +54,21 @@ pub(super) fn wrapped(t: &AddedText) -> Vec<(String, bool, usize)> { let measure = |s: &str| if rtl_script || thai { shaped_width(t, s, rtl) } else { latin_width(t, s) }; let space = measure(" "); let (mut line, mut line_w) = (String::new(), 0.0); + // Widths add up (the standard-14 widths have no kerning), so each word is measured once + // and appended in place: a line that never fills (negative character spacing) stays + // linear instead of being copied and measured again for every word. for word in para.split(' ') { - let candidate = if line.is_empty() { word.to_string() } else { format!("{line} {word}") }; - let candidate_w = - if rtl_script || thai { if line.is_empty() { measure(word) } else { line_w + space + measure(word) } } else { measure(&candidate) }; + let word_w = measure(word); + let candidate_w = if line.is_empty() { word_w } else { line_w + space + word_w }; if !line.is_empty() && candidate_w > width { out.push((std::mem::take(&mut line), rtl, index)); - line = word.to_string(); - line_w = measure(word); + line.push_str(word); + line_w = word_w; } else { - line = candidate; + if !line.is_empty() { + line.push(' '); + } + line.push_str(word); line_w = candidate_w; } } diff --git a/crates/edit/src/added/store.rs b/crates/edit/src/added/store.rs index 3555910f04..a5b53581fc 100644 --- a/crates/edit/src/added/store.rs +++ b/crates/edit/src/added/store.rs @@ -328,8 +328,15 @@ pub fn isolate_content(doc: &mut Document, page: usize, index: usize) -> Result< /// names them so), so they go when the item is replaced or deleted. fn own_fonts(doc: &Document, obj: ObjRef) -> Vec> { let Object::Stream(s) = &*doc.get(obj) else { return Vec::new() }; - let Ok(data) = s.decoded() else { return Vec::new() }; - data.split(u8::is_ascii_whitespace).filter_map(|w| w.strip_prefix(b"/")).filter(|w| w.starts_with(b"PCAr")).map(<[u8]>::to_vec).collect() + // The stream is one this writer put there, so its decode is bounded by the stream itself: the + // page limit alone would refuse an item larger than a page's content, and its fonts would stay + // on the page's /Font, which `write` and `delete_content` are supposed to take them off. + let limit = pdfcraft_cos::MAX_PAGE_DECODED.max(s.raw.len().saturating_mul(16)); + let Ok(data) = s.decoded_within(limit) else { return Vec::new() }; + // Each name once: a stream that repeats a name is not copied out once per use. + let names: std::collections::BTreeSet<&[u8]> = + data.split(u8::is_ascii_whitespace).filter_map(|w| w.strip_prefix(b"/")).filter(|w| w.starts_with(b"PCAr")).collect(); + names.into_iter().map(<[u8]>::to_vec).collect() } fn drop_fonts(doc: &Document, res: &mut Dict, names: &[Vec]) { diff --git a/crates/edit/src/flatten.rs b/crates/edit/src/flatten.rs index ef86d6c67a..c54b062af6 100644 --- a/crates/edit/src/flatten.rs +++ b/crates/edit/src/flatten.rs @@ -183,6 +183,19 @@ fn flatten_which(doc: &mut Document, pages: &[usize], which: Which) -> Result pdfcraft_model::Page { + let mut dict = page.dict.clone(); + if let Some(own) = doc.get(page.obj).as_dict() { + for (k, v) in own.iter() { + dict.set(k.clone(), v.clone()); + } + } + pdfcraft_model::Page { obj: page.obj, dict } +} + fn flatten_in(doc: &mut Document, pages: &[usize], which: Which) -> Result { let all = page_list(doc); check(pages, all.len())?; @@ -322,10 +335,10 @@ fn flatten_in(doc: &mut Document, pages: &[usize], which: Which) -> Result doc.set(r, Object::Array(kept)), None => doc.update_dict(page.obj, |d| { diff --git a/crates/edit/src/tests.rs b/crates/edit/src/tests.rs index 1c4829b418..700117ac6f 100644 --- a/crates/edit/src/tests.rs +++ b/crates/edit/src/tests.rs @@ -9748,6 +9748,20 @@ fn added_text_is_wrapped_once_per_write() { assert!(calls <= 4, "the text was wrapped {calls} times for 40 lines"); } +/// Negative character spacing keeps a line from ever filling its box. Wrapping then holds every +/// word on one line, appended and measured once each rather than copied and measured again for +/// each word (which took minutes for this text); the widths still decide the breaks. +#[test] +fn added_text_that_never_fills_its_line_wraps_in_linear_time() { + let words = vec!["word"; 200_000].join(" "); + let packed = AddedText { rect: [0.0, 0.0, 100.0, 100.0], text: words.clone(), char_spacing: -10.0, ..AddedText::default() }; + let start = std::time::Instant::now(); + assert_eq!(added::lines(&packed), [words]); + assert!(start.elapsed() < std::time::Duration::from_secs(10), "{:?}", start.elapsed()); + let plain = AddedText { rect: [0.0, 0.0, 60.0, 100.0], text: "one two three four".into(), size: 10.0, ..AddedText::default() }; + assert_eq!(added::lines(&plain), ["one two", "three four"]); +} + /// Character spacing on added text (#4245) is the PDF's `Tc`, kept in `/PCAdded` so the item /// reopens with it; an item written without it reads 0. #[test] diff --git a/crates/edit/src/text_regions/grouping.rs b/crates/edit/src/text_regions/grouping.rs index 69b3a16850..1dcb9a9373 100644 --- a/crates/edit/src/text_regions/grouping.rs +++ b/crates/edit/src/text_regions/grouping.rs @@ -366,10 +366,13 @@ fn group_layout(lines: &[TextLine], mixed: bool) -> Vec { if !block.rows.is_empty() { block.editor_text.push('\n'); } + // The run placed before this one in the row, tracked rather than searched for: rows + // don't share runs, so it is the last one this loop pushed. + let mut prev_in_row: Option<&TextLine> = None; for &index in row { let Some(line) = lines.get(index) else { continue }; // Some producers position separate words without encoding a space. - if let Some(prev) = block.lines.last().filter(|i| row.contains(i)).and_then(|i| lines.get(*i)) + if let Some(prev) = prev_in_row && !row_text.ends_with(char::is_whitespace) && !line.text.starts_with(char::is_whitespace) && line.frame_rect()[0] - prev.frame_rect()[2] > line.size * 0.12 @@ -383,6 +386,7 @@ fn group_layout(lines: &[TextLine], mixed: bool) -> Vec { row_text.push_str(&line.text); block.invisible &= line.origin.state.render_mode == 3; block.lines.push(index); + prev_in_row = Some(line); } if !block.text.is_empty() { if block.text.ends_with('-') { From afaaaa641540848e559e430c63e7291deafaf2b3 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:41:04 -0500 Subject: [PATCH 23/75] fix(print): index page labels once and cap how many pages a range selects (#5967) --- crates/print/src/error.rs | 13 +++++- crates/print/src/range.rs | 92 ++++++++++++++++++++++++++++++++++++--- crates/print/src/tests.rs | 2 +- 3 files changed, 97 insertions(+), 10 deletions(-) diff --git a/crates/print/src/error.rs b/crates/print/src/error.rs index 4b1fe95f21..6f87c21c29 100644 --- a/crates/print/src/error.rs +++ b/crates/print/src/error.rs @@ -14,6 +14,9 @@ pub enum PrintError { /// A range token that is neither a page number nor a page label. #[error("{0:?} is not a page number or label")] NotAPage(String), + /// A range that names more pages than a listing will take, with the limit. + #[error("the range selects more than {0} pages")] + TooManySelected(usize), #[error("{0}")] Invalid(String), #[error("{0}")] @@ -33,9 +36,14 @@ pub fn in_english(wording: &str) -> String { /// What a person can get wrong about what to print, in English, with their holes still in them. /// -/// An interface that reports these in another language translates these three, which is the whole +/// An interface that reports these in another language translates these four, which is the whole /// reason they are gathered here; `wordings_are_said` keeps them in step with the errors. -pub const WORDINGS: &[&str] = &["there are no pages to print", "page {n} is out of range (1–{count})", "{page} is not a page number or label"]; +pub const WORDINGS: &[&str] = &[ + "there are no pages to print", + "page {n} is out of range (1–{count})", + "{page} is not a page number or label", + "the range selects more than {limit} pages", +]; impl PrintError { /// This error as one sentence, said in the language `words` translates into. @@ -51,6 +59,7 @@ impl PrintError { } // The token goes in quoted, the way it is typed, as the error itself prints it. PrintError::NotAPage(token) => say(&words("{page} is not a page number or label"), &[("{page}", format!("{token:?}"))]), + PrintError::TooManySelected(limit) => say(&words("the range selects more than {limit} pages"), &[("{limit}", limit.to_string())]), fault => fault.to_string(), } } diff --git a/crates/print/src/range.rs b/crates/print/src/range.rs index f0d8939da6..fffab083af 100644 --- a/crates/print/src/range.rs +++ b/crates/print/src/range.rs @@ -1,7 +1,41 @@ //! Pages to Print: all, a range ("1-3, 6, 9-", page labels allowed), odd or even pages, reverse. +use std::collections::HashMap; + use crate::PrintError; +/// Most pages one range may select (more when the document itself has more): every selection +/// becomes a placement on a sheet, and repeating `1-` costs a few bytes of input for each copy of +/// the whole document. +pub const MAX_SELECTED: usize = 100_000; + +/// The first page with each label, looked up once per token instead of scanning every label. +struct Labels<'a> { + pages: HashMap<&'a str, usize>, + /// The longest label, in bytes: a token longer than that is no label, and is not hashed (a + /// dash-heavy token would otherwise hash every split of itself). + longest: usize, +} + +impl<'a> Labels<'a> { + fn new(labels: &'a [String]) -> Self { + let mut pages = HashMap::with_capacity(labels.len()); + for (i, l) in labels.iter().enumerate() { + pages.entry(l.as_str()).or_insert(i); + } + let longest = pages.keys().map(|l| l.len()).max().unwrap_or(0); + Labels { pages, longest } + } + + fn get(&self, token: &str) -> Option { + if token.len() > self.longest { None } else { self.pages.get(token).copied() } + } + + fn contains_key(&self, token: &str) -> bool { + self.get(token).is_some() + } +} + #[derive(Clone, Copy, Debug, PartialEq, Eq, Default)] pub enum Subset { #[default] @@ -11,10 +45,10 @@ pub enum Subset { } /// Resolve one range token: a page label, or a number (1-based). -fn page_of(tok: &str, count: usize, labels: &[String]) -> Result { +fn page_of(tok: &str, count: usize, labels: &Labels) -> Result { let t = tok.trim(); // Page labels win over numbers (Acrobat's "use logical page numbers"). - if let Some(i) = labels.iter().position(|l| l == t) { + if let Some(i) = labels.get(t) { return Ok(i); } match t.parse::() { @@ -27,12 +61,12 @@ fn page_of(tok: &str, count: usize, labels: &[String]) -> Result Option<(usize, usize)> { +fn hyphenated_label_range(part: &str, count: usize, labels: &Labels) -> Option<(usize, usize)> { for (split, _) in part.match_indices('-') { let (Some(first), Some(last)) = (part.get(..split), part.get(split + 1..)) else { continue }; let (first, last) = (first.trim(), last.trim()); // Avoid changing the meaning of ordinary numeric and open-ended ranges. - if !labels.iter().any(|l| l.contains('-') && (l == first || l == last)) { + if ![first, last].iter().any(|l| labels.contains_key(l) && l.contains('-')) { continue; } let from = if first.is_empty() { Some(0) } else { page_of(first, count, labels).ok() }; @@ -52,12 +86,22 @@ pub fn select_pages(count: usize, range: Option<&str>, labels: &[String], subset match range.map(str::trim).filter(|r| !r.is_empty()) { None => pages.extend(0..count), Some(r) => { + let index = Labels::new(labels); + let labels = &index; + let limit = count.max(MAX_SELECTED); + let mut selected = 0usize; + // Checked before a span is expanded, so the list never grows past the limit. + let mut take = |from: usize, to: usize| { + selected = selected.saturating_add(from.abs_diff(to).saturating_add(1)); + if selected > limit { Err(PrintError::TooManySelected(limit)) } else { Ok(()) } + }; for part in r.split([',', ';']).map(str::trim).filter(|p| !p.is_empty()) { // A complete label (e.g. "A-1") still wins over interpreting it as a range. // A range of such labels (e.g. "A-1-A-3") needs a later '-' split. - if !labels.iter().any(|l| l == part) + if !labels.contains_key(part) && let Some((from, to)) = hyphenated_label_range(part, count, labels) { + take(from, to)?; if from <= to { pages.extend(from..=to); } else { @@ -67,17 +111,24 @@ pub fn select_pages(count: usize, range: Option<&str>, labels: &[String], subset } match part.split_once('-') { // A label may itself contain a dash ("A-1"): try the whole token first. - Some(_) if labels.iter().any(|l| l == part) => pages.push(page_of(part, count, labels)?), + Some(_) if labels.contains_key(part) => { + take(0, 0)?; + pages.push(page_of(part, count, labels)?); + } Some((a, b)) => { let from = if a.trim().is_empty() { 0 } else { page_of(a, count, labels)? }; let to = if b.trim().is_empty() { count.checked_sub(1).ok_or(PrintError::NoPages)? } else { page_of(b, count, labels)? }; + take(from, to)?; if from <= to { pages.extend(from..=to); } else { pages.extend((to..=from).rev()); } } - None => pages.push(page_of(part, count, labels)?), + None => { + take(0, 0)?; + pages.push(page_of(part, count, labels)?); + } } } } @@ -129,4 +180,31 @@ mod tests { fn open_ended_range_covers_all_pages() { assert_eq!(select_pages(5, Some("-"), &[], Subset::All, false).unwrap(), vec![0, 1, 2, 3, 4]); } + + /// Repeating a range is allowed up to the limit, and refused past it before the pages are + /// listed: 1,000 pages selected 100 times is the limit, once more is not. + #[test] + fn a_range_selects_at_most_the_limit() { + let whole = vec!["1-"; MAX_SELECTED / 1000].join(","); + assert_eq!(select_pages(1000, Some(&whole), &[], Subset::All, false).unwrap().len(), MAX_SELECTED); + let e = select_pages(1000, Some(&format!("{whole},1")), &[], Subset::All, false).unwrap_err(); + assert!(e.to_string().contains("more than 100000 pages"), "{e}"); + // A document longer than the limit can still be printed whole, once. + assert_eq!(select_pages(MAX_SELECTED + 1, Some("1-"), &[], Subset::All, false).unwrap().len(), MAX_SELECTED + 1); + } + + /// Labels resolve through one index: the first page with a label wins, and a hyphenated + /// label works as an endpoint, among many labels and many dashes. + #[test] + fn labels_resolve_through_the_index() { + let labels: Vec = (0..50_000).map(|i| format!("A-{i}")).collect(); + assert_eq!(select_pages(labels.len(), Some("A-3-A-5"), &labels, Subset::All, false).unwrap(), vec![3, 4, 5]); + let dashes = "-".repeat(10_000); + assert!(select_pages(labels.len(), Some(&dashes), &labels, Subset::All, false).is_err()); + // a token with no label as long as any of its halves costs no hashing of them + let long_token = format!("{}A-3", "-".repeat(200_000)); + assert!(select_pages(labels.len(), Some(&long_token), &labels, Subset::All, false).is_err()); + let twice = vec!["x".to_string(), "x".to_string()]; + assert_eq!(select_pages(2, Some("x"), &twice, Subset::All, false).unwrap(), vec![0]); + } } diff --git a/crates/print/src/tests.rs b/crates/print/src/tests.rs index 642af0b8c9..f51e505251 100644 --- a/crates/print/src/tests.rs +++ b/crates/print/src/tests.rs @@ -84,7 +84,7 @@ fn page_selection() { /// and the numbers apart from it, and says itself in English exactly as the error prints itself. #[test] fn wordings_are_said() { - let mistakes = [PrintError::NoPages, PrintError::OutOfRange(7, 5), PrintError::NotAPage("x".into())]; + let mistakes = [PrintError::NoPages, PrintError::OutOfRange(7, 5), PrintError::NotAPage("x".into()), PrintError::TooManySelected(1_000)]; for e in &mistakes { assert_eq!(e.said(&in_english), e.to_string(), "{e:?} says itself in English the way it prints itself"); } From ee6ae8eac4da186b9551818a8d6cbf2a8cd5b546 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:41:24 -0500 Subject: [PATCH 24/75] fix(engine): coalesce script resetForm requests and bound console output per edit (#5975) --- crates/engine/src/js.rs | 121 +++++++++++++++++++++++++++++++++++----- 1 file changed, 108 insertions(+), 13 deletions(-) diff --git a/crates/engine/src/js.rs b/crates/engine/src/js.rs index f581504204..7f0561d196 100644 --- a/crates/engine/src/js.rs +++ b/crates/engine/src/js.rs @@ -26,29 +26,70 @@ pub struct JsOutput { pub doc_errors: Vec, } +/// Console lines one output keeps, from every script an edit runs together (a form's +/// calculations, say): each run's own cap ([`pdfcraft_js::xfa::MAX_CONSOLE`]) would otherwise +/// be multiplied by the number of fields a document declares. +pub const MAX_CONSOLE_LINES: usize = 10_000; +/// Bytes of console text one output keeps, for the same reason. +pub const MAX_CONSOLE_BYTES: usize = 1 << 20; +/// Alerts one output keeps. +pub const MAX_ALERTS: usize = 1_000; + impl JsOutput { /// Whether scripts produced nothing: no alerts, console lines, requests or errors. pub fn is_empty(&self) -> bool { self.alerts.is_empty() && self.console.is_empty() && self.requests.is_empty() && self.errors.is_empty() } + /// Adds `lines` to the console within [`MAX_CONSOLE_LINES`] and [`MAX_CONSOLE_BYTES`]; what + /// is past them is left out, which one closing line says. + fn say(&mut self, lines: impl IntoIterator) { + let note = format!("more console lines were left out: the scripts printed more than {MAX_CONSOLE_LINES} lines or {MAX_CONSOLE_BYTES} bytes"); + // Once the notice is written the console is closed: nothing may follow it, and a later + // rejection must not add a second copy. + if self.console.last() == Some(¬e) { + return; + } + let mut bytes: usize = self.console.iter().map(String::len).sum(); + let mut dropped = false; + for l in lines { + if self.console.len() >= MAX_CONSOLE_LINES || bytes.saturating_add(l.len()) > MAX_CONSOLE_BYTES { + dropped = true; + break; + } + bytes += l.len(); + self.console.push(l); + } + if dropped && !self.console.iter().any(|l| l == ¬e) { + self.console.push(note); + } + } + + /// Adds `alerts` within [`MAX_ALERTS`]. + fn alert(&mut self, alerts: impl IntoIterator) { + let room = MAX_ALERTS.saturating_sub(self.alerts.len()); + self.alerts.extend(alerts.into_iter().take(room)); + } + /// Adds what a field's own script produced: its alerts, console lines, errors and requests, /// except a request to show a side panel or open Save As, which is noted on the console /// and dropped. fn absorb(&mut self, o: &Outcome) { - self.alerts.extend(o.alerts.iter().cloned()); - self.console.extend(o.console.iter().cloned()); + self.alert(o.alerts.iter().cloned()); + self.say(o.console.iter().cloned()); // A field's own scripts (keystroke, format, validate, calculate) run unasked: only a // button's may show or hide a side panel, or open Save As. + // The warnings go through `say` like any other console line, so they count against the + // line and byte caps instead of growing the console past them. + let mut notes = Vec::new(); for r in &o.requests { match r { - Request::ShowPanel(_) => self.console.push("A field script asked to show a side panel; only a button may show a side panel".into()), - Request::SaveAs => { - self.console.push("A field script asked to save the document as a new file; only a button may open Save As".into()) - } + Request::ShowPanel(_) => notes.push("A field script asked to show a side panel; only a button may show a side panel".to_string()), + Request::SaveAs => notes.push("A field script asked to save the document as a new file; only a button may open Save As".to_string()), r => self.requests.push(r.clone()), } } + self.say(notes); self.errors.extend(o.error.iter().cloned()); // The document scripts run before every field script, so a failing one would repeat // once per calculation; it is reported once. @@ -61,8 +102,8 @@ impl JsOutput { } pub(crate) fn append(&mut self, other: JsOutput) { - self.alerts.extend(other.alerts); - self.console.extend(other.console); + self.alert(other.alerts); + self.say(other.console); self.requests.extend(other.requests); // As in `absorb`: a document script's error, already reported, is not repeated; a field // script's errors are each their own, even one worded like a document script's. @@ -330,11 +371,9 @@ impl Session { if !changes.is_empty() { edits.push(Edit::ApplyScriptChanges { changes }); } - for names in resets { - let all: Vec = fields.iter().map(|f| f.name.clone()).collect(); - let names = (!names.is_empty()) - .then(|| all.iter().filter(|n| names.iter().any(|x| *n == x || n.starts_with(&format!("{x}.")))).cloned().collect()); - edits.push(Edit::ResetForm { names }); + if !resets.is_empty() { + let all: Vec<&str> = fields.iter().map(|f| f.name.as_str()).collect(); + edits.push(Edit::ResetForm { names: reset_targets(&all, &resets) }); } if !edits.is_empty() { self.apply(id, Edit::Batch { label: "Run JavaScript".into(), edits })?; @@ -343,6 +382,21 @@ impl Session { } } +/// The fields every `resetForm` call of one script asked for, as one reset: a reset is to the +/// defaults, so resetting the union once does what each call in turn would, and recalculates +/// once. `None` is the whole form (a call with no names). A field is named by a selector that +/// is its name or the name of a field above it, looked up in a set of the selectors, so the +/// work is the fields times their depth, not the fields times the selectors, which a script +/// controls. +fn reset_targets(fields: &[&str], requests: &[Vec]) -> Option> { + if requests.iter().any(Vec::is_empty) { + return None; + } + let wanted: std::collections::HashSet<&str> = requests.iter().flatten().map(String::as_str).collect(); + let matched = |name: &str| name.match_indices('.').map(|(i, _)| &name[..i]).chain(std::iter::once(name)).any(|p| wanted.contains(p)); + Some(fields.iter().filter(|n| matched(n)).map(|n| n.to_string()).collect()) +} + /// A page's words for form-field detection, in user space (underscore runs split from the /// text around them). fn detection_words(text: &pdfcraft_render::PageText, info: &pdfcraft_render::PageInfo) -> Vec { @@ -492,6 +546,47 @@ impl Session { mod review_tests { use super::*; + /// Many `resetForm` calls with many selectors over a big form: one reset of the union, found + /// by looking each field's name and parents up, not by comparing every field with every + /// selector. + #[test] + fn reset_requests_are_coalesced_and_matched_through_an_index() { + let fields: Vec = (0..20_000).map(|i| format!("group{}.row{i}.cell", i % 100)).collect(); + let names: Vec<&str> = fields.iter().map(String::as_str).collect(); + let requests: Vec> = (0..1_000).map(|k| (0..20).map(|j| format!("nothing{}", k * 20 + j)).collect()).collect(); + let started = std::time::Instant::now(); + assert_eq!(reset_targets(&names, &requests), Some(Vec::new()), "no selector names a field"); + let mut requests = requests; + requests.push(vec!["group7".into(), "group7.row7.cell".into(), "group8.row8".into()]); + let hit = reset_targets(&names, &requests).unwrap(); + assert_eq!(hit.len(), 200 + 1, "group7 has 200 fields, one more by its exact name"); + assert!(hit.iter().all(|n| n.starts_with("group7.") || n == "group8.row8.cell")); + assert!(started.elapsed() < std::time::Duration::from_secs(5), "{:?}", started.elapsed()); + requests.push(Vec::new()); + assert_eq!(reset_targets(&names, &requests), None, "a call without names resets the whole form"); + } + + /// A form of many calculated fields, each printing its own cap of console lines, keeps a + /// bounded console: past the budget, lines are left out and one closing line says so. + #[test] + fn console_output_of_one_edit_is_bounded() { + let mut out = JsOutput::default(); + let line = "x".repeat(4_096); + let run = Outcome { console: vec![line; 1_000], alerts: vec![Message::plain("hi"); 100], ..Default::default() }; + for _ in 0..1_000 { + out.absorb(&run); + } + let bytes: usize = out.console.iter().map(String::len).sum(); + assert!(bytes <= MAX_CONSOLE_BYTES + 256, "{bytes} bytes kept"); + assert!(out.console.len() <= MAX_CONSOLE_LINES + 1, "{} lines kept", out.console.len()); + assert!(out.console.last().unwrap().starts_with("more console lines were left out")); + assert_eq!(out.console.iter().filter(|l| l.starts_with("more console lines")).count(), 1, "said once"); + assert_eq!(out.alerts.len(), MAX_ALERTS); + let mut merged = JsOutput::default(); + merged.append(out); + assert!(merged.console.len() <= MAX_CONSOLE_LINES + 1); + } + /// A long run of underscores (a form's blank) is read in one pass: each glyph is compared with /// the word so far without reading the word again. #[test] From 85cd33e9c606a2ed777d2727612d28df8882f419 Mon Sep 17 00:00:00 2001 From: Harrison Booth <73434116+Harrison-Booth@users.noreply.github.com> Date: Sun, 8 Nov 2026 18:41:38 -0600 Subject: [PATCH 25/75] Harden Bates sequence validation (#6003) --- crates/automation/src/bates.rs | 264 ++++++++++++++++++++--- crates/automation/src/file_access.rs | 5 +- crates/automation/src/tools.rs | 4 +- crates/automation/tests/it/automation.rs | 34 ++- 4 files changed, 271 insertions(+), 36 deletions(-) diff --git a/crates/automation/src/bates.rs b/crates/automation/src/bates.rs index 5267716e85..ffb2d42723 100644 --- a/crates/automation/src/bates.rs +++ b/crates/automation/src/bates.rs @@ -8,6 +8,8 @@ use super::json::Args; use super::paths::read_input; use super::{Arc, Automation, Path, PathBuf}; +const MAX_BATES_BYTES: u64 = 512 * 1024 * 1024; + struct FilePlan { input: String, source: PathBuf, @@ -26,19 +28,23 @@ impl Automation { } let start = match args.0.get("start_number") { None | Some(Value::Null) => 1, - Some(value) => value - .as_u64() - .filter(|n| *n >= 1) - .ok_or_else(|| ToolError::InvalidArgs("start_number must be a whole number of at least 1".into()))?, + Some(value) => { + value.as_u64().filter(|n| *n >= 1).ok_or_else(|| ToolError::InvalidArgs("start_number must be an integer of at least 1".into()))? + } }; let digits = args.opt_int("digits")?.unwrap_or(6); let digits = u8::try_from(digits).ok().filter(|d| (1..=12).contains(d)).ok_or_else(|| ToolError::InvalidArgs("digits must be from 1 to 12".into()))?; let prefix = args.opt_str("prefix")?.unwrap_or(""); let suffix = args.opt_str("suffix")?.unwrap_or(""); - // `#` separates the token's fields and `>>` ends it: either would cut the token short. - if prefix.len() > 128 || suffix.len() > 128 || [prefix, suffix].iter().any(|s| s.contains('#') || s.contains(">>")) { - return Err(ToolError::InvalidArgs("prefix and suffix must be at most 128 bytes and cannot contain # or >>".into())); + if prefix.chars().count() > 128 + || suffix.chars().count() > 128 + || prefix.contains('#') + || suffix.contains('#') + || prefix.contains(">>") + || suffix.contains(">>") + { + return Err(ToolError::InvalidArgs("prefix and suffix must be at most 128 characters and cannot contain # or >>".into())); } let dry_run = args.opt_bool("dry_run")?.unwrap_or(false); @@ -70,16 +76,7 @@ impl Automation { targets.push((output.to_string(), target)); passwords.push(entry.get("password").and_then(Value::as_str).map(str::to_owned)); } - let input_paths: std::collections::HashSet<_> = sources.iter().map(|(_, path)| path.clone()).collect(); - let mut output_paths = std::collections::HashSet::new(); - for (name, path) in &targets { - if input_paths.contains(path) { - return Err(failed(format!("output {name} names an input file; choose a separate output path"))); - } - if !output_paths.insert(path.clone()) { - return Err(failed(format!("two files use the same output path: {name}"))); - } - } + check_unique_outputs(&targets, dry_run)?; let mut plans = Vec::with_capacity(files.len()); let mut next = start; @@ -88,13 +85,13 @@ impl Automation { let size = std::fs::metadata(input_path).map_err(failed)?.len(); input_bytes = input_bytes .checked_add(size) - .filter(|total| *total <= 512 * 1024 * 1024) + .filter(|total| *total <= MAX_BATES_BYTES) .ok_or_else(|| failed("combined Bates input exceeds the 512 MiB safety limit"))?; let bytes = read_input(input_path)?; input_bytes = input_bytes .saturating_sub(size) .checked_add(u64::try_from(bytes.len()).unwrap_or(u64::MAX)) - .filter(|total| *total <= 512 * 1024 * 1024) + .filter(|total| *total <= MAX_BATES_BYTES) .ok_or_else(|| failed("combined Bates input exceeds the 512 MiB safety limit"))?; let display_name = input_path.file_name().map(|n| n.to_string_lossy().into_owned()).unwrap_or_default(); let password = passwords.get(index).and_then(Option::as_deref); @@ -106,7 +103,8 @@ impl Automation { return Err(failed(format!("{input_name} has no pages to number"))); } let count = u64::try_from(pages).map_err(|_| failed(format!("page count is too large for {input_name}")))?; - let last = next.checked_add(count.saturating_sub(1)).ok_or_else(|| failed(format!("Bates number range overflows for {input_name}")))?; + let last_offset = count.checked_sub(1).ok_or_else(|| failed(format!("{input_name} has no pages to number")))?; + let last = next.checked_add(last_offset).ok_or_else(|| failed(format!("Bates number range overflows for {input_name}")))?; plans.push(FilePlan { input: input_name.clone(), source: input_path.clone(), @@ -116,7 +114,7 @@ impl Automation { first: next, last, }); - if index + 1 < sources.len() { + if index + 1 < files.len() { next = last.checked_add(1).ok_or_else(|| failed(format!("Bates sequence overflows after {input_name}")))?; } } @@ -163,12 +161,14 @@ impl Automation { let result = applied.and_then(|()| self.session.save_full_bytes(id)); match result { Ok(bytes) => { - let total = output_bytes.saturating_add(u64::try_from(bytes.len()).unwrap_or(u64::MAX)); - if total > 512 * 1024 * 1024 { - results.push(json!({"input":plan.input,"output":plan.output,"status":"error","pages":plan.pages,"first":plan.first,"last":plan.last,"error":"combined Bates output exceeds the 512 MiB safety limit"})); - } else { + let candidate = output_bytes + .checked_add(u64::try_from(bytes.len()).unwrap_or(u64::MAX)) + .filter(|total| *total <= MAX_BATES_BYTES); + if let Some(total) = candidate { output_bytes = total; staged.push((plan.output.clone(), bytes)); + } else { + results.push(json!({"input":plan.input,"output":plan.output,"status":"error","pages":plan.pages,"first":plan.first,"last":plan.last,"error":"combined Bates output exceeds the 512 MiB safety limit"})); } } Err(error) => results.push(json!({"input":plan.input,"output":plan.output,"status":"error","pages":plan.pages,"first":plan.first,"last":plan.last,"error":error.to_string()})), @@ -198,6 +198,123 @@ impl Automation { } } +fn check_unique_outputs(targets: &[(String, PathBuf)], dry_run: bool) -> Result<()> { + if targets.len() < 2 { + return Ok(()); + } + + // Probe exact basenames in a directory under each canonical parent, so the filesystem applies its own name rules. + let mut groups: std::collections::HashMap)> = std::collections::HashMap::new(); + for (name, target) in targets { + let parent = target.parent().ok_or_else(|| failed(format!("output has no parent directory: {name}")))?; + let identity = parent.canonicalize().map_err(failed)?; + let probe_name = output_probe_name(target).map_err(failed)?; + groups.entry(identity.clone()).or_insert_with(|| (identity, Vec::new())).1.push((name.clone(), probe_name)); + } + + if dry_run { + // Dry runs cannot probe the volume: Windows uses a conservative case-insensitive comparison, + // while Unix preserves case, so unusual case-sensitive Windows or case-insensitive Unix volumes can differ. + for (_, (_, names)) in groups { + for (index, (name, candidate)) in names.iter().enumerate() { + if names.iter().take(index).any(|(_, prior)| same_output_name(prior, candidate)) { + return Err(failed(format!("two files use the same output path: {name}"))); + } + } + } + return Ok(()); + } + + for (_, (parent, names)) in groups { + match tempfile::tempdir_in(&parent) { + Ok(probes) => { + for (name, probe_name) in names { + let probe = probes.path().join(probe_name); + match std::fs::OpenOptions::new().write(true).create_new(true).open(&probe) { + Ok(file) => drop(file), + Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => { + return Err(failed(format!("two files use the same output path: {name}"))); + } + Err(error) => return Err(failed(format!("could not validate output path {name}: {error}"))), + } + } + } + #[cfg(windows)] + Err(error) if error.kind() == std::io::ErrorKind::PermissionDenied => probe_files_for_resolved_outputs(&parent, names)?, + Err(error) => return Err(failed(error)), + } + } + Ok(()) +} + +#[cfg(windows)] +fn probe_files_for_resolved_outputs(parent: &Path, names: Vec<(String, std::ffi::OsString)>) -> Result<()> { + use std::os::windows::fs::OpenOptionsExt; + + const FILE_FLAG_DELETE_ON_CLOSE: u32 = 0x0400_0000; + let mut probes = Vec::with_capacity(names.len()); + for (name, probe_name) in names { + let probe = parent.join(probe_name); + let mut options = std::fs::OpenOptions::new(); + options.write(true).create_new(true).share_mode(0).custom_flags(FILE_FLAG_DELETE_ON_CLOSE); + match options.open(&probe) { + Ok(file) => probes.push(file), + Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => { + return Err(failed(format!("two files use the same output path: {name}"))); + } + Err(error) => return Err(failed(format!("could not validate output path {name}: {error}"))), + } + } + drop(probes); + Ok(()) +} + +fn same_output_name(left: &std::ffi::OsStr, right: &std::ffi::OsStr) -> bool { + #[cfg(windows)] + { + use std::os::windows::ffi::OsStrExt; + fn key(name: &std::ffi::OsStr) -> Vec { + std::char::decode_utf16(name.encode_wide()) + .map(|decoded| match decoded { + Ok(character) => { + let mut uppercase = character.to_uppercase(); + match (uppercase.next(), uppercase.next()) { + (Some(single), None) => single as u32, + _ => character as u32, + } + } + Err(error) => 0x11_0000 + u32::from(error.unpaired_surrogate()), + }) + .collect() + } + return key(left) == key(right); + } + #[cfg(not(windows))] + { + left == right + } +} + +fn output_probe_name(path: &Path) -> std::io::Result { + let name = path.file_name().ok_or_else(|| std::io::Error::new(std::io::ErrorKind::InvalidInput, "output has no file name"))?; + #[cfg(windows)] + { + use std::os::windows::ffi::{OsStrExt, OsStringExt}; + let is_verbatim = path.components().any(|component| matches!(component, std::path::Component::Prefix(prefix) if matches!(prefix.kind(), std::path::Prefix::Verbatim(_) | std::path::Prefix::VerbatimUNC(_, _) | std::path::Prefix::VerbatimDisk(_)))); + let mut wide: Vec = name.encode_wide().collect(); + if !is_verbatim { + while wide.last().is_some_and(|unit| *unit == u16::from(b'.') || *unit == u16::from(b' ')) { + wide.pop(); + } + } + return Ok(std::ffi::OsString::from_wide(&wide)); + } + #[cfg(not(windows))] + { + Ok(name.to_os_string()) + } +} + /// Publish a PDF atomically while refusing a target created by another process after preflight. pub(super) fn publish_new(path: &Path, bytes: &[u8]) -> Result<()> { let parent = path.parent().ok_or_else(|| failed("output has no parent directory"))?; @@ -212,3 +329,100 @@ pub(super) fn publish_new(path: &Path, bytes: &[u8]) -> Result<()> { } }) } + +#[cfg(test)] +mod tests { + use super::check_unique_outputs; + + #[test] + fn dry_run_path_check_creates_no_probe_files() { + let dir = tempfile::tempdir().unwrap(); + let path = dir.path().join("same.pdf"); + let targets = vec![("first.pdf".into(), path.clone()), ("second.pdf".into(), path)]; + assert!(check_unique_outputs(&targets, true).is_err()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 0); + } + + #[cfg(unix)] + #[test] + fn dry_run_detects_output_parent_aliases_without_writing() { + let dir = tempfile::tempdir().unwrap(); + let alias = dir.path().join("alias"); + std::os::unix::fs::symlink(dir.path(), &alias).unwrap(); + let first = dir.path().join("same.pdf"); + let second = alias.join("same.pdf"); + let targets = vec![("first.pdf".into(), first), ("second.pdf".into(), second)]; + assert!(check_unique_outputs(&targets, true).is_err()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 1); + } + + #[test] + fn long_output_names_fit_in_probe_directory() { + let dir = tempfile::tempdir().unwrap(); + let first = dir.path().join(format!("{}.pdf", "a".repeat(220))); + let second = dir.path().join(format!("{}.pdf", "b".repeat(220))); + let targets = vec![("first.pdf".into(), first), ("second.pdf".into(), second)]; + assert!(check_unique_outputs(&targets, false).is_ok()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 0); + } + + #[cfg(unix)] + #[test] + fn output_probe_paths_preserve_non_utf8_names() { + use std::os::unix::ffi::OsStringExt; + let dir = tempfile::tempdir().unwrap(); + let first = dir.path().join(std::ffi::OsString::from_vec(vec![b'o', 0xff])); + let second = dir.path().join(std::ffi::OsString::from_vec(vec![b'o', 0xfe])); + let targets = vec![("first.pdf".into(), first), ("second.pdf".into(), second)]; + assert!(check_unique_outputs(&targets, false).is_ok()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 0); + } + + #[cfg(windows)] + #[test] + fn windows_probe_names_trim_only_non_verbatim_paths() { + use std::ffi::OsString; + use std::path::Path; + + let normal = Path::new(r"C:\root\report.pdf. "); + assert_eq!(super::output_probe_name(normal).unwrap(), OsString::from("report.pdf")); + + let verbatim = Path::new(r"\\?\C:\root\report.pdf. "); + assert_eq!(super::output_probe_name(verbatim).unwrap(), OsString::from("report.pdf. ")); + } + + #[cfg(windows)] + #[test] + fn dry_run_detects_case_only_output_collisions_without_writing() { + let dir = tempfile::tempdir().unwrap(); + let first = dir.path().join("Case.pdf"); + let second = dir.path().join("CASE.pdf"); + let targets = vec![("first.pdf".into(), first), ("second.pdf".into(), second)]; + assert!(check_unique_outputs(&targets, true).is_err()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 0); + } + + #[cfg(windows)] + #[test] + fn permission_denied_child_directory_uses_delete_on_close_file_probes() { + use std::process::Command; + + let dir = tempfile::tempdir().unwrap(); + let identity = Command::new("whoami.exe").args(["/user", "/fo", "csv", "/nh"]).output().unwrap(); + assert!(identity.status.success()); + let identity = String::from_utf8(identity.stdout).unwrap(); + let sid = identity.split('"').nth(3).unwrap(); + let acl = Command::new("icacls.exe").arg(dir.path()).arg("/deny").arg(format!("*{sid}:(AD)")).status().unwrap(); + assert!(acl.success()); + + assert!(matches!(tempfile::tempdir_in(dir.path()), Err(error) if error.kind() == std::io::ErrorKind::PermissionDenied)); + let direct_probe = dir.path().join("direct-probe"); + let file = std::fs::OpenOptions::new().write(true).create_new(true).open(&direct_probe).unwrap(); + drop(file); + std::fs::remove_file(&direct_probe).unwrap(); + + let targets = vec![("first.pdf".into(), dir.path().join("first.pdf")), ("second.pdf".into(), dir.path().join("second.pdf"))]; + assert!(check_unique_outputs(&targets, false).is_ok()); + assert_eq!(std::fs::read_dir(dir.path()).unwrap().count(), 0); + } +} diff --git a/crates/automation/src/file_access.rs b/crates/automation/src/file_access.rs index dab2760f5d..ff91152567 100644 --- a/crates/automation/src/file_access.rs +++ b/crates/automation/src/file_access.rs @@ -393,7 +393,8 @@ mod tests { fn tools_do_not_touch_the_file_system_without_resolving_the_path() { // Helpers that are only called with a resolved path. `doc_batch` (batch.rs) resolves its // inputs and out_dir, walks only inside them (skipping links) and resolves every - // discovered file again before `process` reads it or `publish` writes it. + // discovered file again before `process` reads it or `publish` writes it. `doc_bates_multi` + // resolves its paths before checking output collisions and using the probe helpers. const RESOLVED_BY_CALLER: &[&str] = &[ "write_atomic", "write_atomic_with", @@ -405,6 +406,8 @@ mod tests { "linked", "walk", "publish", + "check_unique_outputs", + "probe_files_for_resolved_outputs", ]; let mut dirs = vec![Path::new(env!("CARGO_MANIFEST_DIR")).join("src")]; let mut files = vec![]; diff --git a/crates/automation/src/tools.rs b/crates/automation/src/tools.rs index 16371c6a70..8188045b87 100644 --- a/crates/automation/src/tools.rs +++ b/crates/automation/src/tools.rs @@ -140,8 +140,8 @@ pub fn tools() -> Vec { .with(schema(json!({ "files": {"type":"array","minItems":1,"maxItems":1000,"description":"PDFs in Bates sequence order.","items":{"type":"object","properties":{"input":{"type":"string","minLength":1,"description":"Input PDF path."},"output":{"type":"string","minLength":1,"description":"New output PDF path; must not exist."},"password":{"type":"string","description":"Open or permissions password, if needed."}},"required":["input","output"],"additionalProperties":false}}, "start_number": {"type":"integer","minimum":1,"description":"First number in the sequence (default 1)."}, "digits": {"type":"integer","minimum":1,"maximum":12,"description":"Zero-padded number width (default 6)."}, - "prefix": {"type":"string","maxLength":128,"description":"Text before the number: at most 128 UTF-8 bytes, without # or >>."}, "suffix": {"type":"string","maxLength":128,"description":"Text after the number: at most 128 UTF-8 bytes, without # or >>."}, - "dry_run": {"type":"boolean","description":"Return each file's inclusive number interval without writing (default false)."} + "prefix": {"type":"string","maxLength":128,"description":"Bates prefix, up to 128 characters; cannot contain # or >>."}, "suffix": {"type":"string","maxLength":128,"description":"Bates suffix, up to 128 characters; cannot contain # or >>."}, + "dry_run": {"type":"boolean","description":"Return each file's inclusive number interval without writing (default false). Output collisions are checked conservatively: Windows compares case-insensitively, while Unix preserves case, so case-sensitive Windows and case-insensitive Unix volumes can differ from normal runs."} }), &["files"])), t("doc_batch", "Process files and folders", "Run convert, ocr and/or optimize over files and folders, one file at a time. recursive defaults to false. Preserve paths relative to each selected folder inside out_dir; duplicate inputs run once. Skip unsupported files, links and the output subtree. Existing outputs receive numbered names and are never replaced. Continue after per-file errors unless stop_on_error is true. Return completed/failed/skipped counts and per-file results. Discovery is bounded to 10000 entries and 64 folder levels. Convert must be first; each step can appear once. OCR uses installed models. Optimize uses Reduce File Size settings, optionally overriding target ppi (threshold 1.5 times ppi; black-and-white images go to the larger of ppi and 300) and JPEG quality; a larger candidate keeps the original bytes. Sources and open documents are unchanged.") .with(schema(json!({ diff --git a/crates/automation/tests/it/automation.rs b/crates/automation/tests/it/automation.rs index da634490e4..1192050e51 100644 --- a/crates/automation/tests/it/automation.rs +++ b/crates/automation/tests/it/automation.rs @@ -2969,6 +2969,15 @@ fn bates_numbers_continue_across_files_and_report_preflight_and_permission_failu {"input":"a.pdf","output":"a-numbered.pdf"} ]); let mut a = auto(&dir); + for invalid_start in [json!(0), json!(-1), json!(1.5), json!("17")] { + let error = err(&mut a, "doc_bates_multi", json!({"files":files,"start_number":invalid_start})); + assert!(matches!(error, ToolError::InvalidArgs(_))); + } + let unicode = ok(&mut a, "doc_bates_multi", json!({"files":files,"prefix":"é".repeat(65),"dry_run":true})); + assert_eq!(unicode["files"].as_array().unwrap().len(), 2); + assert!(matches!(err(&mut a, "doc_bates_multi", json!({"files":files,"prefix":"CASE>>"})), ToolError::InvalidArgs(_))); + assert!(matches!(err(&mut a, "doc_bates_multi", json!({"files":files,"suffix":"A>>B"})), ToolError::InvalidArgs(_))); + let allocation = ok(&mut a, "doc_bates_multi", json!({"files":files,"start_number":50,"dry_run":true})); assert_eq!(allocation["files"][0]["first"], 50); assert_eq!(allocation["files"][0]["last"], 51); @@ -2987,15 +2996,24 @@ fn bates_numbers_continue_across_files_and_report_preflight_and_permission_failu let overflow = err(&mut a, "doc_bates_multi", json!({"files":[{"input":"b.pdf","output":"overflow.pdf"}],"start_number":u64::MAX})); assert!(overflow.to_string().contains("overflows")); assert!(!dir.join("overflow.pdf").exists()); - // A range may end at u64::MAX itself. - let at_max = ok(&mut a, "doc_bates_multi", json!({"files":[{"input":"b.pdf","output":"max.pdf"}],"start_number":u64::MAX - 1,"dry_run":true})); - assert_eq!(at_max["files"][0]["last"], u64::MAX); - for bad in - [json!({"start_number":-3}), json!({"start_number":2.5}), json!({"start_number":"7"}), json!({"prefix":"CASE>>"}), json!({"suffix":"A>>B"})] + std::fs::write(dir.join("one.pdf"), fixture(1)).unwrap(); + let max_number = + ok(&mut a, "doc_bates_multi", json!({"files":[{"input":"one.pdf","output":"one-numbered.pdf"}],"start_number":u64::MAX,"dry_run":true})); + assert_eq!(max_number["files"][0]["first"], u64::MAX); + assert_eq!(max_number["files"][0]["last"], u64::MAX); + + #[cfg(windows)] { - let mut args = json!({"files":[{"input":"b.pdf","output":"bad.pdf"}],"dry_run":true}); - args.as_object_mut().unwrap().extend(bad.as_object().unwrap().clone()); - assert!(a.call("doc_bates_multi", &args).is_err(), "{bad}"); + let collision = err( + &mut a, + "doc_bates_multi", + json!({"files":[ + {"input":"b.pdf","output":"Case.pdf"}, + {"input":"a.pdf","output":"CASE.pdf"} + ]}), + ); + assert!(collision.to_string().contains("same output path")); + assert!(!dir.join("Case.pdf").exists()); } use pdfcraft_engine::{Changes, Edit, Protection, Session}; From 9642ed6f87584908a65d5bf78726e102c94a0f54 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:42:09 -0500 Subject: [PATCH 26/75] fix(sign): read each existing DSS stream once and within a decoded-size allowance (#6040) --- crates/sign/src/dss.rs | 87 +++++++++++++++++++++++++++++++++++++----- 1 file changed, 77 insertions(+), 10 deletions(-) diff --git a/crates/sign/src/dss.rs b/crates/sign/src/dss.rs index 0d82c797a7..0c76ca1967 100644 --- a/crates/sign/src/dss.rs +++ b/crates/sign/src/dss.rs @@ -37,9 +37,12 @@ pub fn embed(doc: &Document, evidence: &Evidence) -> Result, SignError> let mut ocsps: Vec> = Vec::new(); let mut crls: Vec> = Vec::new(); if let Some(d) = &existing { - pull(&doc, d, b"Certs", &mut certs)?; - pull(&doc, d, b"OCSPs", &mut ocsps)?; - pull(&doc, d, b"CRLs", &mut crls)?; + let mut budget = MAX_EXISTING_EVIDENCE; + // Shared by the three reads: a stream listed under two keys is decoded and charged once. + let mut decoded = std::collections::HashMap::new(); + pull(&doc, d, b"Certs", &mut certs, &mut budget, &mut decoded)?; + pull(&doc, d, b"OCSPs", &mut ocsps, &mut budget, &mut decoded)?; + pull(&doc, d, b"CRLs", &mut crls, &mut budget, &mut decoded)?; } let add = |list: &mut Vec>, blob: &[u8]| { if !list.iter().any(|b| b.as_slice() == blob) { @@ -105,19 +108,83 @@ fn embed_stream(doc: &mut Document, blob: &[u8]) -> pdfcraft_cos::ObjRef { doc.add(Object::Stream(Stream::from_raw(d, blob.to_vec()))) } -/// Pull the raw bytes of an existing DSS array into `into`. -fn pull(doc: &Document, dss: &Dict, key: &[u8], into: &mut Vec>) -> Result<(), SignError> { +/// Most decoded bytes of the store's existing evidence one [`embed`] reads. The store is rewritten +/// as a union, so what is there has to be read; a file whose evidence is larger than this is +/// refused rather than extended, never silently trimmed. +const MAX_EXISTING_EVIDENCE: usize = 256 << 20; + +/// Pull the raw bytes of an existing DSS array into `into`, each stream once (however often the +/// array lists it) and each distinct content once, spending the decoded size from `budget`. +fn pull( + doc: &Document, + dss: &Dict, + key: &[u8], + into: &mut Vec>, + budget: &mut usize, + decoded: &mut std::collections::HashMap>, +) -> Result<(), SignError> { let Some(a) = dss.get(key).map(|o| doc.resolve(o)).and_then(|o| o.as_array().cloned()) else { return Ok(()); }; + let mut seen_refs = std::collections::HashSet::new(); + let mut seen_content: std::collections::HashSet> = into.iter().map(|b| crate::keys::DigestAlg::Sha256.digest(&[b.as_slice()])).collect(); for item in a { - let o = doc.resolve(&item); - let Object::Stream(s) = &*o else { continue }; - if let Ok(bytes) = s.decoded() - && !into.iter().any(|b| b == &bytes) - { + if item.as_ref().is_some_and(|r| !seen_refs.insert(r)) { + continue; + } + // A stream another DSS array already read is reused: it is neither decoded again nor + // charged twice, so a store that lists the same evidence under two keys still fits. + let bytes = match item.as_ref().and_then(|r| decoded.get(&r).cloned()) { + Some(bytes) => bytes, + None => { + let o = doc.resolve(&item); + let Object::Stream(s) = &*o else { continue }; + let too_big = || SignError::Pdf("the document's existing validation data is too large to extend".into()); + match s.decoded_within(*budget) { + Ok(bytes) => { + *budget = budget.checked_sub(bytes.len()).ok_or_else(too_big)?; + if let Some(r) = item.as_ref() { + decoded.insert(r, bytes.clone()); + } + bytes + } + // The allowance is spent: refuse, rather than drop evidence the store holds. + Err(e) if e.to_string().contains("exceeds the limit") => return Err(too_big()), + // Not decodable at all: skipped, as before. + Err(_) => continue, + } + } + }; + if seen_content.insert(crate::keys::DigestAlg::Sha256.digest(&[bytes.as_slice()])) { into.push(bytes); } } Ok(()) } + +#[cfg(test)] +mod tests { + use super::*; + + fn store(blobs: &[&[u8]], listed: &[usize]) -> (Document, Dict) { + let mut doc = Document::new_empty(); + let refs: Vec<_> = blobs.iter().map(|b| doc.add(Object::Stream(Stream::from_raw(Dict::new(), b.to_vec())))).collect(); + let mut dss = Dict::new(); + dss.set(b"Certs".to_vec(), Object::Array(listed.iter().map(|i| Object::Ref(refs[*i])).collect())); + (doc, dss) + } + + #[test] + fn a_stream_listed_many_times_is_read_once_and_the_decoded_size_is_spent() { + let (doc, dss) = store(&[&[1; 100], &[1; 100], &[2; 50]], &[0, 0, 0, 1, 2, 2]); + let mut out = Vec::new(); + let mut budget = 1000; + pull(&doc, &dss, b"Certs", &mut out, &mut budget, &mut std::collections::HashMap::new()).unwrap(); + // Two distinct streams are listed with the same bytes: one copy; the repeats were not read. + assert_eq!(out, [vec![1; 100], vec![2; 50]]); + assert_eq!(budget, 1000 - 100 - 100 - 50); + // Past the allowance the store is refused, not trimmed. + let mut budget = 120; + assert!(pull(&doc, &dss, b"Certs", &mut Vec::new(), &mut budget, &mut std::collections::HashMap::new()).is_err()); + } +} From 74e1e40e676456d4eb92556db097f96c7ed24731 Mon Sep 17 00:00:00 2001 From: "Dylan M. Taylor" Date: Sun, 8 Nov 2026 19:42:24 -0500 Subject: [PATCH 27/75] fix(a11y): heading titles decode their source strings within a shared allowance (#6049) --- crates/a11y/src/headings.rs | 33 +++++++++++++++++++++++++++++---- crates/a11y/src/tests.rs | 18 ++++++++++++++++++ 2 files changed, 47 insertions(+), 4 deletions(-) diff --git a/crates/a11y/src/headings.rs b/crates/a11y/src/headings.rs index 0fa31ccbb1..235ba4b452 100644 --- a/crates/a11y/src/headings.rs +++ b/crates/a11y/src/headings.rs @@ -37,9 +37,16 @@ struct Walker<'a> { visits: usize, /// Per page: the text each MCID shows (read once per page). texts: HashMap>, + /// What is left of [`MAX_TITLE_SOURCE_BYTES`]. + title_left: usize, out: Vec, } +/// Bytes of `/ActualText`, `/Alt` and `/T` strings the headings of one document decode in all. +/// Titles are cut to [`MAX_TITLE`] characters, but many headings can name one large string, and +/// each would otherwise be decoded in full first. +const MAX_TITLE_SOURCE_BYTES: usize = 32 << 20; + /// The headings, in document order. Headings without text or a page are left out. pub fn headings(doc: &Document) -> Vec { let Some(root) = tree_root(doc) else { return Vec::new() }; @@ -52,6 +59,7 @@ pub fn headings(doc: &Document) -> Vec { seen: HashSet::new(), visits: 0, texts: HashMap::new(), + title_left: MAX_TITLE_SOURCE_BYTES, out: Vec::new(), }; if let Some(k) = root.get(b"K") { @@ -131,17 +139,34 @@ impl Walker<'_> { self.marks(kids, pg, depth + 1, &mut marks); } let Some(page) = pg.or_else(|| marks.iter().find_map(|m| m.0)).and_then(|p| self.page_index.get(&p).copied()) else { return }; - let doc = self.doc; - let attr = |key: &[u8]| d.get(key).map(|o| doc.resolve(o)).and_then(|o| o.as_string().map(|s| s.to_text())).filter(|t| !t.trim().is_empty()); - let title = match attr(b"ActualText") { + let title = match self.attr(d, b"ActualText") { Some(t) => Some(t), - None => self.marked_text(&marks, page).or_else(|| attr(b"Alt")).or_else(|| attr(b"T")), + None => match self.marked_text(&marks, page) { + Some(t) => Some(t), + None => self.attr(d, b"Alt").or_else(|| self.attr(d, b"T")), + }, }; let Some(title) = title else { return }; let title: String = title.split_whitespace().collect::>().join(" ").chars().take(MAX_TITLE).collect(); self.out.push(Heading { obj, level, title, page }); } + /// The non-blank text of the string `key` holds, charged against the title allowance by the + /// longer of its raw and decoded size (a string that does not fit is not decoded). + fn attr(&mut self, d: &Dict, key: &[u8]) -> Option { + let o = self.doc.resolve(d.get(key)?); + let s = o.as_string()?; + // A text string decodes to at most three bytes per raw byte, so one whose worst case does + // not fit the remaining allowance is never decoded: the raw check alone let a string at + // the raw limit decode to three times it before the allowance was reduced. + if s.bytes.len().saturating_mul(3) > self.title_left { + return None; + } + let text = s.to_text(); + self.title_left = self.title_left.saturating_sub(text.len().max(s.bytes.len())); + Some(text).filter(|t| !t.trim().is_empty()) + } + /// The text the marked content shows (`page`: where marks without a page of their own are). fn marked_text(&mut self, marks: &[(Option, i64)], page: usize) -> Option { let mut parts = Vec::new(); diff --git a/crates/a11y/src/tests.rs b/crates/a11y/src/tests.rs index d8566eddf1..66ae06fce4 100644 --- a/crates/a11y/src/tests.rs +++ b/crates/a11y/src/tests.rs @@ -1559,6 +1559,24 @@ fn q_inside_a_text_object_restores_the_text_matrices() { assert_eq!(titles, ["Chapter One", "Chapter Two"]); } +#[test] +fn headings_naming_one_large_string_stop_decoding_it_at_the_allowance() { + let count = 100; + let kids = (0..count).map(|i| format!("{} 0 R", 8 + i)).collect::>().join(" "); + let mut objects = vec![ + "<< /Type /Catalog /Pages 2 0 R /MarkInfo << /Marked true >> /StructTreeRoot 5 0 R >>".to_string(), + "<< /Type /Pages /Kids [3 0 R] /Count 1 >>".to_string(), + "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /Contents 4 0 R >>".to_string(), + stream("BT ET"), + "<< /Type /StructTreeRoot /K 6 0 R >>".to_string(), + format!("<< /S /Document /P 5 0 R /K [{kids}] >>"), + format!("({})", "a".repeat(512 << 10)), + ]; + objects.extend((0..count).map(|_| "<< /S /H1 /P 6 0 R /Pg 3 0 R /ActualText 7 0 R >>".to_string())); + let found = headings(&pdf(&objects, "")); + assert!(!found.is_empty() && found.len() < count, "{}", found.len()); +} + #[test] fn an_untagged_document_has_no_headings() { let mut objs = headings_doc(); From 3b731f9458e74a9abf406fb0e759bcc8993c5855 Mon Sep 17 00:00:00 2001 From: Harrison Booth <73434116+Harrison-Booth@users.noreply.github.com> Date: Sun, 8 Nov 2026 18:42:39 -0600 Subject: [PATCH 28/75] Expose widget rectangles to field scripts (#6093) --- crates/engine/src/js.rs | 4 +++- crates/engine/src/tests.rs | 14 ++++++++++++++ crates/forms/src/scripting.rs | 5 +++++ crates/forms/src/tests.rs | 1 + crates/js/src/document.rs | 31 +++++++++++++++++++++++++++++++ crates/js/src/model.rs | 3 +++ crates/js/src/tests.rs | 16 ++++++++++++++++ crates/ui-egui/src/canvas/area.rs | 1 + parity/acrobat-features.toml | 4 ++-- 9 files changed, 76 insertions(+), 3 deletions(-) diff --git a/crates/engine/src/js.rs b/crates/engine/src/js.rs index 7f0561d196..fead2dbebc 100644 --- a/crates/engine/src/js.rs +++ b/crates/engine/src/js.rs @@ -159,6 +159,7 @@ fn field_state_with_display(f: &Field, display: i32) -> pdfcraft_js::FieldState s.readonly = f.read_only(); s.required = f.has(pdfcraft_forms::flags::REQUIRED); s.display = display; + s.rect = f.widgets.first().map(|w| w.rect); s.char_limit = f.max_len; s.options = match f.kind { FieldKind::CheckBox | FieldKind::Radio => f.widgets.iter().filter_map(|w| w.on_state.clone()).map(|o| (o.clone(), o)).collect(), @@ -188,8 +189,9 @@ fn changes(before: &[pdfcraft_js::FieldState], after: &[pdfcraft_js::FieldState] read_only: (a.readonly != b.readonly).then_some(a.readonly), required: (a.required != b.required).then_some(a.required), display: (a.display != b.display).then_some(a.display), + rect: (a.rect != b.rect).then_some(a.rect).flatten(), }; - (c.value.is_some() || c.read_only.is_some() || c.required.is_some() || c.display.is_some()).then_some(c) + (c.value.is_some() || c.read_only.is_some() || c.required.is_some() || c.display.is_some() || c.rect.is_some()).then_some(c) }) .collect() } diff --git a/crates/engine/src/tests.rs b/crates/engine/src/tests.rs index 4df8608946..181d5de452 100644 --- a/crates/engine/src/tests.rs +++ b/crates/engine/src/tests.rs @@ -6080,6 +6080,20 @@ fn form_javascript_validates_calculates_and_formats() { assert!(s.run_javascript(id, "1", None).is_err()); } +#[test] +fn javascript_field_rect_moves_first_widget_and_undoes() { + let mut s = Session::new(); + let id = s.open("rect.pdf", None, Arc::new(scripted_form()), None).unwrap(); + let initial = s.get(id).unwrap().form.iter().find(|f| f.name == "price").unwrap().widgets[0].rect; + s.run_javascript(id, "getField('price').rect = [30, 40, 130, 60];", Some("go")).unwrap(); + assert_eq!(s.get(id).unwrap().form.iter().find(|f| f.name == "price").unwrap().widgets[0].rect, [30.0, 40.0, 130.0, 60.0]); + assert_eq!(s.get(id).unwrap().can_undo(), Some("Run JavaScript")); + s.undo(id).unwrap(); + assert_eq!(s.get(id).unwrap().form.iter().find(|f| f.name == "price").unwrap().widgets[0].rect, initial); + s.redo(id).unwrap(); + assert_eq!(s.get(id).unwrap().form.iter().find(|f| f.name == "price").unwrap().widgets[0].rect, [30.0, 40.0, 130.0, 60.0]); +} + /// XFA forms check `xfa_installed` and `xfa_version` in a document script that runs before /// each of their scripts, and alert when either is missing. #[test] diff --git a/crates/forms/src/scripting.rs b/crates/forms/src/scripting.rs index 31e6812e6b..0f7452d96e 100644 --- a/crates/forms/src/scripting.rs +++ b/crates/forms/src/scripting.rs @@ -42,6 +42,8 @@ pub struct FieldChange { pub required: Option, /// `display.visible` 0, `hidden` 1, `noPrint` 2, `noView` 3. pub display: Option, + /// Move/resize the first widget, in PDF user space. + pub rect: Option<[f64; 4]>, } /// What a script did. @@ -173,6 +175,9 @@ pub(crate) fn apply_changes(doc: &mut Document, changes: &[FieldChange], except: } } } + if let Some(rect) = c.rect { + crate::set_props(doc, &c.name, &crate::FieldProps { rect: Some((0, rect)), ..Default::default() })?; + } } Ok(()) } diff --git a/crates/forms/src/tests.rs b/crates/forms/src/tests.rs index 6cf207d0de..5080ed03bd 100644 --- a/crates/forms/src/tests.rs +++ b/crates/forms/src/tests.rs @@ -2994,6 +2994,7 @@ fn recalc_chain_where_script_changes_checkbox() { read_only: None, required: None, display: None, + rect: None, }], message: None, } diff --git a/crates/js/src/document.rs b/crates/js/src/document.rs index e977457e2b..54274bb140 100644 --- a/crates/js/src/document.rs +++ b/crates/js/src/document.rs @@ -94,6 +94,36 @@ pub(crate) fn field_object(ctx: &mut Context, name: &str) -> JsObject { } function(ctx, NativeFunction::from_copy_closure_with_captures(set, n.clone())) }; + let rect_get = getter!(ctx, n, |f, c| match f.rect { + Some(rect) => JsArray::from_iter(rect.into_iter().map(JsValue::from).collect::>(), c).into(), + None => JsValue::null(), + }); + let rect_set = { + fn set(_: &JsValue, args: &[JsValue], name: &JsString, c: &mut Context) -> JsResult { + let value = arg(args, 0); + let object = value.as_object().filter(|o| o.is_array()).ok_or_else(|| error("rect must be an array of four finite numbers"))?; + let array = JsArray::from_object(object.clone())?; + if array.length(c)? != 4 { + return Err(error("rect must be an array of four finite numbers")); + } + let mut rect = [0.0; 4]; + for (i, coordinate) in rect.iter_mut().enumerate() { + *coordinate = array.get(i as u64, c)?.to_number(c)?; + } + if !rect.iter().all(|coordinate| coordinate.is_finite()) { + return Err(error("rect coordinates must be finite")); + } + let width = (rect[2] - rect[0]).abs(); + let height = (rect[3] - rect[1]).abs(); + if !width.is_finite() || !height.is_finite() || width < 4.0 || height < 4.0 { + return Err(error("rect must be at least 4 points wide and high")); + } + let rect = [rect[0].min(rect[2]), rect[1].min(rect[3]), rect[0].max(rect[2]), rect[1].max(rect[3])]; + write(c, name, |f| f.rect = Some(rect))?; + Ok(JsValue::undefined()) + } + function(ctx, NativeFunction::from_copy_closure_with_captures(set, n.clone())) + }; fn get_item_at(_: &JsValue, args: &[JsValue], name: &JsString, c: &mut Context) -> JsResult { let f = read(c, name)?; @@ -150,6 +180,7 @@ pub(crate) fn field_object(ctx: &mut Context, name: &str) -> JsObject { .accessor(js_string!("currentValueIndices"), Some(indices), None, rw) .accessor(js_string!("textColor"), Some(text_color_get), Some(text_color_set), rw) .accessor(js_string!("fillColor"), Some(fill_color_get), Some(fill_color_set), rw) + .accessor(js_string!("rect"), Some(rect_get), Some(rect_set), rw) .function(get_item_at, js_string!("getItemAt"), 2) .function(is_box_checked, js_string!("isBoxChecked"), 1) .function(check_this_box, js_string!("checkThisBox"), 2) diff --git a/crates/js/src/model.rs b/crates/js/src/model.rs index 36b09bec97..44d5ac2dee 100644 --- a/crates/js/src/model.rs +++ b/crates/js/src/model.rs @@ -52,6 +52,8 @@ pub struct FieldState { /// `textColor` / `fillColor` as Acrobat colour arrays (`["RGB", r, g, b]` etc.), when set. pub text_color: Option>, pub fill_color: Option>, + /// Rectangle of the first widget, in PDF user space. + pub rect: Option<[f64; 4]>, } impl FieldState { @@ -68,6 +70,7 @@ impl FieldState { char_limit: None, text_color: None, fill_color: None, + rect: None, } } diff --git a/crates/js/src/tests.rs b/crates/js/src/tests.rs index b61649b729..7f8595ab56 100644 --- a/crates/js/src/tests.rs +++ b/crates/js/src/tests.rs @@ -73,6 +73,22 @@ fn scripts_change_other_fields_and_their_properties() { assert_eq!(o.changed.iter().find(|f| f.name == "size").unwrap().value, ["l"]); } +#[test] +fn widget_rect_reads_and_writes_a_four_number_array() { + let o = go( + "var t = getField('total'); t.rect = [10, 20, 110, 50]; event.value = getField('total').rect.join(',');", + &Event::field("Mouse Up", "agree", ""), + ); + assert_eq!(o.error, None); + assert_eq!(o.value, "10,20,110,50"); + assert_eq!(o.changed.iter().find(|f| f.name == "total").and_then(|f| f.rect), Some([10.0, 20.0, 110.0, 50.0])); + let o = go("getField('total').rect = [110, 50, 10, 20];", &Event::field("Mouse Up", "agree", "")); + assert_eq!(o.changed.iter().find(|f| f.name == "total").and_then(|f| f.rect), Some([10.0, 20.0, 110.0, 50.0])); + let o = go("getField('total').rect = [0, 0, Infinity, 10];", &Event::field("Mouse Up", "agree", "")); + assert!(o.error.is_some()); + assert!(o.changed.iter().all(|f| f.name != "total")); +} + #[test] fn choice_and_check_box_object_model() { let o = go( diff --git a/crates/ui-egui/src/canvas/area.rs b/crates/ui-egui/src/canvas/area.rs index 9fd1cdd615..64cbd6e939 100644 --- a/crates/ui-egui/src/canvas/area.rs +++ b/crates/ui-egui/src/canvas/area.rs @@ -89,6 +89,7 @@ pub(super) fn run_button(app: &mut PdfCraftApp, index: usize, name: &str, action read_only: None, required: None, display: Some(if hide { 1 } else { 0 }), + rect: None, }) .collect(); if !changes.is_empty() { diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index e0c7adf36a..99a30c0b0f 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -5351,8 +5351,8 @@ milestone = "M6" status = "partial" commands = ["tools.js_console"] tools = ["js_run"] -evidence = ["crates/js/src/tests.rs::calculate_scripts_read_fields_and_set_the_value", "crates/engine/src/tests.rs::form_javascript_validates_calculates_and_formats", "crates/automation/tests/it/automation.rs::javascript_through_tools", "crates/js/src/tests.rs::an_xfa_form_sees_the_xfa_plugin_installed", "crates/js/src/tests.rs::a_plain_form_sees_the_xfa_version_but_not_installed", "crates/engine/src/tests.rs::scripts_in_an_xfa_form_see_the_xfa_plugin_installed", "crates/js/src/tests.rs::scand_allows_separators_between_fields_but_nothing_before_or_after_them", "crates/js/src/tests.rs::app_alert_keeps_its_title_icon_and_buttons", "crates/js/src/tests.rs::app_alert_reads_the_object_form", "crates/js/src/tests.rs::app_alert_falls_back_on_bad_icon_and_button_numbers", "crates/js/src/tests.rs::app_alert_titles_are_clipped_and_charged", "crates/ui-egui/tests/it/javascript.rs::every_message_a_script_shows_takes_the_toast_in_turn"] -notes = "boa (pure Rust) with the forms object model: event, getField/Field (value, readonly, display, colours, items, check boxes), app (with execMenuItem for the side panels and Save As, from a button only, and Print where this.print() works), util.printf/printd/printx/scand (scand reads dates as another viewer was observed to, in AcroForm and XFA scripts), console, display, color; xfa_version (3.5) in every document and xfa_installed in XFA forms, so their version check passes; no annotations, layers or dialogs (app.response returns null) yet, and no app.findComponent. app.alert keeps its title, icon (error, warning, question, status) and buttons (OK, OK/Cancel, Yes/No, Yes/No/Cancel), given in order or as one object; js_run reports them with each message, and the app shows every message in turn in the toast with its title and icon. Still missing: answers. Every app.alert returns 1 (OK), so a script is never told No or Cancel (#4291)." +evidence = ["crates/js/src/tests.rs::calculate_scripts_read_fields_and_set_the_value", "crates/engine/src/tests.rs::form_javascript_validates_calculates_and_formats", "crates/automation/tests/it/automation.rs::javascript_through_tools", "crates/js/src/tests.rs::an_xfa_form_sees_the_xfa_plugin_installed", "crates/js/src/tests.rs::a_plain_form_sees_the_xfa_version_but_not_installed", "crates/engine/src/tests.rs::scripts_in_an_xfa_form_see_the_xfa_plugin_installed", "crates/js/src/tests.rs::scand_allows_separators_between_fields_but_nothing_before_or_after_them", "crates/js/src/tests.rs::app_alert_keeps_its_title_icon_and_buttons", "crates/js/src/tests.rs::app_alert_reads_the_object_form", "crates/js/src/tests.rs::app_alert_falls_back_on_bad_icon_and_button_numbers", "crates/js/src/tests.rs::app_alert_titles_are_clipped_and_charged", "crates/ui-egui/tests/it/javascript.rs::every_message_a_script_shows_takes_the_toast_in_turn", "crates/js/src/tests.rs::widget_rect_reads_and_writes_a_four_number_array", "crates/engine/src/tests.rs::javascript_field_rect_moves_first_widget_and_undoes"] +notes = "boa (pure Rust) with the forms object model: event, getField/Field (value, readonly, display, colours, items, check boxes, rect), app (with execMenuItem for the side panels and Save As, from a button only, and Print where this.print() works), util.printf/printd/printx/scand (scand reads dates as another viewer was observed to, in AcroForm and XFA scripts), console, display, color; xfa_version (3.5) in every document and xfa_installed in XFA forms, so their version check passes. field.rect reads and writes the first widget in PDF user space. no annotations, layers or dialogs (app.response returns null) yet, and no app.findComponent. app.alert keeps its title, icon (error, warning, question, status) and buttons (OK, OK/Cancel, Yes/No, Yes/No/Cancel), given in order or as one object; js_run reports them with each message, and the app shows every message in turn in the toast with its title and icon. Still missing: answers. Every app.alert returns 1 (OK), so a script is never told No or Cancel (#4291)." [[feature]] id = "form.js-console" From 9b2c8d6026aa8fa454fb1ffb3a7e4299bce559b7 Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:42:54 +0800 Subject: [PATCH 29/75] fix(xfa): read and validate the first alternative in typed_date (#6120) --- crates/xfa/src/picture.rs | 50 +++++++++++++++++++++++++++++++++++++-- 1 file changed, 48 insertions(+), 2 deletions(-) diff --git a/crates/xfa/src/picture.rs b/crates/xfa/src/picture.rs index 3ddb392114..f1296e495a 100644 --- a/crates/xfa/src/picture.rs +++ b/crates/xfa/src/picture.rs @@ -39,14 +39,15 @@ pub fn typed_date(tpl: &Template, som: &str, text: &str) -> Option { } let picture = f.picture.as_deref()?; let reads = |pattern: &str| pattern_to_iso(text, pattern).filter(|iso| iso_to_pattern(iso, pattern).eq_ignore_ascii_case(text.trim())); - let first = date_pattern(Some(picture)); + let mut alts = alternatives(picture).into_iter(); + let first = alts.next().filter(|a| is_date_clause(a) && date_letters(a).is_some()).map(|a| date_pattern(Some(a)))?; if reads(&first).is_some() { return None; } // The first later date pattern that reads the text and gives a date the first pattern keeps: // written in the first, it must read back as the same date (a two-digit year can't hold 1901 // or 1203), else the next pattern is tried. - alternatives(picture).into_iter().skip(1).filter(|a| is_date_clause(a)).filter_map(date_letters).find_map(|p| { + alts.filter(|a| is_date_clause(a)).filter_map(date_letters).find_map(|p| { let iso = reads(&p)?; let shown = iso_to_pattern(&iso, &first); (pattern_to_iso(&shown, &first).as_deref() == Some(iso.as_str())).then_some(shown) @@ -63,3 +64,48 @@ fn is_date_clause(clause: &str) -> bool { None => true, } } + +#[cfg(test)] +mod tests { + use super::*; + + fn tpl_with_pictures(fields: &[(&str, &str)]) -> Template { + let mut body = String::new(); + for (name, pic) in fields { + body.push_str(&format!( + "{pic}" + )); + } + let xml = format!( + "\ + \ + " + ); + crate::parse(&xml).expect("valid template").0 + } + + #[test] + fn typed_date_reads_bare_first_alternative_instead_of_whole_picture_or_later_braces() { + let tpl = tpl_with_pictures(&[("bare", "YYYYMMDD|YYMMDD"), ("mixed", "YYYYMMDD|date{YYMMDD}")]); + assert_eq!(typed_date(&tpl, "form[0].bare[0]", "261018").as_deref(), Some("20261018")); + assert_eq!(typed_date(&tpl, "form[0].bare[0]", "20261018"), None); + assert_eq!(typed_date(&tpl, "form[0].mixed[0]", "261018").as_deref(), Some("20261018")); + assert_eq!(typed_date(&tpl, "form[0].mixed[0]", "20261018"), None); + } + + #[test] + fn typed_date_rejects_non_date_or_unreadable_first_alternative() { + let tpl = tpl_with_pictures(&[ + ("text_first", "text{YYYYMMDD}|date{YYMMDD}"), + ("num_first", "num{99999999}|date{YYMMDD}"), + ("datetime_first", "date{YYYY-MM-DD HH:MM}|date{YYMMDD}"), + ]); + assert_eq!(typed_date(&tpl, "form[0].text_first[0]", "261018"), None); + assert_eq!(typed_date(&tpl, "form[0].num_first[0]", "261018"), None); + assert_eq!(typed_date(&tpl, "form[0].datetime_first[0]", "261018"), None); + } +} From 19619ad2ccd85d9caba26a554aca6bda91874391 Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:43:09 +0800 Subject: [PATCH 30/75] fix(js): decode FormCalc Unicode string escapes and preserve chained unary not/! (#6122) --- crates/js/src/formcalc/lexer.rs | 167 +++++++++++++++++++++++-------- crates/js/src/formcalc/parser.rs | 33 +++++- 2 files changed, 152 insertions(+), 48 deletions(-) diff --git a/crates/js/src/formcalc/lexer.rs b/crates/js/src/formcalc/lexer.rs index 46af19f761..9d01abf331 100644 --- a/crates/js/src/formcalc/lexer.rs +++ b/crates/js/src/formcalc/lexer.rs @@ -49,52 +49,12 @@ pub(super) fn lex(src: &str) -> Result, String> { } } '"' => { - let mut s = String::new(); - i += 1; - loop { - match chars.get(i) { - None => return Err(format!("line {line}: unterminated string")), - Some('"') if chars.get(i + 1) == Some(&'"') => { - s.push('"'); - i += 2; - } - Some('"') => { - i += 1; - break; - } - Some(&ch) => { - if ch == '\n' { - line += 1; - } - s.push(ch); - i += 1; - } - } - if s.len() > MAX_STRING { - return Err(format!("line {line}: a string is too long")); - } - } - out.push(Token { tok: Tok::Str(s), line }); + let tok_line = line; + let s = lex_string(&chars, &mut i, &mut line)?; + out.push(Token { tok: Tok::Str(s), line: tok_line }); } c if c.is_ascii_digit() || (c == '.' && chars.get(i + 1).is_some_and(char::is_ascii_digit)) => { - let start = i; - while i < chars.len() && (chars[i].is_ascii_digit() || (chars[i] == '.' && chars.get(i + 1) != Some(&'.'))) { - i += 1; - } - if i < chars.len() && (chars[i] == 'e' || chars[i] == 'E') { - let mut j = i + 1; - if j < chars.len() && (chars[j] == '+' || chars[j] == '-') { - j += 1; - } - if j < chars.len() && chars[j].is_ascii_digit() { - i = j; - while i < chars.len() && chars[i].is_ascii_digit() { - i += 1; - } - } - } - let text: String = chars[start..i].iter().collect(); - let n: f64 = text.parse().map_err(|_| format!("line {line}: bad number {text}"))?; + let n = lex_number(&chars, &mut i, line)?; out.push(Token { tok: Tok::Num(n), line }); } '!' => { @@ -127,3 +87,122 @@ pub(super) fn lex(src: &str) -> Result, String> { out.push(Token { tok: Tok::Eof, line }); Ok(out) } + +fn hex_u32(digits: &[char]) -> Option { + let mut v = 0u32; + for &c in digits { + v = (v << 4) | c.to_digit(16)?; + } + Some(v) +} + +/// FormCalc string escape (`\uXXXX` or `\UXXXXXXXX`, XFA 3.3 part 8 p. 1054); any other `\` stays literal. +fn unicode_escape(chars: &[char], at: usize) -> Option<(char, usize)> { + if chars.get(at) != Some(&'\\') { + return None; + } + match chars.get(at + 1)? { + 'u' => { + let hi = hex_u32(chars.get(at + 2..at + 6)?)?; + if (0xD800..=0xDBFF).contains(&hi) + && chars.get(at + 6..at + 8) == Some(&['\\', 'u']) + && let Some(lo) = chars.get(at + 8..at + 12).and_then(hex_u32).filter(|lo| (0xDC00..=0xDFFF).contains(lo)) + { + let scalar = 0x1_0000 + ((hi - 0xD800) << 10) + (lo - 0xDC00); + return Some((char::from_u32(scalar)?, 12)); + } + Some((char::from_u32(hi)?, 6)) + } + 'U' => { + let code = hex_u32(chars.get(at + 2..at + 10)?)?; + Some((char::from_u32(code)?, 10)) + } + _ => None, + } +} + +fn lex_string(chars: &[char], i: &mut usize, line: &mut usize) -> Result { + let mut s = String::new(); + *i += 1; + loop { + match chars.get(*i) { + None => return Err(format!("line {line}: unterminated string")), + Some('"') if chars.get(*i + 1) == Some(&'"') => { + s.push('"'); + *i += 2; + } + Some('"') => { + *i += 1; + return Ok(s); + } + Some(&ch) => { + if let Some((decoded, step)) = unicode_escape(chars, *i) { + s.push(decoded); + *i += step; + } else { + if ch == '\n' { + *line += 1; + } + s.push(ch); + *i += 1; + } + } + } + if s.len() > MAX_STRING { + return Err(format!("line {line}: a string is too long")); + } + } +} + +/// Whether `c` may start a FormCalc identifier. +fn is_ident_start(c: char) -> bool { + c.is_alphabetic() || c == '_' || c == '$' +} + +/// End index of a numeric exponent starting at `at` (`e`/`E`, optional sign, digits), or `None` +/// if no valid exponent starts there or if identifier characters immediately follow the digits. +fn exponent_end(chars: &[char], at: usize) -> Option { + if !matches!(chars.get(at), Some('e' | 'E')) { + return None; + } + let mut j = at + 1; + if matches!(chars.get(j), Some('+' | '-')) { + j += 1; + } + if !chars.get(j).is_some_and(char::is_ascii_digit) { + return None; + } + while chars.get(j).is_some_and(char::is_ascii_digit) { + j += 1; + } + if chars.get(j).is_some_and(|&c| c.is_alphanumeric() || c == '_' || c == '$') { + return None; + } + Some(j) +} + +/// Scan a FormCalc number literal starting at `*i`, stopping before `.member` or `..` while +/// rejecting malformed multi-dot literals such as `1.2.3`. +fn lex_number(chars: &[char], i: &mut usize, line: usize) -> Result { + let start = *i; + let mut seen_dot = false; + while *i < chars.len() { + let next = chars.get(*i + 1).copied(); + let next_is_member = next.is_some_and(is_ident_start) && exponent_end(chars, *i + 1).is_none(); + if chars[*i].is_ascii_digit() { + *i += 1; + } else if chars[*i] == '.' && !seen_dot && next != Some('.') && !next_is_member { + seen_dot = true; + *i += 1; + } else if chars[*i] == '.' && seen_dot && next.is_some_and(|c| c.is_ascii_digit()) { + return Err(format!("line {line}: bad number")); + } else { + break; + } + } + if let Some(end) = exponent_end(chars, *i) { + *i = end; + } + let text: String = chars[start..*i].iter().collect(); + text.parse().map_err(|_| format!("line {line}: bad number {text}")) +} diff --git a/crates/js/src/formcalc/parser.rs b/crates/js/src/formcalc/parser.rs index fceedadf5f..871d3874de 100644 --- a/crates/js/src/formcalc/parser.rs +++ b/crates/js/src/formcalc/parser.rs @@ -341,10 +341,7 @@ impl Parser { Ok(Expr::Neg(Box::new(self.unary()?))) } else if self.eat_op("+") { self.unary() - } else if self.eat_kw("not") || matches!(self.peek(), Tok::Ident(w) if w == "!") { - if matches!(self.peek(), Tok::Ident(w) if w == "!") { - self.at += 1; - } + } else if self.eat_kw("not") || self.eat_kw("!") { Ok(Expr::Not(Box::new(self.unary()?))) } else { self.postfix() @@ -449,3 +446,31 @@ pub(super) fn parse(src: &str) -> Result, String> { t => Err(format!("line {}: unexpected {}", p.line(), describe(t))), } } + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn unary_not_does_not_swallow_following_bang() { + assert_eq!(parse("not !1"), Ok(vec![Stmt::Expr(Expr::Not(Box::new(Expr::Not(Box::new(Expr::Num(1.0))))))])); + assert_eq!(parse("not !flag"), Ok(vec![Stmt::Expr(Expr::Not(Box::new(Expr::Not(Box::new(Expr::Ident("flag".into()))))))])); + assert_eq!(parse("!!1"), Ok(vec![Stmt::Expr(Expr::Not(Box::new(Expr::Not(Box::new(Expr::Num(1.0))))))])); + } + + #[test] + fn string_literals_decode_xfa_unicode_escapes_and_keep_other_backslashes() { + assert_eq!(parse(r#""\u0041\u20ac\U00000042\U0001F600""#), Ok(vec![Stmt::Expr(Expr::Str("A\u{20ac}B\u{1F600}".into()))])); + assert_eq!(parse(r#""\uD83D\uDE00""#), Ok(vec![Stmt::Expr(Expr::Str("\u{1F600}".into()))])); + assert_eq!(parse(r#""C:\temp\new\u12""#), Ok(vec![Stmt::Expr(Expr::Str(r"C:\temp\new\u12".into()))])); + } + + #[test] + fn number_literals_stop_before_second_dot_for_member_access() { + assert_eq!(parse("1.5.rawValue"), Ok(vec![Stmt::Expr(Expr::Member(Box::new(Expr::Num(1.5)), "rawValue".into()))])); + assert_eq!(parse("1.rawValue"), Ok(vec![Stmt::Expr(Expr::Member(Box::new(Expr::Num(1.0)), "rawValue".into()))])); + assert_eq!(parse("1.e2foo"), Ok(vec![Stmt::Expr(Expr::Member(Box::new(Expr::Num(1.0)), "e2foo".into()))])); + assert_eq!(parse("1.e2"), Ok(vec![Stmt::Expr(Expr::Num(100.0))])); + assert!(parse("1.2.3").is_err(), "a second decimal point followed by digits is a bad number"); + } +} From d22aab133c0148ad45921893be48a82b1c718b5f Mon Sep 17 00:00:00 2001 From: ".CooCoo" Date: Sun, 8 Nov 2026 16:43:49 -0800 Subject: [PATCH 31/75] fix(ui): two-page continuous view prefetches whole rows ahead (#6126) --- crates/ui-egui/src/canvas.rs | 18 ++++++++++++++++-- crates/ui-egui/src/canvas/tests.rs | 15 +++++++++++++++ 2 files changed, 31 insertions(+), 2 deletions(-) diff --git a/crates/ui-egui/src/canvas.rs b/crates/ui-egui/src/canvas.rs index 3e51dac3bf..b5cfd14910 100644 --- a/crates/ui-egui/src/canvas.rs +++ b/crates/ui-egui/src/canvas.rs @@ -2262,13 +2262,27 @@ fn scroll_prefetch(view: &DocView, visible_now: &[usize]) -> Vec { return Vec::new(); } let n = PREFETCH_ROWS * if view.layout == PageLayout::Single { 1 } else { 2 }; + // At high zoom the other page of a two-page row can be off to the side and not on screen, so + // the rows are counted from the edges of the rows shown, and that page is prefetched too. match (view.scroll_dir, visible_now.iter().min(), visible_now.iter().max()) { - (1, _, Some(&last)) => (last.saturating_add(1)..=last.saturating_add(n)).collect(), - (-1, Some(&first), _) => (first.saturating_sub(n)..first).rev().collect(), + (1, _, Some(&last)) => (last.saturating_add(1)..=row_bounds(view, last).1.saturating_add(n)).collect(), + (-1, Some(&first), _) => (row_bounds(view, first).0.saturating_sub(n)..first).rev().collect(), _ => Vec::new(), } } +/// The first and last page of the row that page `i` is in, paired as [`shown_row`] pairs them. +fn row_bounds(view: &DocView, i: usize) -> (usize, usize) { + match view.layout { + PageLayout::Single => (i, i), + PageLayout::TwoUp if view.cover && i == 0 => (0, 0), + PageLayout::TwoUp => { + let first = view.spread_first(i); + (first, first.saturating_add(1)) + } + } +} + /// The whole-page raster a prefetch asks for: the page at this scale, or the backdrop of a page /// that is tiled at this scale. A cached raster sharper than the backdrop stays demanded as it /// is, as `paint_page_surface` demands it under tiles. diff --git a/crates/ui-egui/src/canvas/tests.rs b/crates/ui-egui/src/canvas/tests.rs index 62c00d6b7a..0e86db7914 100644 --- a/crates/ui-egui/src/canvas/tests.rs +++ b/crates/ui-egui/src/canvas/tests.rs @@ -1536,6 +1536,21 @@ fn continuous_view_prefetches_the_rows_ahead_of_the_way_it_moved() { assert!(scroll_prefetch(&single, &[9]).is_empty(), "single-page view keeps its own prefetch"); } +#[test] +fn continuous_two_page_prefetch_counts_rows_from_the_end_of_the_row_shown() { + let mut two = view(20, PageLayout::TwoUp, true); + two.scroll_dir = 1; + assert_eq!(scroll_prefetch(&two, &[8]), [9, 10, 11, 12, 13], "the right page off to the side, then two rows"); + two.scroll_dir = -1; + assert_eq!(scroll_prefetch(&two, &[9]), [8, 7, 6, 5, 4], "the left page off to the side, then two rows"); + two.cover = true; + two.scroll_dir = 1; + assert_eq!(scroll_prefetch(&two, &[0]), [1, 2, 3, 4], "the cover is a row of its own"); + assert_eq!(scroll_prefetch(&two, &[3]), [4, 5, 6, 7, 8]); + two.scroll_dir = -1; + assert_eq!(scroll_prefetch(&two, &[4]), [3, 2, 1, 0]); +} + #[test] fn scrolling_the_other_way_drops_the_prefetch_from_the_queue() { let info = DocInfo { From e877b0822a79a18c6ea501b0bbf1435e833b41cb Mon Sep 17 00:00:00 2001 From: ".CooCoo" Date: Sun, 8 Nov 2026 16:44:04 -0800 Subject: [PATCH 32/75] Scroll: a touch zoom keeps zooming after its keys are let go (#6130) --- crates/ui-egui/src/canvas.rs | 33 +++++---- crates/ui-egui/src/platform/motion.rs | 76 ++++++++++++++++++--- crates/ui-egui/src/platform/motion/tests.rs | 31 +++++++-- crates/ui-egui/tests/it/view.rs | 45 ++++++++++++ 4 files changed, 152 insertions(+), 33 deletions(-) diff --git a/crates/ui-egui/src/canvas.rs b/crates/ui-egui/src/canvas.rs index b5cfd14910..dd0fa285a0 100644 --- a/crates/ui-egui/src/canvas.rs +++ b/crates/ui-egui/src/canvas.rs @@ -2141,19 +2141,18 @@ fn single_page_paging(ui: &egui::Ui, view: &mut DocView, cx: &PagingCx<'_>) -> ( single_paging = can_turn && fits; // Every wheel event goes to the pager, so it follows each trackpad touch to its end // even while the page can't turn. - ui.input(|i| { - for e in &i.events { - if let egui::Event::MouseWheel { unit, delta, phase, modifiers, .. } = e { - // Zooming (⌘ or Ctrl) and sideways scrolling (Shift) never turn pages. - let dy = if modifiers.command || modifiers.ctrl || modifiers.shift { 0.0 } else { delta.y }; - if fits { - view.single_page_wheel(*unit, dy, *phase, i.time, can_turn); - } else { - edge_turned |= view.single_page_wheel_edge(*unit, dy, *phase, i.time, can_turn); - } + let now = ui.input(|i| i.time); + for e in &crate::motion::gesture_wheel_events(ui.ctx()) { + if let egui::Event::MouseWheel { unit, delta, phase, modifiers, .. } = e { + // Zooming (⌘ or Ctrl) and sideways scrolling (Shift) never turn pages. + let dy = if modifiers.command || modifiers.ctrl || modifiers.shift { 0.0 } else { delta.y }; + if fits { + view.single_page_wheel(*unit, dy, *phase, now, can_turn); + } else { + edge_turned |= view.single_page_wheel_edge(*unit, dy, *phase, now, can_turn); } } - }); + } if single_paging { // Paging owns vertical wheel motion here. A diagonal gesture keeps its sideways // part, as `ScrollArea` itself handles each axis. @@ -2202,10 +2201,10 @@ fn read_wheel(ui: &mut egui::Ui, view: &mut DocView, avail: Rect, single_paging: } return Wheel { precise_scroll: Vec2::ZERO, line_base: Vec2::ZERO, coarse_scroll: Vec2::ZERO, has_page: false, paging_owns_vertical }; } - let consumed = ui.input(|i| { - i.events.iter().any(|event| { - matches!(event, egui::Event::MouseWheel { delta, modifiers, .. } if *delta != Vec2::ZERO && delta.is_finite() && !modifiers.command && !modifiers.ctrl) - }) + // A touch gesture zooms to its end once it began with ⌘ or Ctrl, the keys let go or not. + let events = crate::motion::gesture_wheel_events(ui.ctx()); + let consumed = events.iter().any(|event| { + matches!(event, egui::Event::MouseWheel { delta, modifiers, .. } if *delta != Vec2::ZERO && delta.is_finite() && !modifiers.command && !modifiers.ctrl) }); if consumed { wheel_reset::consumed(ui.ctx()); @@ -2214,7 +2213,7 @@ fn read_wheel(ui: &mut egui::Ui, view: &mut DocView, avail: Rect, single_paging: // Match the line distance used by egui's other scroll areas. let live = ui.ctx().options(|o| o.input_options.line_scroll_speed); let line_points = if live.is_finite() && live > 0.0 { live } else { crate::scroll_speed::EGUI_LINE_POINTS }; - ui.input(|i| crate::wheel_boost::accumulate(&i.events, line_points, single_paging)) + crate::wheel_boost::accumulate(&events, line_points, single_paging) }; let mut precise_scroll = input.precise; let mut line_base = input.line; @@ -2870,7 +2869,7 @@ struct GlideCx { fn step_glide(ui: &egui::Ui, view: &mut DocView, avail: Rect, cx: &GlideCx) -> Vec2 { let GlideCx { now, may_glide, key_scroll, key_side, auto_delta, zoom_delta, edge_delta, wheel_scrolled } = *cx; let over_page = ui.rect_contains_pointer(avail); - ui.input(|i| view.fling.read(&i.events, now, may_glide && over_page)); + view.fling.read(&crate::motion::gesture_wheel_events(ui.ctx()), now, may_glide && over_page); let interrupted = glide_interrupted(ui); if !may_glide || interrupted || key_scroll != 0.0 || key_side != 0.0 || auto_delta != Vec2::ZERO || zoom_delta != 1.0 { view.fling.stop(); diff --git a/crates/ui-egui/src/platform/motion.rs b/crates/ui-egui/src/platform/motion.rs index b1d615df14..79878de20b 100644 --- a/crates/ui-egui/src/platform/motion.rs +++ b/crates/ui-egui/src/platform/motion.rs @@ -19,7 +19,7 @@ //! How far anything scrolls or zooms doesn't change, only how long it takes. use egui::style::ScrollAnimation; -use egui::{Event, Modifiers, MouseWheelUnit, TouchPhase, Vec2, vec2}; +use egui::{Event, Modifiers, MouseWheelSource, MouseWheelUnit, TouchPhase, Vec2, vec2}; /// Preferences ▸ Animation effects: whether anything in the window is animated, the wheel's /// scrolling and zooming included. @@ -119,12 +119,6 @@ impl Motion { } let animate = effects.animates(self.reduced); self.sync_style(ctx, animate); - if animate { - // A gesture begun while easing is egui's to finish. - self.wheel = WheelReader::default(); - self.frame = None; - return; - } if crate::canvas::wheel_reset::reset_this_pass(ctx) { // The document's hidden cancel resets egui before fresh input. Its app reader // must forget the same older gesture modifiers before reading those events. @@ -146,11 +140,18 @@ impl Motion { let motion = match self.frame { Some((read_in, motion)) if read_in == frame => motion, _ => { - let motion = ctx.input(|i| self.wheel.read(&i.events, i.viewport_rect().height(), &options)); + let motion = ctx.input(|i| self.wheel.read(&i.events, i.time, i.viewport_rect().height(), &options)); self.frame = Some((frame, motion)); + let modifiers = self.wheel.modifiers.clone(); + ctx.data_mut(|d| d.insert_temp(gesture_id(), (frame, modifiers))); motion } }; + if animate { + // egui eases the wheel itself. The reader still follows each touch gesture, as + // egui does, for the modifiers [`gesture_wheel_events`] hands the document. + return; + } let (scroll, zoom) = if first_pass { (motion.scroll, motion.zoom) } else { (Vec2::ZERO, 1.0) }; ctx.input_mut(|i| i.smooth_scroll_delta = scroll); let pass = ctx.cumulative_pass_nr(); @@ -202,6 +203,34 @@ pub(crate) fn zoom_delta(ctx: &egui::Context) -> f32 { }) } +/// Return the stable egui id holding the modifiers of this frame's wheel events. +fn gesture_id() -> egui::Id { + egui::Id::unique("pdfcraft-motion-wheel-gesture") +} + +/// This pass's wheel events, each with the modifiers it acts with: a touch gesture keeps +/// ⌘/Ctrl or Shift from its `Start` until its `End`, as egui and [`WheelReader`] read it. A view +/// that reads the raw wheel then scrolls none of a gesture that egui or [`zoom_delta`] zooms, +/// after the keys are let go too. The events keep their own modifiers when +/// [`Motion::begin_pass`] did not read this frame's wheel. +pub(crate) fn gesture_wheel_events(ctx: &egui::Context) -> Vec { + let read: Option<(u64, Vec)> = ctx.data(|d| d.get_temp(gesture_id())); + let frame = ctx.cumulative_frame_nr(); + let mut events: Vec = ctx.input(|i| i.events.iter().filter(|e| matches!(e, Event::MouseWheel { .. })).cloned().collect()); + if let Some((at, held)) = read + && at == frame + // A later pass of the frame has no events left; only the pass that was read matches. + && events.len() == held.len() + { + for (event, held) in events.iter_mut().zip(held) { + if let Event::MouseWheel { modifiers, .. } = event { + *modifiers = held; + } + } + } + events +} + /// One pass's wheel input. #[derive(Clone, Copy, Debug, PartialEq)] struct WheelMotion { @@ -221,20 +250,44 @@ struct WheelReader { /// The modifiers of the touch gesture in progress (`TouchPhase::Start` to `End`). They hold /// for the whole gesture, as in egui: its momentum arrives after the keys are let go. touch: Option, + /// The time of the last wheel event and what drove it. A touch whose `End` never arrives + /// (some mouse wheels on Wayland send `Start` alone) ends after egui's timeout. + last: Option<(f64, MouseWheelSource)>, + /// The modifiers each wheel event of the last [`Self::read`] acted with, in order. + modifiers: Vec, } +/// egui's `SCROLL_ACTION_TIMEOUT`: seconds without a wheel event that end a touch. +const TOUCH_TIMEOUT: f64 = 0.150; + +/// egui's `RESTING_FINGER_TIMEOUT`: the same for a touch driven by fingers on a trackpad. +const RESTING_FINGER_TIMEOUT: f64 = 1.0; + impl WheelReader { - /// `page_height` is what one `MouseWheelUnit::Page` scrolls: the window's height in points. - fn read(&mut self, events: &[Event], page_height: f32, options: &egui::InputOptions) -> WheelMotion { + /// `time` is the pass's input time; `page_height` is what one `MouseWheelUnit::Page` + /// scrolls: the window's height in points. + fn read(&mut self, events: &[Event], time: f64, page_height: f32, options: &egui::InputOptions) -> WheelMotion { let mut delta = Vec2::ZERO; let mut zoom = 1.0_f32; let mut modifiers = self.touch.unwrap_or_default(); + self.modifiers.clear(); for event in events { + if let Event::MouseWheel { source, .. } = event { + if let Some((at, last_source)) = self.last { + let timeout = if last_source == MouseWheelSource::Trackpad { RESTING_FINGER_TIMEOUT } else { TOUCH_TIMEOUT }; + // `>` is false for a NaN time, which then keeps the touch, as egui does. + if time - at > timeout { + self.touch = None; + } + } + self.last = Some((time, *source)); + } match event { Event::MouseWheel { unit, delta: step, phase, modifiers: latest, .. } => match phase { TouchPhase::Start => { self.touch = Some(*latest); modifiers = *latest; + self.modifiers.push(modifiers); } TouchPhase::Move => { modifiers = match &mut self.touch { @@ -244,6 +297,7 @@ impl WheelReader { } None => *latest, }; + self.modifiers.push(modifiers); let mut step = match unit { MouseWheelUnit::Point => *step, MouseWheelUnit::Line => options.line_scroll_speed * *step, @@ -264,6 +318,8 @@ impl WheelReader { } // egui drops what the gesture moved earlier in the pass it ends in. TouchPhase::End | TouchPhase::Cancel => { + // The end belongs to the gesture: it carries the gesture's modifiers. + self.modifiers.push(self.touch.unwrap_or(*latest)); self.touch = None; modifiers = Modifiers::NONE; delta = Vec2::ZERO; diff --git a/crates/ui-egui/src/platform/motion/tests.rs b/crates/ui-egui/src/platform/motion/tests.rs index 2db565df6a..4ccb3c0230 100644 --- a/crates/ui-egui/src/platform/motion/tests.rs +++ b/crates/ui-egui/src/platform/motion/tests.rs @@ -11,7 +11,7 @@ fn notch(dy: f32, modifiers: Modifiers) -> Event { } fn read(events: &[Event]) -> WheelMotion { - WheelReader::default().read(events, 600.0, &egui::InputOptions::default()) + WheelReader::default().read(events, 0.0, 600.0, &egui::InputOptions::default()) } #[test] @@ -46,14 +46,33 @@ fn a_touch_gesture_keeps_its_modifiers_and_ends_clean() { let options = egui::InputOptions::default(); let point = |dy: f32, phase, modifiers| wheel(MouseWheelUnit::Point, vec2(0.0, dy), phase, modifiers); // ⌘ held as the fingers land, let go before the momentum: still zooming. - let began = reader.read(&[point(0.0, TouchPhase::Start, Modifiers::COMMAND), point(20.0, TouchPhase::Move, Modifiers::COMMAND)], 600.0, &options); + let began = + reader.read(&[point(0.0, TouchPhase::Start, Modifiers::COMMAND), point(20.0, TouchPhase::Move, Modifiers::COMMAND)], 0.0, 600.0, &options); assert!(began.zoom > 1.0 && began.scroll == Vec2::ZERO, "{began:?}"); - let momentum = reader.read(&[point(20.0, TouchPhase::Move, Modifiers::NONE)], 600.0, &options); + let momentum = reader.read(&[point(20.0, TouchPhase::Move, Modifiers::NONE)], 0.0, 600.0, &options); assert!(momentum.zoom > 1.0 && momentum.scroll == Vec2::ZERO, "{momentum:?}"); + assert_eq!(reader.modifiers, vec![Modifiers::COMMAND], "the momentum acts with the gesture's modifiers"); // The pass it ends in moves nothing, and the next gesture starts without the modifier. - let ended = reader.read(&[point(20.0, TouchPhase::Move, Modifiers::NONE), point(0.0, TouchPhase::End, Modifiers::NONE)], 600.0, &options); + let ended = reader.read(&[point(20.0, TouchPhase::Move, Modifiers::NONE), point(0.0, TouchPhase::End, Modifiers::NONE)], 0.0, 600.0, &options); assert_eq!(ended, WheelMotion { scroll: Vec2::ZERO, zoom: 1.0 }); - assert_eq!(reader.read(&[point(-20.0, TouchPhase::Move, Modifiers::NONE)], 600.0, &options).scroll, vec2(0.0, -20.0)); + assert_eq!(reader.read(&[point(-20.0, TouchPhase::Move, Modifiers::NONE)], 0.0, 600.0, &options).scroll, vec2(0.0, -20.0)); +} + +/// A touch whose `End` never arrives (some mouse wheels on Wayland) ends after egui's timeout, +/// so its modifiers don't latch onto later scrolling. +#[test] +fn a_touch_without_an_end_lets_its_modifiers_go_after_a_pause() { + let options = egui::InputOptions::default(); + for (source, pause) in [(egui::MouseWheelSource::Unknown, 0.2), (egui::MouseWheelSource::Trackpad, 1.1)] { + let mut reader = WheelReader::default(); + let point = |dy: f32, phase, modifiers| Event::MouseWheel { unit: MouseWheelUnit::Point, delta: vec2(0.0, dy), phase, modifiers, source }; + reader.read(&[point(0.0, TouchPhase::Start, Modifiers::COMMAND), point(20.0, TouchPhase::Move, Modifiers::COMMAND)], 1.0, 600.0, &options); + let soon = reader.read(&[point(20.0, TouchPhase::Move, Modifiers::NONE)], 1.1, 600.0, &options); + assert!(soon.zoom > 1.0 && soon.scroll == Vec2::ZERO, "{source:?}: still the same touch: {soon:?}"); + let later = reader.read(&[point(-20.0, TouchPhase::Move, Modifiers::NONE)], 1.1 + pause, 600.0, &options); + assert_eq!(later, WheelMotion { scroll: vec2(0.0, -20.0), zoom: 1.0 }, "{source:?}"); + assert_eq!(reader.modifiers, vec![Modifiers::NONE], "{source:?}"); + } } #[test] @@ -103,7 +122,7 @@ fn reading_at_once_goes_as_far_as_egui_eases() { }); } let page_height = ctx.input(|i| i.viewport_rect().height()); - let at_once = WheelReader::default().read(&[event], page_height, &options); + let at_once = WheelReader::default().read(&[event], 0.0, page_height, &options); assert!((scrolled - at_once.scroll).length() < 0.01, "{modifiers:?} {unit:?}: egui scrolled {scrolled:?}, read {:?}", at_once.scroll); assert!((zoomed / at_once.zoom - 1.0).abs() < 1e-3, "{modifiers:?} {unit:?}: egui zoomed {zoomed}, read {}", at_once.zoom); // What makes the reader necessary: egui's first pass holds part of the step back. diff --git a/crates/ui-egui/tests/it/view.rs b/crates/ui-egui/tests/it/view.rs index f6384127e8..d912fa18a4 100644 --- a/crates/ui-egui/tests/it/view.rs +++ b/crates/ui-egui/tests/it/view.rs @@ -1016,6 +1016,51 @@ fn the_wheel_zooms_a_whole_step_at_once_with_animation_effects_off() { } } +/// A touch gesture begun with ⌘/Ctrl zooms to its end after the keys are let go, as egui reads +/// it: its later motion neither scrolls the page as well nor turns the rest of it into scrolling. +#[test] +fn a_zoom_gesture_zooms_to_its_end_after_its_keys_are_let_go() { + /// The view's zoom and page 1 on screen, settled, after a gesture of six 5-point motions, + /// the keys let go after the second one or held throughout. + fn zoom_gesture(effects: &'static str, let_go: bool) -> (f32, f32, egui::Rect) { + let mut h = harness_stepping(1.0 / 60.0, &[("layout", "continuous"), ("zoom", "300")]); + h.state_mut().set_option("animation_effects", effects).unwrap(); + let at = h.state().views[0].viewport_rect().center(); + h.hover_at(at); + h.run_steps(2); + let before = h.state().views[0].zoom; + let touch = |phase, dy, modifiers| egui::Event::MouseWheel { + unit: MouseWheelUnit::Point, + delta: egui::vec2(0.0, dy), + phase, + modifiers, + source: egui::MouseWheelSource::Unknown, + }; + h.event(touch(TouchPhase::Start, 0.0, Modifiers::COMMAND)); + h.step(); + for k in 0..6 { + let modifiers = if let_go && k >= 2 { Modifiers::NONE } else { Modifiers::COMMAND }; + h.event(touch(TouchPhase::Move, 5.0, modifiers)); + h.step(); + } + h.event(touch(TouchPhase::End, 0.0, Modifiers::NONE)); + h.run_steps(40); + (before, h.state().views[0].zoom, rect(&h, 0).expect("page 1 on screen")) + } + for effects in ["off", "on"] { + let (before, held, held_page) = zoom_gesture(effects, false); + assert!(held > before, "{effects}: the gesture zoomed in: {before} to {held}"); + let (_, let_go, let_go_page) = zoom_gesture(effects, true); + assert!((let_go / held - 1.0).abs() < 1e-4, "{effects}: the whole gesture zooms: {let_go}, held throughout {held}"); + assert!( + (let_go_page.min - held_page.min).length() < 0.5, + "{effects}: and nothing scrolls: page 1 at {:?}, held throughout {:?}", + let_go_page.min, + held_page.min + ); + } +} + #[test] fn page_display_commands_switch_layouts() { let mut h = harness(&[]); From 5d3f3a2e42414dd7ab13e0e3b9fb943b0f99ee92 Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:44:20 +0800 Subject: [PATCH 33/75] fix(export): skip empty Markdown runs before merging and emphasis lookahead (#6134) --- crates/export/src/as_markdown.rs | 41 ++++++++++++++++++++++++++------ 1 file changed, 34 insertions(+), 7 deletions(-) diff --git a/crates/export/src/as_markdown.rs b/crates/export/src/as_markdown.rs index a5c986c6b3..949264494d 100644 --- a/crates/export/src/as_markdown.rs +++ b/crates/export/src/as_markdown.rs @@ -60,20 +60,23 @@ pub(crate) fn emphasis(text: &str, bold: bool, italic: bool, open: bool, close: pub(crate) fn paragraph(runs: &[(&str, bool, bool)]) -> String { let mut merged: Vec<(String, bool, bool)> = Vec::new(); for &(text, bold, italic) in runs { + let text = md_text(text); + if text.is_empty() { + continue; + } match merged.last_mut() { - Some(last) if last.1 == bold && last.2 == italic => last.0.push_str(text), - _ => merged.push((text.to_string(), bold, italic)), + Some(last) if last.1 == bold && last.2 == italic => last.0.push_str(&text), + _ => merged.push((text, bold, italic)), } } - let texts: Vec = merged.iter().map(|(text, ..)| md_text(text)).collect(); let mut out = String::new(); - for (k, ((_, bold, italic), text)) in merged.iter().zip(&texts).enumerate() { + for (k, (text, bold, italic)) in merged.iter().enumerate() { let core = text.trim(); let open = marks(core.chars().next(), out.chars().last()); // Markers of a styled neighbour right next to these would join them into one run. - let close = match (merged.get(k.saturating_add(1)), texts.get(k.saturating_add(1))) { - (Some((_, b, i)), Some(next)) if (*b || *i) && !next.starts_with(char::is_whitespace) => false, - (_, next) => marks(core.chars().last(), next.and_then(|n| n.chars().next())), + let close = match merged.get(k.saturating_add(1)) { + Some((next, b, i)) if (*b || *i) && !next.starts_with(char::is_whitespace) => false, + next => marks(core.chars().last(), next.and_then(|(n, ..)| n.chars().next())), }; out.push_str(&emphasis(text, *bold, *italic, open, close)); } @@ -149,3 +152,27 @@ pub fn markdown(pages: &[Page]) -> String { s.push('\n'); s } + +#[cfg(test)] +mod tests { + use super::paragraph; + + #[test] + fn empty_or_control_run_does_not_split_styled_runs() { + assert_eq!(paragraph(&[("hello", true, false), ("", false, false), ("world", true, false)]), "**helloworld**"); + assert_eq!(paragraph(&[("hello", true, false), ("\u{1}", false, false), ("world", true, false)]), "**helloworld**"); + } + + #[test] + fn empty_styled_run_does_not_affect_neighbor_emphasis() { + assert_eq!(paragraph(&[("x", true, false), ("", false, true)]), "**x**"); + assert_eq!(paragraph(&[("foo", false, false), (".", true, false), ("", true, true), ("bar", false, false)]), "foo.bar"); + // Empty/control runs between differently styled runs still look ahead to the next non-empty run. + assert_eq!(paragraph(&[("bold", true, false), ("", false, false), ("italic", false, true)]), "bold*italic*"); + assert_eq!(paragraph(&[("bold", true, false), ("\u{1}", false, false), (" italic", false, true)]), "**bold** *italic*"); + // Leading empty/control runs do not affect the opening delimiter. + assert_eq!(paragraph(&[("", false, true), ("\u{1}", true, false), ("lead", true, false)]), "**lead**"); + // All-empty/control paragraphs produce empty output. + assert_eq!(paragraph(&[("", true, false), ("\u{1}", false, true), ("\u{7f}", false, false)]), ""); + } +} From 5ee4a85876d0254e72bb21cdc6cd5c9b9bf63988 Mon Sep 17 00:00:00 2001 From: jonathanwxh-cell Date: Mon, 9 Nov 2026 08:44:34 +0800 Subject: [PATCH 34/75] fix(sign): merge CMS unsignedAttrs into one SET and validate SignerIdentifier (#6137) --- crates/sign/src/cms.rs | 224 +++++++++++++++++++++++++++++++---------- 1 file changed, 172 insertions(+), 52 deletions(-) diff --git a/crates/sign/src/cms.rs b/crates/sign/src/cms.rs index 4e232af616..ee1882ccdc 100644 --- a/crates/sign/src/cms.rs +++ b/crates/sign/src/cms.rs @@ -230,66 +230,94 @@ impl SignedData { } } -fn strip(b: &[u8]) -> &[u8] { - match b { - [0, rest @ ..] if !rest.is_empty() => rest, - _ => b, +fn strip(mut b: &[u8]) -> &[u8] { + while let [0, rest @ ..] = b + && !rest.is_empty() + { + b = rest; } + b +} + +fn parse_signer_id(sid: &Tlv<'_>) -> Result { + match sid.tag { + tag::SEQUENCE => { + let p = sid.children()?; + let [issuer, serial] = p.as_slice() else { return Err(bad("issuerAndSerialNumber")) }; + if issuer.tag != tag::SEQUENCE || serial.tag != tag::INTEGER || serial.value.is_empty() { + return Err(bad("issuerAndSerialNumber")); + } + Ok(SignerId::IssuerSerial { issuer: issuer.raw.to_vec(), serial: serial.value.to_vec() }) + } + t if t == tag::ctx_prim(0) => Ok(SignerId::KeyId(sid.value.to_vec())), + t if t == tag::ctx(0) => { + let key_id = match sid.inner() { + Ok(inner) if matches!(inner.tag, tag::OCTET_STRING | tag::OCTET_STRING_SEGMENTS) => inner.octets("sid")?.into_owned(), + _ => { + let mut seg = *sid; + seg.tag = tag::OCTET_STRING_SEGMENTS; + seg.octets("sid")?.into_owned() + } + }; + Ok(SignerId::KeyId(key_id)) + } + _ => Err(bad("sid")), + } +} + +type SignedAttrs = (Vec, Option>, Option, Option