diff --git a/crates/cos/src/document.rs b/crates/cos/src/document.rs index 7a2d1edcfb..a053ea150a 100644 --- a/crates/cos/src/document.rs +++ b/crates/cos/src/document.rs @@ -20,6 +20,7 @@ mod state; mod xref; pub use objects::DECODE_PARMS_KEYS; +pub(crate) use objects::Held; #[cfg(test)] pub(crate) use reader::find; #[cfg(test)] @@ -63,6 +64,10 @@ pub struct Document { security: Option>, /// Object number of the `/Encrypt` dictionary (never encrypted itself). encrypt_num: Option, + /// What that dictionary and `/ID` refer to, which the key comes from, and so is as stored. + encrypt_members: Arc>, + /// Object numbers a scan of the file didn't find: the file doesn't change, so it never will. + unfound: Arc>>, /// Encryption was added, changed or removed since opening: only a full save can apply it. encryption_changed: bool, /// Set by edits that must not leave earlier revisions in the file (redaction). diff --git a/crates/cos/src/document/objects.rs b/crates/cos/src/document/objects.rs index 9f55f65fff..911a6704f5 100644 --- a/crates/cos/src/document/objects.rs +++ b/crates/cos/src/document/objects.rs @@ -156,12 +156,16 @@ impl Document { self.try_get(r.num).unwrap_or_else(|_| Arc::new(Object::Null)) } - /// Follow a reference if `o` is one; otherwise return `o` itself. + /// Follow a reference if `o` is one; otherwise return `o` itself. An object can be nothing but + /// a reference to another, which is followed too, up to [`MAX_REFERENCE_HOPS`] of them. pub fn resolve(&self, o: &Object) -> Arc { - match o { - Object::Ref(r) => self.get(*r), - other => Arc::new(other.clone()), + let Object::Ref(r) = o else { return Arc::new(o.clone()) }; + let mut held = self.get(*r); + for _ in 0..MAX_REFERENCE_HOPS { + let Object::Ref(r) = &*held else { break }; + held = self.get(*r); } + held } /// `s` with the `/DecodeParms` its filters use made direct, so [`Stream::decoded`] reads them: @@ -218,7 +222,7 @@ impl Document { } /// `o`, or the object it refers to, without copying either. - fn held<'o>(&self, o: &'o Object) -> Held<'o> { + pub(crate) fn held<'o>(&self, o: &'o Object) -> Held<'o> { match o { Object::Ref(_) => Held::Resolved(self.resolve(o)), o => Held::Direct(o), @@ -384,6 +388,7 @@ impl Document { // cycles such as `6 0 obj << /Length 6 0 R >>`. Bound the nesting on every path. let _guard = LoadGuard::enter(num).ok_or_else(|| CosError::Syntax { offset: 0, detail: format!("reference cycle while loading object {num}") })?; + let _scope = crate::security::FollowScope::enter(); let obj = self.load(num, 0)?; let obj = Arc::new(obj); // The editing cache is bounded like the scan reader's: re-parsing a file-backed view @@ -551,7 +556,11 @@ impl Document { let Object::Stream(s) = &*loaded else { return Err(CosError::MissingObject(num)) }; let n = s.dict.int(b"N").unwrap_or(0).clamp(0, 1_000_000) as usize; let first = usize::try_from(s.dict.int(b"First").unwrap_or(0).max(0)).ok(); - let data = s.decoded_within(self.stream_limit)?; + // Its parameters can be given by reference, and can sit in another object stream. A scan of + // the file's object streams comes here without a read around it. + let _scope = crate::security::FollowScope::enter(); + let readable = crate::security::following(Cow::Borrowed(s), || self.with_direct_parms(s)); + let data = readable.decoded_within(self.stream_limit)?; let mut lx = Lexer::new(&data, 0); let mut index = Vec::with_capacity(n); for _ in 0..n { @@ -566,8 +575,14 @@ impl Document { Ok(stm) } - /// Last-resort lookup: scan the file for `num G obj`. + /// Last-resort lookup: scan the file for `num G obj`. What isn't there is remembered: each miss + /// reads the whole file, and a file can refer to such an object from everywhere. fn scan_for(&self, num: u32) -> Option<(ObjRef, Object)> { + if self.unfound.lock().is_ok_and(|unfound| unfound.contains(&num)) { + return None; + } + #[cfg(test)] + super::tests::FILE_SCANS.with(|scans| scans.set(scans.get() + 1)); let needle = format!("{num} "); let data = &self.data; let mut found = None; @@ -583,6 +598,11 @@ impl Document { found = Some((id, o)); // keep the last (newest) definition } } + if found.is_none() + && let Ok(mut unfound) = self.unfound.lock() + { + unfound.insert(num); + } found } @@ -775,9 +795,13 @@ pub const DECODE_PARMS_KEYS: [&[u8]; 13] = [ b"ColorTransform", ]; +/// How many references in a row [`Document::resolve`] follows: enough for any file, and it ends a +/// cycle of them. +const MAX_REFERENCE_HOPS: usize = 32; + /// An object read in place, or the object a reference names as the document holds it: either way /// nothing is copied ([`Document::held`]). -enum Held<'o> { +pub(crate) enum Held<'o> { Direct(&'o Object), Resolved(Arc), } diff --git a/crates/cos/src/document/open.rs b/crates/cos/src/document/open.rs index 395424b771..70bbcc313e 100644 --- a/crates/cos/src/document/open.rs +++ b/crates/cos/src/document/open.rs @@ -1,7 +1,11 @@ use super::reader::find; use super::state::Credentials; use super::xref::XrefIndex; -use super::{Arc, BTreeMap, CosError, Dict, Document, ObjRef, Object}; +use super::{Arc, BTreeMap, CosError, Dict, Document, HashSet, ObjRef, Object}; + +/// More objects than `/Encrypt` and `/ID` refer to in any file, so that one which refers to the page +/// tree, say, doesn't have a whole document read to open it. +const MAX_ENCRYPT_MEMBERS: usize = 64; impl Document { /// A new document with an empty page tree (for split / extract / combine). It has no @@ -27,6 +31,8 @@ impl Document { version: "1.7".into(), security: None, encrypt_num: None, + encrypt_members: Arc::default(), + unfound: Arc::default(), encryption_changed: false, full_save: false, out_security: None, @@ -111,6 +117,8 @@ impl Document { version, security: None, encrypt_num: None, + encrypt_members: Arc::default(), + unfound: Arc::default(), encryption_changed: false, full_save: false, out_security: None, @@ -185,6 +193,8 @@ impl Document { }; let params = crate::security::encrypt_dict(self, &dict); let id0 = self.first_id(); + // Read here, while nothing is decrypted yet. + let members = self.encryption_members(enc); let handler = match credentials { Credentials::Password(password) => Arc::new(pdfcraft_crypt::SecurityHandler::open(params, &id0, *password).map_err(|e| match e { pdfcraft_crypt::CryptError::WrongPassword if password.is_none() => CosError::NeedsPassword, @@ -197,6 +207,7 @@ impl Document { }; self.security = Some(handler); self.encrypt_num = num; + self.encrypt_members = Arc::new(members); // Objects read while locating the catalog were read before decryption was possible. self.cache = Arc::default(); self.objstms = Arc::default(); @@ -205,10 +216,42 @@ impl Document { /// The first file identifier, part of the encryption key (empty when there is none). fn first_id(&self) -> Vec { - match self.trailer.get(b"ID") { - Some(Object::Array(a)) => a.first().and_then(|s| s.as_string()).map(|s| s.bytes.clone()).unwrap_or_default(), - _ => Vec::new(), + let id = self.trailer.get(b"ID").map(|id| self.resolve(id)); + let first = id.as_deref().and_then(Object::as_array).and_then(|a| a.first()).map(|s| self.resolve(s)); + first.as_deref().and_then(Object::as_string).map(|s| s.bytes.clone()).unwrap_or_default() + } + + /// The objects `/Encrypt`, given as `enc`, and `/ID` refer to, the ones they refer to, and so + /// on, up to [`MAX_ENCRYPT_MEMBERS`]. Their strings aren't encrypted (ISO 32000-2 §7.6.2): the + /// key is made from them, so they have to read as stored, and stay so. A number with no object + /// behind it, and a stream, which those dictionaries never hold, isn't one: an object added + /// under that number later, or a page's content a hostile `/Encrypt` names, is encrypted. + fn encryption_members(&self, enc: &Object) -> HashSet { + let mut pending = Vec::new(); + crate::writer::collect_refs(enc, &mut |r| pending.push(r)); + if let Some(id) = self.trailer.get(b"ID") { + crate::writer::collect_refs(id, &mut |r| pending.push(r)); } + let mut members = HashSet::new(); + while let Some(r) = pending.pop() { + if members.len() >= MAX_ENCRYPT_MEMBERS || members.contains(&r.num) { + continue; + } + let held = self.get(r); + if matches!(&*held, Object::Null | Object::Stream(_)) { + continue; + } + members.insert(r.num); + crate::writer::collect_refs(&held, &mut |x| pending.push(x)); + } + members + } + + /// Whether object `num` is written as it is held, being one of [`Document::encryption_members`], + /// unedited, and the save writing the same `/Encrypt`. What was edited or added is encrypted + /// whatever its number. + pub(crate) fn stored_as_is(&self, num: u32) -> bool { + !self.encryption_changed && !self.is_edited(num) && self.encrypt_members.contains(&num) } /// Lift the restrictions of a document opened with its user password by entering the @@ -270,8 +313,9 @@ impl Document { /// always a full rewrite. Returns the handler (authenticated as owner). pub fn set_encryption(&mut self, params: &pdfcraft_crypt::NewEncryption) -> Result<(), CosError> { // The file identifier is part of the key; make sure it exists and keep it. - let id0 = match self.trailer.get(b"ID") { - Some(Object::Array(a)) if a.len() == 2 => a[0].as_string().map(|s| s.bytes.clone()).unwrap_or_default(), + let id = self.trailer.get(b"ID").map(|id| self.resolve(id)); + let id0 = match id.as_deref() { + Some(Object::Array(a)) if a.len() == 2 => self.first_id(), _ => { let id = pdfcraft_crypt::file_id(¶ms.seed); let s = Object::String(crate::PdfString { bytes: id.clone(), hex: true }); @@ -338,7 +382,9 @@ impl Document { /// Decrypt a freshly parsed indirect object if the document is encrypted. pub(super) fn decrypted(&self, id: ObjRef, o: Object) -> Object { match &self.security { - Some(h) if Some(id.num) != self.encrypt_num => crate::security::transform(h, &o, id.num, id.generation, crate::security::Pass::Decrypt), + Some(h) if Some(id.num) != self.encrypt_num && !self.encrypt_members.contains(&id.num) => { + crate::security::transform(self, h, &o, id.num, id.generation, crate::security::Pass::Decrypt) + } _ => o, } } diff --git a/crates/cos/src/document/tests.rs b/crates/cos/src/document/tests.rs index 77c0e9f3a7..6d9c98eb95 100644 --- a/crates/cos/src/document/tests.rs +++ b/crates/cos/src/document/tests.rs @@ -21,6 +21,148 @@ fn what_is_reachable_is_what_a_full_save_keeps() { assert!(holds(7) && !holds(8)); } +std::thread_local! { + /// How many times `Document::scan_for` read the file on this thread. + pub(super) static FILE_SCANS: std::cell::Cell = const { std::cell::Cell::new(0) }; +} + +/// A scan that doesn't find an object isn't made again: each reads the whole file, and a file can +/// refer to such an object from every object it has. +#[test] +fn what_a_scan_of_the_file_did_not_find_is_remembered() { + let body = b"%PDF-1.7\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n2 0 obj << /Type /Pages /Kids [] /Count 0 >> endobj\n"; + let mut out = body.to_vec(); + let xref = out.len(); + out.extend_from_slice(b"xref\n0 4\n0000000000 65535 f \n"); + out.extend_from_slice(format!("{:010} 00000 n \n", 9).as_bytes()); + out.extend_from_slice(format!("{:010} 00000 n \n", body.iter().position(|b| *b == b'\n').unwrap() + 49).as_bytes()); + out.extend_from_slice(b"0000000001 00000 n \n"); + out.extend_from_slice(format!("trailer << /Size 4 /Root 1 0 R >>\nstartxref\n{xref}\n%%EOF\n").as_bytes()); + let doc = Document::open(Arc::new(out)).unwrap(); + FILE_SCANS.with(|scans| scans.set(0)); + for _ in 0..5 { + assert!(doc.try_get(3).is_err()); + } + assert_eq!(FILE_SCANS.with(std::cell::Cell::get), 1); + let clone = doc.clone(); + assert!(clone.try_get(3).is_err()); + assert_eq!(FILE_SCANS.with(std::cell::Cell::get), 1, "a clone has the same file"); +} + +/// An object can be nothing but a reference to another. A chain of them is followed, and one that +/// goes round ends in what isn't a dictionary or a name, without hanging. +#[test] +fn references_to_references_are_followed() { + let mut doc = Document::new_empty(); + let name = doc.add(Object::name("Target")); + let one = doc.add(Object::Ref(name)); + let two = doc.add(Object::Ref(one)); + assert_eq!(doc.resolve(&Object::Ref(two)).as_name(), Some(b"Target".as_slice())); + let a = doc.add(Object::Null); + let b = doc.add(Object::Ref(a)); + doc.set(a, Object::Ref(b)); + assert_eq!(doc.resolve(&Object::Ref(a)).as_name(), None); +} + +/// An object stream's `/DecodeParms` can be given by reference: without them a predictor isn't +/// undone, so its members were lost. +#[test] +fn an_object_stream_with_parameters_by_reference_is_read() { + let members = b"3 0 << /Hidden (yes) >>"; + // PNG Sub, one row. + let mut rows = vec![1u8]; + let mut before = 0u8; + for b in members { + rows.push(b.wrapping_sub(before)); + before = *b; + } + let coded = pdfcraft_filters::encode_flate(&rows); + let mut out = b"%PDF-1.5\n".to_vec(); + let mut offs = [0usize; 7]; + let mut obj = |out: &mut Vec, n: usize, body: &[u8]| { + offs[n] = out.len(); + out.extend_from_slice(format!("{n} 0 obj\n").as_bytes()); + out.extend_from_slice(body); + out.extend_from_slice(b"\nendobj\n"); + }; + obj(&mut out, 1, b"<< /Type /Catalog /Pages 2 0 R >>"); + obj(&mut out, 2, b"<< /Type /Pages /Kids [] /Count 0 /Extra 3 0 R >>"); + let mut stm = format!("<< /Type /ObjStm /N 1 /First 4 /Filter /FlateDecode /DecodeParms 6 0 R /Length {} >>\nstream\n", coded.len()).into_bytes(); + stm.extend_from_slice(&coded); + stm.extend_from_slice(b"\nendstream"); + obj(&mut out, 4, &stm); + let data = [2u8, 0, 4, 0]; + let mut xs = format!("<< /Type /XRef /Size 7 /W [1 2 1] /Index [3 1] /Length {} >>\nstream\n", data.len()).into_bytes(); + xs.extend_from_slice(&data); + xs.extend_from_slice(b"\nendstream"); + obj(&mut out, 5, &xs); + obj(&mut out, 6, format!("<< /Predictor 11 /Columns {} >>", members.len()).as_bytes()); + let table = out.len(); + out.extend_from_slice(b"xref\n0 7\n0000000000 65535 f \n"); + for (n, off) in offs.iter().enumerate().skip(1) { + let line = if n == 3 { "0000000000 65535 f \n".to_string() } else { format!("{off:010} 00000 n \n") }; + out.extend_from_slice(line.as_bytes()); + } + out.extend_from_slice(format!("trailer << /Size 7 /Root 1 0 R /XRefStm {} >>\nstartxref\n{table}\n%%EOF\n", offs[5]).as_bytes()); + let doc = Document::open(Arc::new(out)).unwrap(); + let hidden = doc.get(ObjRef::new(3, 0)); + assert_eq!(hidden.as_dict().and_then(|d| d.get(b"Hidden")).and_then(Object::as_string).map(|s| s.bytes.clone()), Some(b"yes".to_vec())); +} + +/// Each entry of an `/Encrypt` dictionary, and of a crypt filter in it, reads the same by reference. +/// Read as absent, `/StmF` or `/CFM` was `/Identity`, so a whole file's streams stayed ciphertext. +#[test] +fn encrypt_dictionary_entries_read_the_same_by_reference() { + let string = |s: &str| Object::String(crate::PdfString::text(s)); + let mut filter = Dict::new(); + filter.set(b"CFM".to_vec(), Object::name("AESV2")); + filter.set(b"Length".to_vec(), Object::Int(5)); + let cf = |filter: Dict| { + let mut cf = Dict::new(); + cf.set(b"StdCF".to_vec(), Object::Dict(filter)); + cf.set(b"Other".to_vec(), Object::Dict(Dict::new())); + Object::Dict(cf) + }; + let mut d = Dict::new(); + let entries = [ + ("Filter", Object::name("Standard")), + ("V", Object::Int(4)), + ("R", Object::Int(4)), + ("Length", Object::Int(128)), + ("P", Object::Int(-44)), + ("O", string("owner")), + ("U", string("user")), + ("EncryptMetadata", Object::Bool(false)), + ("CF", cf(filter.clone())), + ("StmF", Object::name("StdCF")), + ("StrF", Object::name("Other")), + ("EFF", Object::name("StdCF")), + ]; + for (key, value) in &entries { + d.set(key.as_bytes().to_vec(), value.clone()); + } + let mut doc = Document::new_empty(); + let direct = crate::security::encrypt_dict(&doc, &d); + assert_eq!((direct.stm_f.as_slice(), direct.str_f.as_slice(), direct.ef_f.as_slice()), (&b"StdCF"[..], &b"Other"[..], &b"StdCF"[..])); + assert_eq!(direct.length_bits, 40, "the filter's own length, in bytes"); + assert!(!direct.encrypt_metadata); + assert_eq!(direct.crypt_filters.iter().filter(|(_, m)| *m == pdfcraft_crypt::Method::Aes128).count(), 1); + for (key, value) in &entries { + let mut one = d.clone(); + let r = doc.add(value.clone()); + one.set(key.as_bytes().to_vec(), Object::Ref(r)); + assert_eq!(crate::security::encrypt_dict(&doc, &one), direct, "/{key}"); + } + for key in ["CFM", "Length"] { + let mut by_reference = filter.clone(); + let r = doc.add(filter.get(key.as_bytes()).unwrap().clone()); + by_reference.set(key.as_bytes().to_vec(), Object::Ref(r)); + let mut one = d.clone(); + one.set(b"CF".to_vec(), cf(by_reference)); + assert_eq!(crate::security::encrypt_dict(&doc, &one), direct, "/{key} in the filter"); + } +} + /// /DecodeParms given by reference, whole, as array entries or as the numbers inside, are read /// once made direct: here a Flate predictor, which the stream alone doesn't see. Only the entries the filters use are /// resolved, and nothing is for a chain longer than a decoder takes. diff --git a/crates/cos/src/security.rs b/crates/cos/src/security.rs index 22c85c7f68..61c4f92a4c 100644 --- a/crates/cos/src/security.rs +++ b/crates/cos/src/security.rs @@ -7,10 +7,69 @@ //! false. Streams naming a crypt filter (`/Crypt` in `/Filter`) use that filter; embedded files //! use `/EFF`. -use pdfcraft_crypt::{EncryptDict, Method, SecurityHandler, StreamKind}; +use std::cell::Cell; +use pdfcraft_crypt::{EncryptDict, Method, SecurityHandler, StreamKind}; +use pdfcraft_filters::MAX_CHAIN; + +use crate::document::Held; use crate::{Dict, Document, Object, PdfString, Stream}; +/// How many references one read or write of an object may follow, with all that it loads on the +/// way. How an object is treated can be given by reference, and what that loads can do the same: a +/// file could make one read a tree of them, each object read again once the cache has let it go. +/// An object takes a few of these, and two dozen if every part of it is given by reference; past +/// this they read as absent. The parameters of an object stream, read together, count as one. +const MAX_FOLLOWED: u32 = 32; + +std::thread_local! { + /// How many [`FollowScope`]s this thread is inside. + static SCOPES: Cell = const { Cell::new(0) }; + /// The references followed in the outermost of them. + static FOLLOWED: Cell = const { Cell::new(0) }; + /// The references refused in the outermost of them, past [`MAX_FOLLOWED`]. + static REFUSED: Cell = const { Cell::new(0) }; +} + +/// One read or write of an object, with everything it loads on the way: what [`MAX_FOLLOWED`] counts. +pub(crate) struct FollowScope; + +impl FollowScope { + pub(crate) fn enter() -> Self { + if SCOPES.replace(SCOPES.get().saturating_add(1)) == 0 { + FOLLOWED.set(0); + REFUSED.set(0); + } + FollowScope + } +} + +impl Drop for FollowScope { + fn drop(&mut self) { + SCOPES.set(SCOPES.get().saturating_sub(1)); + } +} + +/// What `f` gives, counted against [`MAX_FOLLOWED`], or `spent` past it. Called inside a +/// [`FollowScope`], which is what starts the count again. +pub(crate) fn following(spent: T, f: impl FnOnce() -> T) -> T { + if FOLLOWED.get() >= MAX_FOLLOWED { + REFUSED.set(REFUSED.get().saturating_add(1)); + return spent; + } + FOLLOWED.set(FOLLOWED.get() + 1); + f() +} + +/// `o`, or the object it refers to, for what decides how a stream or dictionary is treated. Past +/// [`MAX_FOLLOWED`] a reference reads as absent. +fn followed<'o>(doc: &Document, o: &'o Object) -> Held<'o> { + match o { + Object::Ref(_) => following(Held::Direct(&Object::Null), || doc.held(o)), + o => Held::Direct(o), + } +} + fn bytes_of(doc: &Document, d: &Dict, key: &[u8]) -> Vec { d.get(key).map(|v| doc.resolve(v)).and_then(|v| v.as_string().map(|s| s.bytes.clone())).unwrap_or_default() } @@ -18,12 +77,12 @@ fn bytes_of(doc: &Document, d: &Dict, key: &[u8]) -> Vec { /// Read the handler parameters from an `/Encrypt` dictionary. pub(crate) fn encrypt_dict(doc: &Document, d: &Dict) -> EncryptDict { let int = |k: &[u8], default: i64| d.get(k).map(|v| doc.resolve(v)).and_then(|v| v.as_int()).unwrap_or(default); - let name = |k: &[u8]| d.get(k).and_then(|v| v.as_name()).map(|n| n.to_vec()).unwrap_or_else(|| b"Identity".to_vec()); + let name = |d: &Dict, k: &[u8]| d.get(k).and_then(|v| doc.resolve(v).as_name().map(<[u8]>::to_vec)); let mut crypt_filters = Vec::new(); if let Some(cf) = d.get(b"CF").map(|c| doc.resolve(c)).and_then(|c| c.as_dict().cloned()) { for (k, v) in cf.iter() { let Some(f) = doc.resolve(v).as_dict().cloned() else { continue }; - let method = match f.name(b"CFM") { + let method = match name(&f, b"CFM").as_deref() { Some(b"V2") => Method::Rc4, Some(b"AESV2") => Method::Aes128, Some(b"AESV3") => Method::Aes256, @@ -40,7 +99,8 @@ pub(crate) fn encrypt_dict(doc: &Document, d: &Dict) -> EncryptDict { .get(b"CF") .map(|c| doc.resolve(c)) .and_then(|c| c.as_dict().and_then(|c| c.get(b"StdCF").map(|f| doc.resolve(f)))) - .and_then(|f| f.as_dict().and_then(|f| f.int(b"Length"))) + .and_then(|f| f.as_dict().and_then(|f| f.get(b"Length").map(|n| doc.resolve(n)))) + .and_then(|n| n.as_int()) .filter(|n| *n > 0); let length_bits = match (v, cf_len) { (5.., _) => 256, @@ -50,7 +110,7 @@ pub(crate) fn encrypt_dict(doc: &Document, d: &Dict) -> EncryptDict { _ => int(b"Length", 40), }; EncryptDict { - filter: d.name(b"Filter").map(|n| n.to_vec()).unwrap_or_default(), + filter: name(d, b"Filter").unwrap_or_default(), v, r: int(b"R", 0), length_bits, @@ -60,28 +120,46 @@ pub(crate) fn encrypt_dict(doc: &Document, d: &Dict) -> EncryptDict { ue: bytes_of(doc, d, b"UE"), perms: bytes_of(doc, d, b"Perms"), p: int(b"P", -1) as i32, - encrypt_metadata: !matches!(d.get(b"EncryptMetadata"), Some(Object::Bool(false))), + encrypt_metadata: !matches!(d.get(b"EncryptMetadata").map(|v| doc.resolve(v)).as_deref(), Some(Object::Bool(false))), crypt_filters, - stm_f: name(b"StmF"), - str_f: name(b"StrF"), - ef_f: d.get(b"EFF").and_then(|v| v.as_name()).map(|n| n.to_vec()).unwrap_or_default(), + stm_f: name(d, b"StmF").unwrap_or_else(|| b"Identity".to_vec()), + str_f: name(d, b"StrF").unwrap_or_else(|| b"Identity".to_vec()), + ef_f: name(d, b"EFF").unwrap_or_default(), } } -/// The crypt filter a stream names in its `/Filter` chain (`/Crypt` + `/DecodeParms /Name`). -fn named_crypt_filter(dict: &Dict) -> Option> { - let names: Vec<&[u8]> = match dict.get(b"Filter") { - Some(Object::Name(n)) => vec![n], - Some(Object::Array(a)) => a.iter().filter_map(Object::as_name).collect(), +/// Whether `f`, an entry of a `/Filter` chain or its only filter, is `/Crypt`. +fn is_crypt(doc: &Document, f: &Object) -> bool { + followed(doc, f).as_name() == Some(b"Crypt") +} + +/// The crypt filter a stream names in its `/Filter` chain (`/Crypt` + `/DecodeParms /Name`), each +/// given directly or by reference, in an array or not. A chain longer than any decoder takes names +/// none: nothing can read that stream anyway, and every stream of a file can share one long array. +/// A `/Name` that reads as absent means `/Identity`, unless a reference on the way to it was refused +/// (see [`MAX_FOLLOWED`]): that says nothing about the stream, so it names none, and the document's +/// own filter is used. +fn named_crypt_filter(doc: &Document, dict: &Dict) -> Option> { + let refused = REFUSED.get(); + let filter = followed(doc, dict.get(b"Filter")?); + let i = match &*filter { + Object::Array(a) if a.len() <= MAX_CHAIN + 1 => a.iter().position(|f| is_crypt(doc, f))?, + f if is_crypt(doc, f) => 0, _ => return None, }; - let i = names.iter().position(|n| *n == b"Crypt")?; - let parms = match dict.get(b"DecodeParms") { - Some(Object::Dict(d)) if i == 0 => Some(d.clone()), - Some(Object::Array(a)) => a.get(i).and_then(|p| p.as_dict().cloned()), + let parms = dict.get(b"DecodeParms").map(|p| followed(doc, p)); + let entry = match parms.as_deref() { + Some(Object::Array(a)) => a.get(i).map(|p| followed(doc, p)), + Some(p) if i == 0 => Some(Held::Direct(p)), _ => None, }; - Some(parms.and_then(|p| p.name(b"Name").map(|n| n.to_vec())).unwrap_or_else(|| b"Identity".to_vec())) + let name = entry.as_deref().and_then(Object::as_dict).and_then(|p| p.get(b"Name")).map(|n| followed(doc, n)); + match name.as_deref().and_then(Object::as_name) { + Some(name) => Some(name.to_vec()), + // /Identity here would read ciphertext as data, and write the stream in the clear. + None if REFUSED.get() > refused => None, + None => Some(b"Identity".to_vec()), + } } /// A signature dictionary (field signature, document timestamp or usage-rights signature), @@ -91,42 +169,67 @@ fn named_crypt_filter(dict: &Dict) -> Option> { /// `/Contents` alongside an array `/ByteRange` (`/Type` is optional). A signature dictionary /// has no `/Subtype`; a dictionary with one (an annotation, whose `/Contents` is the comment /// text) or with another `/Type` is never exempt, whatever else it carries. -fn is_signature_dict(d: &Dict) -> bool { - if d.get(b"Subtype").is_some() { +/// With `doc` its entries are read through references; a dictionary nested in another object +/// has none (see [`transform`]). +fn is_signature_dict(doc: Option<&Document>, d: &Dict) -> bool { + // All this decides is what happens to such a /Contents. + if !matches!(d.get(b"Contents"), Some(Object::String(_))) { return false; } - match d.get(b"Type") { - Some(_) => matches!(d.name(b"Type"), Some(b"Sig" | b"DocTimeStamp")), - None => matches!(d.get(b"Contents"), Some(Object::String(_))) && matches!(d.get(b"ByteRange"), Some(Object::Array(_))), + // A null entry, or one that refers to nothing, is an absent one (ISO 32000-2 §7.3.7, §7.3.10). + let entry = |key: &[u8]| { + let value = d.get(key)?; + let held = doc.map_or(Held::Direct(value), |doc| followed(doc, value)); + (!matches!(*held, Object::Null)).then_some(held) + }; + if entry(b"Subtype").is_some() { + return false; + } + match entry(b"Type") { + Some(t) => matches!(t.as_name(), Some(b"Sig" | b"DocTimeStamp")), + None => matches!(entry(b"ByteRange").as_deref(), Some(Object::Array(_))), } } -/// Remove the `/Crypt` entry from a stream's filter chain, with its decode parameters. -fn drop_crypt_filter(dict: &mut Dict) { - let i = match dict.get(b"Filter") { - Some(Object::Array(a)) => a.iter().position(|n| n.as_name() == Some(b"Crypt")), - Some(Object::Name(n)) if n.as_slice() == b"Crypt" => Some(0), - _ => None, +/// Longer than any filter's name. A name can be of any length, and a chain every stream shares can +/// hold a long one. +const MAX_FILTER_NAME: usize = 127; + +/// The dictionary of `source` without the `/Crypt` entries of its filter chain and their decode +/// parameters. The two come out direct when given by reference, which every stream of a file can +/// share, so of what was shared only small values are copied: filter names, and the parameters +/// decoders read ([`Document::with_direct_parms`]). The chain is short, as [`named_crypt_filter`] +/// found a filter in it. +fn drop_crypt_filter(doc: &Document, source: &Stream) -> Dict { + let entry = |f: &Object| match f { + f if is_crypt(doc, f) => Object::name("Crypt"), + Object::Name(n) if n.len() <= MAX_FILTER_NAME => f.clone(), + Object::Ref(_) => f.clone(), + _ => Object::Null, }; - let Some(i) = i else { return }; - let mut filters = match dict.get(b"Filter") { - Some(Object::Array(a)) => a.clone(), - _ => Vec::new(), + let filter = source.dict.get(b"Filter").map(|f| followed(doc, f)); + let chain: Vec = match filter.as_deref() { + Some(Object::Array(a)) => a.iter().map(entry).collect(), + Some(f) => vec![entry(f)], + None => Vec::new(), }; - if i < filters.len() { - filters.remove(i); - } + let crypt: Vec = chain.iter().map(|f| f.as_name() == Some(b"Crypt")).collect(); + let mut chained = Stream::from_raw(source.dict.clone(), Vec::new()); + chained.dict.set(b"Filter".to_vec(), Object::Array(chain.clone())); // A lone parameter dictionary belongs to the first filter (as `named_crypt_filter` reads it). - match dict.get(b"DecodeParms").cloned() { - Some(Object::Array(mut parms)) if i < parms.len() => { - parms.remove(i); - dict.set(b"DecodeParms".to_vec(), Object::Array(parms)); - } - Some(Object::Dict(_)) if i == 0 => { + let parms = match doc.with_direct_parms(&chained).dict.get(b"DecodeParms") { + Some(Object::Array(p)) => Some(Object::Array(p.iter().zip(&crypt).filter(|(_, c)| !**c).map(|(p, _)| p.clone()).collect())), + Some(Object::Dict(_)) if crypt.first() == Some(&true) => None, + _ => source.dict.get(b"DecodeParms").cloned(), + }; + let mut dict = source.dict.clone(); + match parms { + Some(p) => dict.set(b"DecodeParms".to_vec(), p), + None => { dict.remove(b"DecodeParms"); } - _ => {} } + let filters: Vec = chain.into_iter().zip(&crypt).filter(|(_, c)| !**c).map(|(f, _)| f).collect(); if filters.is_empty() { dict.remove(b"Filter"); if matches!(dict.get(b"DecodeParms"), Some(Object::Array(p)) if p.is_empty()) { @@ -135,6 +238,7 @@ fn drop_crypt_filter(dict: &mut Dict) { } else { dict.set(b"Filter".to_vec(), Object::Array(filters)); } + dict } /// What [`transform`] does to an object. @@ -152,51 +256,81 @@ pub(crate) enum Pass { /// Decrypt or encrypt every string and stream in an indirect object, except a signature /// dictionary's `/Contents` (see [`is_signature_dict`]). -pub(crate) fn transform(h: &SecurityHandler, o: &Object, num: u32, generation: u16, pass: Pass) -> Object { +/// +/// `o` is as `doc` holds it: what a stream's dictionary gives by reference is read from `doc`, so +/// its references have to be the document's, not a save's renumbered ones. +pub(crate) fn transform(doc: &Document, h: &SecurityHandler, o: &Object, num: u32, generation: u16, pass: Pass) -> Object { + let _scope = FollowScope::enter(); let decrypt = pass == Pass::Decrypt; let strings = |s: &PdfString| { let bytes = if decrypt { h.decrypt_string(num, generation, &s.bytes) } else { h.encrypt_string(num, generation, &s.bytes) }; PdfString { bytes, hex: s.hex || !decrypt } }; - walk(o, &strings, h, num, generation, pass) + // References are followed here only, for the object itself, never for what is nested in it. + // `walk` goes deep, and a load from down there adds that depth again for each object loaded, + // which overflowed the stack on a 10 KB file. A stream can't be nested, and a dictionary that + // is reads as it says directly. + match o { + Object::Stream(s) => stream(doc, s, &strings, h, num, generation, pass), + Object::Dict(d) if is_signature_dict(Some(doc), d) => signature_dict(d, &strings), + o => walk(o, &strings), + } } -fn walk(o: &Object, strings: &dyn Fn(&PdfString) -> PdfString, h: &SecurityHandler, num: u32, generation: u16, pass: Pass) -> Object { - let decrypt = pass == Pass::Decrypt; +/// `d`, a signature dictionary, with the strings of everything but its `/Contents` transformed. +fn signature_dict(d: &Dict, strings: &dyn Fn(&PdfString) -> PdfString) -> Object { + Object::Dict( + d.iter() + .map(|(k, v)| match (k.as_slice(), v) { + (b"Contents", Object::String(_)) => (k.clone(), v.clone()), + _ => (k.clone(), walk(v, strings)), + }) + .collect(), + ) +} + +fn walk(o: &Object, strings: &dyn Fn(&PdfString) -> PdfString) -> Object { match o { Object::String(s) => Object::String(strings(s)), - Object::Array(a) => Object::Array(a.iter().map(|x| walk(x, strings, h, num, generation, pass)).collect()), - Object::Dict(d) if is_signature_dict(d) => Object::Dict( - d.iter() - .map(|(k, v)| match (k.as_slice(), v) { - (b"Contents", Object::String(_)) => (k.clone(), v.clone()), - _ => (k.clone(), walk(v, strings, h, num, generation, pass)), - }) - .collect(), - ), - Object::Dict(d) => Object::Dict(d.iter().map(|(k, v)| (k.clone(), walk(v, strings, h, num, generation, pass))).collect()), - Object::Stream(s) => { - let mut dict: Dict = s.dict.iter().map(|(k, v)| (k.clone(), walk(v, strings, h, num, generation, pass))).collect(); - let ty = s.dict.name(b"Type"); - // Cross-reference streams are never encrypted (§7.5.8), but only reading may trust the - // label: a stream a producer calls /XRef stays plaintext if writing believes it, and the - // writer builds its own cross-reference streams outside this transform. - if (decrypt && ty == Some(b"XRef")) || (ty == Some(b"Metadata") && !h.encrypts_metadata()) { - return Object::Stream(Stream { dict, raw: s.raw.clone() }); - } - let mut named = named_crypt_filter(&s.dict); - if pass == Pass::EncryptReplacing && named.as_deref().is_some_and(|n| !h.dict().crypt_filters.iter().any(|(f, _)| f == n)) { - drop_crypt_filter(&mut dict); - named = None; - } - let kind = match (&named, ty) { - (Some(n), _) => StreamKind::Named(n), - (None, Some(b"EmbeddedFile")) => StreamKind::EmbeddedFile, - _ => StreamKind::Normal, - }; - let raw = if decrypt { h.decrypt_stream(num, generation, &s.raw, kind) } else { h.encrypt_stream(num, generation, &s.raw, kind) }; - Object::Stream(Stream { dict, raw: raw.into() }) - } + Object::Array(a) => Object::Array(a.iter().map(|x| walk(x, strings)).collect()), + Object::Dict(d) if is_signature_dict(None, d) => signature_dict(d, strings), + Object::Dict(d) => Object::Dict(d.iter().map(|(k, v)| (k.clone(), walk(v, strings))).collect()), other => other.clone(), } } + +fn stream( + doc: &Document, + s: &Stream, + strings: &dyn Fn(&PdfString) -> PdfString, + h: &SecurityHandler, + num: u32, + generation: u16, + pass: Pass, +) -> Object { + let decrypt = pass == Pass::Decrypt; + let walked = |d: &Dict| -> Dict { d.iter().map(|(k, v)| (k.clone(), walk(v, strings))).collect() }; + let ty = s.dict.get(b"Type").map(|t| followed(doc, t)); + let ty = ty.as_deref().and_then(Object::as_name); + // Cross-reference streams are never encrypted (§7.5.8), but only reading may trust the + // label: a stream a producer calls /XRef stays plaintext if writing believes it, and the + // writer builds its own cross-reference streams outside this transform. + if (decrypt && ty == Some(b"XRef")) || (ty == Some(b"Metadata") && !h.encrypts_metadata()) { + return Object::Stream(Stream { dict: walked(&s.dict), raw: s.raw.clone() }); + } + let mut named = named_crypt_filter(doc, &s.dict); + // Before the walk, which then encrypts what is in the dictionary this leaves. + let foreign = pass == Pass::EncryptReplacing && named.as_deref().is_some_and(|n| !h.dict().crypt_filters.iter().any(|(f, _)| f == n)); + let dropped = foreign.then(|| drop_crypt_filter(doc, s)); + if dropped.is_some() { + named = None; + } + let dict = walked(dropped.as_ref().unwrap_or(&s.dict)); + let kind = match (&named, ty) { + (Some(n), _) => StreamKind::Named(n), + (None, Some(b"EmbeddedFile")) => StreamKind::EmbeddedFile, + _ => StreamKind::Normal, + }; + let raw = if decrypt { h.decrypt_stream(num, generation, &s.raw, kind) } else { h.encrypt_stream(num, generation, &s.raw, kind) }; + Object::Stream(Stream { dict, raw: raw.into() }) +} diff --git a/crates/cos/src/writer.rs b/crates/cos/src/writer.rs index e34799d5f9..e5ee7530b8 100644 --- a/crates/cos/src/writer.rs +++ b/crates/cos/src/writer.rs @@ -133,11 +133,12 @@ fn write_stream(s: &Stream, out: &mut Vec) { /// The object as written: encrypted with its (output) number when the document is encrypted, /// except the `/Encrypt` dictionary itself (`source_num` is its number in the source document). +/// `o` has the document's numbers in its references ([`crate::security::transform`]). fn prepared<'a>(doc: &Document, source_num: u32, num: u32, generation: u16, o: &'a Object) -> Cow<'a, Object> { match doc.output_security() { - (Some(h), encrypt_num) if Some(source_num) != encrypt_num => { + (Some(h), encrypt_num) if Some(source_num) != encrypt_num && !doc.stored_as_is(source_num) => { let pass = if doc.encryption_changed() { crate::security::Pass::EncryptReplacing } else { crate::security::Pass::Encrypt }; - Cow::Owned(crate::security::transform(h, o, num, generation, pass)) + Cow::Owned(crate::security::transform(doc, h, o, num, generation, pass)) } _ => Cow::Borrowed(o), } @@ -288,7 +289,10 @@ pub fn write_full(doc: &Document, opts: &SaveOptions) -> Result, CosErro // The /Encrypt dictionary and everything it references (e.g. indirect /CF crypt-filter // dictionaries) are needed before anything can be decrypted: never inside object streams. let mut encryption_objects: std::collections::HashSet = std::collections::HashSet::new(); - let mut pending: Vec = trailer_in.reference(b"Encrypt").into_iter().collect(); + let mut pending: Vec = Vec::new(); + if let Some(encrypt) = trailer_in.get(b"Encrypt") { + collect_refs(encrypt, &mut |r| pending.push(r)); + } while let Some(r) = pending.pop() { if encryption_objects.insert(r) { collect_refs(&reader.get(r), &mut |x| pending.push(x)); @@ -309,9 +313,10 @@ pub fn write_full(doc: &Document, opts: &SaveOptions) -> Result, CosErro if in_stream { packed.push((num, *r)); } else { - let o = renumber(&*reread(&mut reader, *r, *is_stream)?, &map); + let o = reread(&mut reader, *r, *is_stream)?; + let o = renumber(&prepared(&reader.document, r.num, num, 0, &o), &map); rows.insert(num, Row::InFile(out.len() as u64, 0)); - write_indirect(num, 0, &prepared(&reader.document, r.num, num, 0, &o), &mut out); + write_indirect(num, 0, &o, &mut out); } } // Object streams take the numbers after the last object. Their contents are not encrypted @@ -341,11 +346,11 @@ pub fn write_full(doc: &Document, opts: &SaveOptions) -> Result, CosErro trailer.set(key.to_vec(), Object::Ref(ObjRef::new(*r, 0))); } } - if let Some(Object::Dict(d)) = trailer_in.get(b"Encrypt") { - trailer.set(b"Encrypt".to_vec(), Object::Dict(d.clone())); + if let Some(encrypt @ Object::Dict(_)) = trailer_in.get(b"Encrypt") { + trailer.set(b"Encrypt".to_vec(), renumber(encrypt, &map)); } if let Some(id) = trailer_in.get(b"ID") { - trailer.set(b"ID".to_vec(), id.clone()); + trailer.set(b"ID".to_vec(), direct_id(&reader.document, id)); } for (key, value) in trailer_in.iter() { if !trailer_key_rewritten(key) { @@ -415,6 +420,22 @@ impl ObjectStreamBatch { } } +/// `/ID` for the trailer, direct, as it has to be with `/Encrypt`: nothing else in the file holds +/// what it refers to now. Given by reference it is two entries at most, however many references a +/// file gave, so that one large string isn't copied for each. An entry that isn't a string stays in +/// its place as an empty one, which is how the key read it. +fn direct_id(doc: &Document, id: &Object) -> Object { + if matches!(id, Object::Array(a) if !a.iter().any(|s| matches!(s, Object::Ref(_)))) { + return id.clone(); + } + let entry = |s: &Object| match &*doc.resolve(s) { + s @ Object::String(_) => s.clone(), + _ => Object::String(crate::PdfString { bytes: Vec::new(), hex: true }), + }; + let held = doc.resolve(id); + Object::Array(held.as_array().into_iter().flatten().take(2).map(entry).collect()) +} + /// Trailer keys a full rewrite generates itself. Everything else is copied, with references /// renumbered, so a save does not drop metadata the reader accepted. fn trailer_key_rewritten(key: &[u8]) -> bool { @@ -471,6 +492,10 @@ fn walk(reader: &mut ObjectReader, trailer: &Dict) -> Vec<(ObjRef, bool)> { queue.push_back(r); } } + // A direct /Encrypt can refer to objects, /CF or /StmF, say. + if let Some(encrypt @ Object::Dict(_)) = trailer.get(b"Encrypt") { + collect_refs(encrypt, &mut |r| queue.push_back(r)); + } // Unknown trailer entries are document data. Objects they alone reference must be written // too, or a full rewrite would keep the key and drop its value. for (key, value) in trailer.iter() { @@ -500,7 +525,7 @@ pub fn reachable(doc: &Document) -> HashSet { walk(&mut reader, &trailer).into_iter().map(|(r, _)| r).collect() } -fn collect_refs(o: &Object, f: &mut impl FnMut(ObjRef)) { +pub(crate) fn collect_refs(o: &Object, f: &mut impl FnMut(ObjRef)) { match o { Object::Ref(r) => f(*r), Object::Array(a) => a.iter().for_each(|x| collect_refs(x, f)), @@ -677,6 +702,15 @@ fn civil_from_days(days: i64) -> (i64, u32, u32) { #[cfg(test)] mod tests { + /// `/ID` by reference comes out direct, each entry where it was: the key is made from the first. + #[test] + fn a_file_identifier_keeps_its_entries_in_their_places() { + let doc = Document::new_empty(); + let string = |bytes: &[u8]| Object::String(crate::PdfString { bytes: bytes.to_vec(), hex: true }); + let nowhere = Object::Ref(ObjRef::new(9_999, 0)); + assert_eq!(super::direct_id(&doc, &Object::Array(vec![nowhere, string(b"ab")])), Object::Array(vec![string(b""), string(b"ab")])); + } + use proptest::prelude::*; use super::*; diff --git a/crates/cos/tests/it/encryption.rs b/crates/cos/tests/it/encryption.rs index 2ee33afc12..b19d618f5c 100644 --- a/crates/cos/tests/it/encryption.rs +++ b/crates/cos/tests/it/encryption.rs @@ -494,3 +494,826 @@ fn a_stream_labelled_xref_is_encrypted_when_saved() { let bytes = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); assert!(!bytes.windows(SECRET.len()).any(|w| w == SECRET.as_bytes()), "the labelled stream was written in the clear"); } + +/// The content stream's reference. +fn contents_of(doc: &Document) -> ObjRef { + let catalog = doc.get(doc.root().unwrap()).as_dict().cloned().unwrap(); + let pages = doc.get(catalog.reference(b"Pages").unwrap()).as_dict().cloned().unwrap(); + let page = doc.resolve(&pages.get(b"Kids").unwrap().as_array().unwrap()[0]).as_dict().cloned().unwrap(); + page.reference(b"Contents").unwrap() +} + +/// The content stream's data as loaded, whatever its filters say. +fn raw_content(doc: &Document) -> String { + match &*doc.get(contents_of(doc)) { + Object::Stream(s) => String::from_utf8_lossy(&s.raw).trim().to_string(), + _ => panic!("no content stream"), + } +} + +/// Gives the content stream the `/Filter` and `/DecodeParms` of `selector`. +fn select(doc: &mut Document, (filter, parms): (Object, Object)) { + let contents = contents_of(doc); + let Object::Stream(mut s) = doc.get(contents).as_ref().clone() else { panic!("no content stream") }; + s.dict.set(b"Filter".to_vec(), filter); + s.dict.set(b"DecodeParms".to_vec(), parms); + doc.set(contents, Object::Stream(s)); +} + +/// `/Crypt` naming `name`, in the `form`th spelling: all direct, then each part by reference. +/// `None` past the last. +fn selector(doc: &mut Document, form: usize, name: &str) -> Option<(Object, Object)> { + let crypt = Object::name("Crypt"); + let parms = |n: Object| { + let mut d = pdfcraft_cos::Dict::new(); + d.set(b"Name".to_vec(), n); + Object::Dict(d) + }; + let named = parms(Object::name(name)); + Some(match form { + 0 => (crypt, named), + 1 => (crypt, Object::Ref(doc.add(named))), + 2 => (Object::Array(vec![crypt]), Object::Array(vec![Object::Ref(doc.add(named))])), + 3 => (Object::Array(vec![crypt]), Object::Ref(doc.add(Object::Array(vec![named])))), + 4 => (crypt, parms(Object::Ref(doc.add(Object::name(name))))), + 5 => (Object::Ref(doc.add(crypt)), named), + 6 => (Object::Ref(doc.add(Object::Array(vec![crypt]))), Object::Array(vec![named])), + 7 => (Object::Array(vec![Object::Ref(doc.add(crypt))]), Object::Array(vec![named])), + // Behind a filter given by reference: places in the two arrays pair up all the same. + 8 => (Object::Array(vec![Object::Ref(doc.add(Object::name("ASCIIHexDecode"))), crypt]), Object::Array(vec![Object::Null, named])), + // The two entries in different forms, which is how a lone filter is usually spelled. + 9 => (Object::Array(vec![crypt]), named), + 10 => (crypt, Object::Array(vec![named])), + // The filter twice, which names the first's. + 11 => (Object::Array(vec![crypt.clone(), crypt]), Object::Array(vec![named.clone(), named])), + // A reference to a reference. + 12 => { + let name = doc.add(Object::name(name)); + let hop = doc.add(Object::Ref(name)); + (crypt, parms(Object::Ref(hop))) + } + _ => return None, + }) +} + +fn holds(bytes: &[u8], text: &str) -> bool { + String::from_utf8_lossy(bytes).contains(text) +} + +/// An object `/Encrypt` names is written as stored only while it is as stored. A file whose +/// `/Encrypt` names its `/Info` reads that dictionary as stored; once the title is edited, both +/// saves encrypt it rather than writing the new title in the clear. +#[test] +fn an_edited_object_that_encrypt_names_is_encrypted() { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes128, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + let info = doc.trailer().reference(b"Info").unwrap(); + let mut dict = doc.get(doc.trailer().reference(b"Encrypt").unwrap()).as_dict().cloned().unwrap(); + dict.set(b"Junk".to_vec(), Object::Ref(info)); + doc.trailer_mut().set(b"Encrypt".to_vec(), Object::Dict(dict)); + let bytes = write_full(&doc, &SaveOptions::default()).unwrap(); + let mut doc = Document::open(Arc::new(bytes)).unwrap(); + let info = doc.trailer().reference(b"Info").unwrap(); + doc.update_dict(info, |d| d.set(b"Title".to_vec(), PdfString::text("Merger plan"))).unwrap(); + for (save, saved) in [("incremental", write_incremental(&doc, &SaveOptions::default())), ("full", write_full(&doc, &SaveOptions::default()))] { + assert!(!holds(&saved.unwrap(), "Merger plan"), "{save}: the edited title was written in the clear"); + } +} + +/// A stream can select its crypt filter by reference at every step: the chain, an entry of it, the +/// parameters, an entry of those, the name (ISO 32000-2 §7.3.10, §7.4.10). Read as absent, the +/// filter was `/Identity`, so such a stream was saved in the clear and, in a file from elsewhere, +/// its ciphertext read as data. A full save renumbers: what it follows has the document's numbers. +#[test] +fn a_crypt_filter_selected_by_reference_is_used() { + for alg in [Algorithm::Aes128, Algorithm::Aes256] { + for form in 0.. { + let mut doc = Document::open(Arc::new(protect(alg, "", "owner", -1))).unwrap(); + let Some(selected) = selector(&mut doc, form, "StdCF") else { break }; + select(&mut doc, selected); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert!(!holds(&saved, "Secret page"), "{alg:?} form {form}: saved encrypted"); + let back = Document::open(Arc::new(saved)).unwrap(); + assert_eq!(raw_content(&back), SECRET, "{alg:?} form {form}: read back"); + let full = write_full(&back, &SaveOptions::default()).unwrap(); + assert!(!holds(&full, "Secret page"), "{alg:?} form {form}: renumbered, encrypted"); + assert_eq!(raw_content(&Document::open(Arc::new(full)).unwrap()), SECRET, "{alg:?} form {form}: renumbered"); + } + } +} + +/// A reference to a stream that takes `follows` references to read cold, this one included: its +/// nine filters are all read, the last leading to the next stream. +fn spends(doc: &mut Document, follows: usize) -> Object { + let left = follows - 1; + let mut filters: Vec = (0..left.min(8)).map(|i| Object::Ref(doc.add(Object::Int(i as i64)))).collect(); + if left > 8 { + filters.push(spends(doc, left - 8)); + } + let mut dict = pdfcraft_cos::Dict::new(); + dict.set(b"Filter".to_vec(), Object::Array(filters)); + Object::Ref(doc.add(Object::Stream(pdfcraft_cos::Stream::from_raw(dict, Vec::new())))) +} + +/// Gives the content stream a `/Type` that takes `follows` references to read, ahead of its filter. +fn spend_first(doc: &mut Document, follows: usize) { + let spent = spends(doc, follows); + let contents = contents_of(doc); + let Object::Stream(mut s) = doc.get(contents).as_ref().clone() else { panic!("no content stream") }; + s.dict.set(b"Type".to_vec(), spent); + doc.set(contents, Object::Stream(s)); +} + +/// Past what one read may follow a reference reads as absent, and a crypt filter with no `/Name` is +/// `/Identity`. Together those would read this stream's ciphertext as data, so what can't be told +/// is the document's own filter, as a `/Filter` that can't be read is. +#[test] +fn a_crypt_filter_past_what_a_read_may_follow_is_the_documents_own() { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let selected = selector(&mut doc, 1, "StdCF").unwrap(); + select(&mut doc, selected); + spend_first(&mut doc, 40); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert!(!holds(&saved, "Secret page"), "saved encrypted"); + let back = Document::open(Arc::new(saved)).unwrap(); + assert_eq!(raw_content(&back), SECRET); + let full = write_full(&back, &SaveOptions::default()).unwrap(); + assert!(!holds(&full, "Secret page"), "renumbered, encrypted"); + assert_eq!(raw_content(&Document::open(Arc::new(full)).unwrap()), SECRET, "renumbered"); +} + +/// A read that takes all it may follow has refused nothing: a crypt filter with no `/Name` after it +/// is still `/Identity`, and the stream is in the clear. +#[test] +fn a_crypt_filter_with_no_name_after_all_a_read_may_follow_is_identity() { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + select(&mut doc, (Object::name("Crypt"), Object::Dict(pdfcraft_cos::Dict::new()))); + spend_first(&mut doc, 32); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert!(holds(&saved, "Secret page"), "written as the stream asks"); + assert_eq!(raw_content(&Document::open(Arc::new(saved)).unwrap()), SECRET); +} + +/// `/Identity` selected by reference keeps a stream in the clear in a document that encrypts the +/// rest, in every spelling, on both saves. +#[test] +fn an_identity_crypt_filter_selected_by_reference_leaves_a_stream_alone() { + for form in 0.. { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let Some(selected) = selector(&mut doc, form, "Identity") else { break }; + select(&mut doc, selected); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert!(holds(&saved, "Secret page"), "form {form}: written as the stream asks"); + let back = Document::open(Arc::new(saved)).unwrap(); + assert_eq!(raw_content(&back), SECRET, "form {form}"); + let full = write_full(&back, &SaveOptions::default()).unwrap(); + assert!(holds(&full, "Secret page"), "form {form}: renumbered"); + assert_eq!(raw_content(&Document::open(Arc::new(full)).unwrap()), SECRET, "form {form}: renumbered"); + } +} + +/// New protection drops a crypt filter it doesn't define, whichever way the stream selected it. The +/// chain comes out direct. Parameters in the stream's own dictionary stay, encrypted with it; of an +/// array in an object of its own, which every stream can share, only what decoders read is copied. +#[test] +fn protecting_drops_a_crypt_filter_selected_by_reference() { + let protected = |doc: &mut Document| { + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [7; 32], + }) + .unwrap(); + write_full(doc, &SaveOptions::default()).unwrap() + }; + for form in 0.. { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let Some(selected) = selector(&mut doc, form, "MyCF") else { break }; + select(&mut doc, selected); + let bytes = protected(&mut doc); + assert!(!holds(&bytes, "Secret page"), "form {form}: encrypted"); + assert!(!holds(&bytes, "MyCF"), "form {form}: no selector left in the stream"); + let back = Document::open(Arc::new(bytes.clone())).unwrap(); + assert_eq!(raw_content(&back), SECRET, "form {form}"); + let contents = back.get(contents_of(&back)); + let filter = contents.as_dict().unwrap().get(b"Filter"); + match form { + 8 => assert!( + matches!(filter, Some(Object::Array(a)) if matches!(a.as_slice(), [Object::Ref(_)])), + "form 8: the other filter stays: {filter:?}" + ), + _ => assert_eq!(filter, None, "form {form}"), + } + assert!(hayro_syntax::Pdf::new(bytes).is_ok(), "form {form}"); + } + for shared in [false, true] { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let mut named = pdfcraft_cos::Dict::new(); + named.set(b"Name".to_vec(), Object::name("MyCF")); + let mut other = pdfcraft_cos::Dict::new(); + other.set(b"Note".to_vec(), Object::String(PdfString::text("Private note"))); + other.set(b"Predictor".to_vec(), Object::Int(1)); + let parms = Object::Array(vec![Object::Dict(named), Object::Dict(other)]); + let parms = if shared { Object::Ref(doc.add(parms)) } else { parms }; + select(&mut doc, (Object::Array(vec![Object::name("Crypt"), Object::name("ASCIIHexDecode")]), parms)); + let bytes = protected(&mut doc); + assert!(!holds(&bytes, "Private note"), "shared: {shared}"); + let back = Document::open(Arc::new(bytes)).unwrap(); + let contents = back.get(contents_of(&back)); + let kept = contents.as_dict().unwrap().get(b"DecodeParms").unwrap().as_array().unwrap(); + assert_eq!(kept.len(), 1, "one filter left, one entry"); + let kept = kept[0].as_dict().unwrap(); + assert_eq!(kept.get(b"Predictor"), Some(&Object::Int(1)), "shared: {shared}"); + assert_eq!(kept.get(b"Note").is_some(), !shared, "shared: {shared}"); + } +} + +/// Every stream of a file can share one large object as its `/DecodeParms` or `/Filter`, which +/// costs nothing until new protection drops a crypt filter from each. Copied whole, a 1 MB file +/// came out at 1 GB. +#[test] +fn protecting_copies_no_large_object_into_every_stream() { + let pad = Object::Array(vec![Object::Int(1); 100_000]); + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let mut large = pdfcraft_cos::Dict::new(); + large.set(b"Pad".to_vec(), pad.clone()); + let parms = doc.add(Object::Array(vec![Object::Null, Object::Dict(large)])); + let filter = doc.add(Object::Array(vec![Object::name("Crypt"), Object::Name(vec![b'A'; 100_000]), pad])); + let mut streams = Vec::new(); + for _ in 0..200 { + let mut by_parms = pdfcraft_cos::Dict::new(); + by_parms.set(b"Filter".to_vec(), Object::Array(vec![Object::name("Crypt"), Object::name("ASCIIHexDecode")])); + by_parms.set(b"DecodeParms".to_vec(), Object::Ref(parms)); + streams.push(Object::Ref(doc.add(Object::Stream(pdfcraft_cos::Stream::from_raw(by_parms, b"41>".to_vec()))))); + let mut by_filter = pdfcraft_cos::Dict::new(); + by_filter.set(b"Filter".to_vec(), Object::Ref(filter)); + streams.push(Object::Ref(doc.add(Object::Stream(pdfcraft_cos::Stream::from_raw(by_filter, b"41>".to_vec()))))); + } + doc.update_dict(doc.root().unwrap(), |c| c.set(b"Test".to_vec(), Object::Array(streams))).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [7; 32], + }) + .unwrap(); + let bytes = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); + assert!(bytes.len() < 1 << 20, "{} bytes", bytes.len()); +} + +/// A file can name one large string thousands of times in `/ID`, which holds two. An encrypted file +/// keeps its `/ID`, so each copy would be in the file a save writes. +#[test] +fn a_file_identifier_of_many_references_is_two_strings() { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes128, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + let large = doc.add(Object::String(PdfString { bytes: vec![0xAB; 100_000], hex: true })); + doc.trailer_mut().set(b"ID".to_vec(), Object::Array(vec![Object::Ref(large); 500])); + let bytes = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); + assert!(bytes.len() < 1 << 20, "{} bytes", bytes.len()); +} + +/// What a file can do with those references: point a stream's selector at the stream itself, or at +/// nothing. They read as absent, so a stream with `/Crypt` is `/Identity` (ISO 32000-2 §7.4.10), +/// and one that names no filter follows the document's. Loading ends either way. +#[test] +fn a_crypt_filter_reference_to_itself_or_nothing_reads_as_absent() { + let named = |n: Object| { + let mut d = pdfcraft_cos::Dict::new(); + d.set(b"Name".to_vec(), n); + Object::Dict(d) + }; + for (what, encrypted) in [ + ("/Filter to itself", true), + ("/Filter nowhere", true), + ("/DecodeParms to itself", false), + ("/DecodeParms nowhere", false), + ("/Name to itself", false), + ("/Name nowhere", false), + ] { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let me = Object::Ref(contents_of(&doc)); + let nowhere = Object::Ref(ObjRef::new(9_999, 0)); + let crypt = Object::name("Crypt"); + select( + &mut doc, + match what { + "/Filter to itself" => (me, Object::Null), + "/Filter nowhere" => (nowhere, Object::Null), + "/DecodeParms to itself" => (crypt, me), + "/DecodeParms nowhere" => (crypt, nowhere), + "/Name to itself" => (crypt, named(me)), + _ => (crypt, named(nowhere)), + }, + ); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert_eq!(!holds(&saved, "Secret page"), encrypted, "{what}"); + assert_eq!(raw_content(&Document::open(Arc::new(saved)).unwrap()), SECRET, "{what}"); + } +} + +/// A lone parameter dictionary belongs to the first filter, so `/Crypt` behind another filter has +/// none of its own and is `/Identity` (ISO 32000-2 §7.4.1), whatever that dictionary says. +#[test] +fn a_lone_parameter_dictionary_is_the_first_filters() { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let mut parms = pdfcraft_cos::Dict::new(); + parms.set(b"Name".to_vec(), Object::name("StdCF")); + let first = Object::Array(vec![Object::name("Crypt"), Object::name("ASCIIHexDecode")]); + let second = Object::Array(vec![Object::name("ASCIIHexDecode"), Object::name("Crypt")]); + select(&mut doc, (first, Object::Dict(parms.clone()))); + assert!(!holds(&write_incremental(&doc, &SaveOptions::default()).unwrap(), "Secret page"), "first: it names StdCF"); + select(&mut doc, (second, Object::Dict(parms))); + assert!(holds(&write_incremental(&doc, &SaveOptions::default()).unwrap(), "Secret page"), "second: it names nothing"); +} + +/// A lone parameter dictionary is the first filter's, so dropping a `/Crypt` behind that filter +/// leaves it, and dropping one in front of it takes it away. +#[test] +fn protecting_keeps_the_parameters_of_the_filter_before_a_crypt_filter() { + let protected = |filters: Vec| { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let mut parms = pdfcraft_cos::Dict::new(); + parms.set(b"Name".to_vec(), Object::name("MyCF")); + select(&mut doc, (Object::Array(filters), Object::Dict(parms))); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [7; 32], + }) + .unwrap(); + let back = Document::open(Arc::new(write_full(&doc, &SaveOptions::default()).unwrap())).unwrap(); + back.get(contents_of(&back)).as_dict().unwrap().contains(b"DecodeParms") + }; + let (hex, crypt) = (Object::name("ASCIIHexDecode"), Object::name("Crypt")); + assert!(protected(vec![hex.clone(), crypt.clone()]), "the first filter's, which stays"); + assert!(!protected(vec![crypt, hex]), "/Crypt's, which goes"); +} + +/// Dropping `/Crypt` takes its entry out of `/DecodeParms`, and an array too short to have one is +/// left as it is, given directly or by reference. +#[test] +fn protecting_leaves_a_parameter_array_too_short_for_the_crypt_filter() { + for by_reference in [false, true] { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let short = Object::Array(vec![Object::Null]); + let parms = if by_reference { Object::Ref(doc.add(short)) } else { short }; + select(&mut doc, (Object::Array(vec![Object::name("ASCIIHexDecode"), Object::name("Crypt")]), parms)); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [7; 32], + }) + .unwrap(); + let bytes = write_full(&doc, &SaveOptions::default()).unwrap(); + assert!(!holds(&bytes, "Secret page"), "by reference: {by_reference}"); + let back = Document::open(Arc::new(bytes)).unwrap(); + assert_eq!(raw_content(&back), SECRET, "by reference: {by_reference}"); + let contents = back.get(contents_of(&back)); + let filter = contents.as_dict().unwrap().get(b"Filter").unwrap().as_array().unwrap(); + assert_eq!(filter.len(), 1, "only /Crypt goes"); + assert_eq!(filter[0].as_name(), Some(b"ASCIIHexDecode".as_slice()), "the other filter stays as it was"); + } +} + +/// A chain of up to [`pdfcraft_filters::MAX_CHAIN`] filters and `/Crypt` is read; a longer one +/// names no crypt filter, so its entries aren't followed for every stream sharing it. +#[test] +fn a_chain_longer_than_any_decoder_takes_names_no_crypt_filter() { + let chain = |others: usize| { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let mut filters = vec![Object::name("ASCIIHexDecode"); others]; + filters.push(Object::name("Crypt")); + let array = doc.add(Object::Array(filters)); + select(&mut doc, (Object::Ref(array), Object::Null)); + holds(&write_incremental(&doc, &SaveOptions::default()).unwrap(), "Secret page") + }; + assert!(chain(pdfcraft_filters::MAX_CHAIN), "the longest chain read: /Identity, in the clear"); + assert!(!chain(pdfcraft_filters::MAX_CHAIN + 1), "one more: the document's filter"); +} + +/// A stream's `/Type` by reference picks its handling like one given directly: metadata left +/// readable under `/EncryptMetadata false`, an embedded file under `/EFF`. +#[test] +fn a_stream_type_given_by_reference_picks_its_crypt_filter() { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let catalog = doc.get(doc.root().unwrap()).as_dict().cloned().unwrap(); + let meta = catalog.reference(b"Metadata").unwrap(); + let ty = doc.add(Object::name("Metadata")); + let Object::Stream(mut s) = doc.get(meta).as_ref().clone() else { panic!("no metadata stream") }; + s.dict.set(b"Type".to_vec(), Object::Ref(ty)); + doc.set(meta, Object::Stream(s)); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "o", + permissions: -1, + encrypt_metadata: false, + seed: [5; 32], + }) + .unwrap(); + let bytes = write_full(&doc, &SaveOptions::default()).unwrap(); + assert!(holds(&bytes, "meta"), "search engines can read XMP"); + + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let encrypt = doc.trailer().reference(b"Encrypt").unwrap(); + doc.update_dict(encrypt, |e| e.set(b"EFF".to_vec(), Object::name("Identity"))).unwrap(); + let mut doc = Document::open(Arc::new(write_incremental(&doc, &SaveOptions::default()).unwrap())).unwrap(); + let contents = contents_of(&doc); + let ty = doc.add(Object::name("EmbeddedFile")); + let Object::Stream(mut s) = doc.get(contents).as_ref().clone() else { panic!("no content stream") }; + s.dict.set(b"Type".to_vec(), Object::Ref(ty)); + doc.set(contents, Object::Stream(s)); + assert!(holds(&write_incremental(&doc, &SaveOptions::default()).unwrap(), "Secret page"), "/EFF /Identity leaves it alone"); +} + +/// What makes a dictionary a signature's can be given by reference, and what is null or refers to +/// nothing is absent: `/Contents` stays as written in each of these. +#[test] +fn a_signature_dictionary_is_known_in_every_form() { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + let ty = doc.add(Object::name("Sig")); + let range = Object::Array([0, 10, 20, 30].into_iter().map(Object::Int).collect()); + let by_reference = doc.add(range.clone()); + let nowhere = Object::Ref(ObjRef::new(9_999, 0)); + let forms: [(Vec<(&str, Object)>, &str); 5] = [ + (vec![("Type", Object::Ref(ty))], "DEAD"), + (vec![("ByteRange", Object::Ref(by_reference))], "BEEF"), + (vec![("Type", Object::Null), ("ByteRange", range.clone())], "CAFE"), + (vec![("Type", nowhere.clone()), ("ByteRange", range.clone())], "F00D"), + (vec![("Subtype", nowhere), ("ByteRange", range)], "D00D"), + ]; + let mut signatures = Vec::new(); + for (entries, contents) in forms { + let mut sig = pdfcraft_cos::Dict::new(); + for (key, value) in entries { + sig.set(key.as_bytes().to_vec(), value); + } + let bytes = (0..contents.len()).step_by(2).map(|i| u8::from_str_radix(&contents[i..i + 2], 16).unwrap()).collect(); + sig.set(b"Contents".to_vec(), Object::String(PdfString { bytes, hex: true })); + signatures.push(Object::Ref(doc.add(Object::Dict(sig)))); + } + // Nested in another object, it follows nothing, but known directly it is one all the same. + let mut nested = pdfcraft_cos::Dict::new(); + nested.set(b"Type".to_vec(), Object::name("Sig")); + nested.set(b"Contents".to_vec(), Object::String(PdfString { bytes: vec![0xAB, 0xBA], hex: true })); + let mut field = pdfcraft_cos::Dict::new(); + field.set(b"V".to_vec(), Object::Dict(nested)); + signatures.push(Object::Ref(doc.add(Object::Dict(field)))); + // Neither of what makes one: a string /Contents alone is anybody's, so it's encrypted. + let mut other = pdfcraft_cos::Dict::new(); + other.set(b"Contents".to_vec(), Object::String(PdfString { bytes: vec![0xAB, 0xCD], hex: true })); + signatures.push(Object::Ref(doc.add(Object::Dict(other)))); + doc.update_dict(doc.root().unwrap(), |c| c.set(b"Test".to_vec(), Object::Array(signatures))).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes256, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [3; 32], + }) + .unwrap(); + let bytes = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); + for contents in ["DEAD", "BEEF", "CAFE", "F00D", "D00D", "ABBA"] { + assert!(holds(&bytes, &format!("<{contents}>")), "{contents}: written as is"); + } + assert!(!holds(&bytes, ""), "without a /ByteRange it is no signature"); +} + +/// What a test needs to write an update of a classic file protected with AES-128. +struct Classic { + bytes: Vec, + root: ObjRef, + pages: ObjRef, + info: ObjRef, + encrypt: ObjRef, + id: Vec, + dict: pdfcraft_cos::Dict, +} + +fn classic_file() -> Classic { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes128, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + let bytes = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); + let doc = Document::open(Arc::new(bytes.clone())).unwrap(); + let encrypt = doc.trailer().reference(b"Encrypt").unwrap(); + let id = doc.trailer().get(b"ID").unwrap().as_array().unwrap()[0].as_string().unwrap().bytes.clone(); + let dict = doc.get(encrypt).as_dict().cloned().unwrap(); + let pages = doc.get(doc.root().unwrap()).as_dict().unwrap().reference(b"Pages").unwrap(); + Classic { root: doc.root().unwrap(), pages, info: doc.trailer().reference(b"Info").unwrap(), bytes, encrypt, id, dict } +} + +fn hex(bytes: &[u8]) -> String { + bytes.iter().map(|b| format!("{b:02X}")).collect() +} + +/// `bytes` with an update appended: `objects`, and a trailer of `trailer` entries, in which `@n` +/// stands for the offset of object `n`. An object with nothing in it gets an entry of the +/// cross-reference table that points at nothing. +fn updated(bytes: &[u8], objects: &[(u32, Vec)], trailer: &str) -> Vec { + let at = bytes.windows(9).rposition(|w| w == b"startxref").unwrap(); + let prev: usize = String::from_utf8_lossy(&bytes[at + 9..]).split_whitespace().next().unwrap().parse().unwrap(); + let mut out = bytes.to_vec(); + let mut table = String::from("xref\n"); + let mut trailer = trailer.to_string(); + for (num, body) in objects { + if body.is_empty() { + table.push_str(&format!("{num} 1\n0000000010 00000 n \n")); + continue; + } + trailer = trailer.replace(&format!("@{num}"), &out.len().to_string()); + table.push_str(&format!("{num} 1\n{:010} 00000 n \n", out.len())); + out.extend_from_slice(format!("{num} 0 obj\n").as_bytes()); + out.extend_from_slice(body); + out.extend_from_slice(b"\nendobj\n"); + } + let xref = out.len(); + out.extend_from_slice(table.as_bytes()); + out.extend_from_slice(format!("trailer\n<< {trailer} /Size 2000 /Prev {prev} >>\nstartxref\n{xref}\n%%EOF\n").as_bytes()); + out +} + +/// The `/Encrypt` dictionary of `file`, with `/O`, `/StmF` and `/StrF` as given. +fn encrypt_with(file: &Classic, o: &str, filter: &str) -> Vec { + let string = |key: &[u8]| hex(&file.dict.get(key).unwrap().as_string().unwrap().bytes); + format!( + "<< /Filter /Standard /V 4 /R 4 /Length 128 /P {} /O {o} /U <{}> /CF << /StdCF << /CFM /AESV2 /Length 16 /AuthEvent /DocOpen >> >> /StmF /{filter} /StrF /{filter} >>", + file.dict.int(b"P").unwrap(), + string(b"U") + ) + .into_bytes() +} + +/// An object stream can have its parameters by reference, and a stream's crypt filter can be +/// selected by a member of it: legal together, though each is rare. What reads the stream first +/// goes through the object stream, and members must come out of it as they do read directly. They +/// didn't, while a read made for a selector wasn't allowed to follow its own: the object stream was +/// kept undecoded, its members lost for good, and a full save wrote nulls without saying so. +#[test] +fn an_object_stream_reached_from_a_selector_keeps_its_members() { + let file = classic_file(); + let parms = "[<< /Name /Identity >> null]"; + let header = format!("902 0 903 {} ", parms.len() + 1); + let members = format!("{header}{parms} << /Keep (yes) >>"); + let (mut rows, mut before) = (vec![1u8], 0u8); + for b in members.bytes() { + rows.push(b.wrapping_sub(before)); + before = b; + } + let coded = pdfcraft_cos::Stream::flate(pdfcraft_cos::Dict::new(), &rows).raw; + let mut object_stream = + format!("<< /Type /ObjStm /N 2 /First {} /Filter /FlateDecode /DecodeParms 911 0 R /Length {} >>\nstream\n", header.len(), coded.len()) + .into_bytes(); + object_stream.extend_from_slice(&coded); + object_stream.extend_from_slice(b"\nendstream"); + let data = [2u8, 0x03, 0x8E, 0, 2, 0x03, 0x8E, 1]; + let mut xref = format!("<< /Type /XRef /Size 2000 /W [1 2 1] /Index [902 2] /Length {} >>\nstream\n", data.len()).into_bytes(); + xref.extend_from_slice(&data); + xref.extend_from_slice(b"\nendstream"); + let string = |key: &[u8]| format!("<{}>", hex(&file.dict.get(key).unwrap().as_string().unwrap().bytes)); + let objects = [ + (file.root.num, format!("<< /Type /Catalog /Pages {} 0 R /A 900 0 R /B 903 0 R >>", file.pages.num).into_bytes()), + (file.encrypt.num, encrypt_with(&file, &string(b"O"), "Identity")), + (900, b"<< /Filter [/Crypt /ASCIIHexDecode] /DecodeParms 902 0 R /Length 1 >>\nstream\n>\nendstream".to_vec()), + (910, object_stream), + (911, format!("<< /Predictor 11 /Columns {} >>", members.len()).into_bytes()), + (912, xref), + ]; + let trailer = format!("/Root {} 0 R /Encrypt {} 0 R /ID [<{id}> <{id}>] /XRefStm @912", file.root.num, file.encrypt.num, id = hex(&file.id)); + let bytes = updated(&file.bytes, &objects, &trailer); + let keep = + |doc: &Document, r: &Object| doc.resolve(r).as_dict().and_then(|d| d.get(b"Keep")).and_then(Object::as_string).map(|s| s.bytes.clone()); + let member = Object::Ref(ObjRef::new(903, 0)); + let doc = Document::open(Arc::new(bytes.clone())).unwrap(); + assert_eq!(keep(&doc, &member).as_deref(), Some(b"yes".as_slice()), "read first"); + let doc = Document::open(Arc::new(bytes.clone())).unwrap(); + let _ = doc.get(ObjRef::new(900, 0)); + assert_eq!(keep(&doc, &member).as_deref(), Some(b"yes".as_slice()), "read after the stream selecting from it"); + let saved = write_full(&Document::open(Arc::new(bytes)).unwrap(), &SaveOptions::default()).unwrap(); + let back = Document::open(Arc::new(saved)).unwrap(); + let catalog = back.get(back.root().unwrap()); + assert_eq!(keep(&back, catalog.as_dict().unwrap().get(b"B").unwrap()).as_deref(), Some(b"yes".as_slice()), "after a full save"); +} + +/// `walk` goes as deep as an object nests, so following a reference from down there adds that depth +/// again for each object it loads: this file, under 10 KB, overflowed a 2 MiB stack and took the +/// process with it. Only the object itself follows references, and it does so in very little stack. +#[test] +fn nested_dictionaries_load_nothing_on_a_small_stack() { + let file = classic_file(); + let first = 960u32; + let objects: Vec<(u32, Vec)> = (first..first + 40) + .map(|num| { + let open = "[".repeat(120); + let close = "]".repeat(120); + (num, format!("{open}<< /Contents () /Type {} 0 R >>{close}", num + 1).into_bytes()) + }) + .collect(); + let trailer = format!("/Root {} 0 R /Encrypt {} 0 R /ID [<{id}> <{id}>]", file.root.num, file.encrypt.num, id = hex(&file.id)); + let bytes = updated(&file.bytes, &objects, &trailer); + let reading = std::thread::Builder::new().stack_size(1 << 20).spawn(move || { + let doc = Document::open(Arc::new(bytes)).unwrap(); + doc.get(ObjRef::new(first, 0)).as_array().is_some() + }); + assert_eq!(reading.unwrap().join().ok(), Some(true)); +} + +/// The object itself can follow references, a ladder of them as long as loading may nest, and that +/// fits a small stack too. +#[test] +fn a_ladder_of_signature_checks_fits_a_small_stack() { + let file = classic_file(); + let first = 960u32; + let objects: Vec<(u32, Vec)> = + (first..first + 40).map(|num| (num, format!("<< /Contents () /Type {} 0 R >>", num + 1).into_bytes())).collect(); + let trailer = format!("/Root {} 0 R /Encrypt {} 0 R /ID [<{id}> <{id}>]", file.root.num, file.encrypt.num, id = hex(&file.id)); + let bytes = updated(&file.bytes, &objects, &trailer); + let reading = std::thread::Builder::new().stack_size(1 << 20).spawn(move || { + let doc = Document::open(Arc::new(bytes)).unwrap(); + doc.get(ObjRef::new(first, 0)).as_dict().is_some() + }); + assert_eq!(reading.unwrap().join().ok(), Some(true)); +} + +/// The strings of `/Encrypt` are never encrypted (ISO 32000-2 §7.6.2), and can be objects of their +/// own. Read as ordinary strings they were decrypted, and a full save encrypted them again under +/// another number, so the file it wrote no longer opened with its own password. +#[test] +fn encrypt_strings_given_by_reference_stay_as_stored() { + let file = classic_file(); + let string = |key: &[u8]| hex(&file.dict.get(key).unwrap().as_string().unwrap().bytes); + let replaced = format!( + "<< /Filter /Standard /V 4 /R 4 /Length 128 /P {} /O 900 0 R /U 901 0 R /CF << /StdCF << /CFM /AESV2 /Length 16 /AuthEvent /DocOpen >> >> /StmF /StdCF /StrF /StdCF >>", + file.dict.int(b"P").unwrap() + ); + let objects = [ + (file.encrypt.num, replaced.into_bytes()), + (900, format!("<{}>", string(b"O")).into_bytes()), + (901, format!("<{}>", string(b"U")).into_bytes()), + ]; + let trailer = + format!("/Root {} 0 R /Info {} 0 R /Encrypt {} 0 R /ID [<{id}> <{id}>]", file.root.num, file.info.num, file.encrypt.num, id = hex(&file.id)); + let source = Document::open(Arc::new(updated(&file.bytes, &objects, &trailer))).unwrap(); + assert_eq!(content(&source), SECRET, "the file reads"); + assert_eq!(title(&source).as_deref(), Some("Quarterly numbers")); + for object_streams in [false, true] { + let saved = write_full(&source, &SaveOptions { object_streams, ..SaveOptions::default() }).unwrap(); + let back = Document::open(Arc::new(saved)).unwrap(); + assert_eq!(content(&back), SECRET, "object streams: {object_streams}"); + assert_eq!(title(&back).as_deref(), Some("Quarterly numbers"), "object streams: {object_streams}"); + } +} + +/// A save keeps what a file is protected with: its `/Encrypt`, which holds the handler and the +/// permissions, its `/ID`, and so its file key, under every algorithm and by either kind of save. +#[test] +fn a_save_keeps_the_handler_the_key_and_the_permissions() { + for alg in ALL { + let doc = Document::open_with_password(Arc::new(protect(alg, "user", "owner", -1 ^ 0b1_0000)), Some("user")).unwrap(); + let held = |d: &Document| { + let encrypt = d.resolve(d.trailer().get(b"Encrypt").unwrap()).as_dict().unwrap().clone(); + (encrypt, d.trailer().get(b"ID").cloned(), d.security().unwrap().file_key().to_vec(), format!("{:?}", d.permissions())) + }; + let before = held(&doc); + let mut edited = doc.clone(); + let _ = edited.add(Object::Int(1)); + for (save, saved) in [ + ("incremental", write_incremental(&edited, &SaveOptions::default()).unwrap()), + ("full", write_full(&doc, &SaveOptions::default()).unwrap()), + ] { + let back = Document::open_with_password(Arc::new(saved), Some("user")).unwrap(); + assert!(held(&back) == before, "{alg:?}: a {save} save"); + assert_eq!(content(&back), SECRET, "{alg:?}: a {save} save"); + } + } +} + +/// `/ID` has to be direct in an encrypted file, but one that isn't has its key made from the same +/// bytes. A full save writes it direct, and an incremental one leaves it as it was. +#[test] +fn a_file_identifier_given_by_reference_is_read() { + let file = classic_file(); + let id = hex(&file.id); + let reference = |entries: &str, objects: &[(u32, Vec)]| { + let trailer = format!("/Root {} 0 R /Encrypt {} 0 R /ID {entries}", file.root.num, file.encrypt.num); + updated(&file.bytes, objects, &trailer) + }; + let whole = reference("900 0 R", &[(900, format!("[<{id}> <{id}>]").into_bytes())]); + let entries = reference("[900 0 R 901 0 R]", &[(900, format!("<{id}>").into_bytes()), (901, format!("<{id}>").into_bytes())]); + for file in [whole, entries] { + let source = Document::open(Arc::new(file)).unwrap(); + assert_eq!(content(&source), SECRET); + let back = Document::open(Arc::new(write_full(&source, &SaveOptions::default()).unwrap())).unwrap(); + assert_eq!(content(&back), SECRET); + assert!(matches!(back.trailer().get(b"ID"), Some(Object::Array(_))), "written direct"); + let mut edited = source.clone(); + let _ = edited.add(Object::Int(1)); + let back = Document::open(Arc::new(write_incremental(&edited, &SaveOptions::default()).unwrap())).unwrap(); + assert_eq!(content(&back), SECRET, "an incremental save"); + assert_eq!(back.trailer().get(b"ID"), edited.trailer().get(b"ID"), "and leaves it as it was"); + } +} + +/// A direct `/Encrypt` can refer to objects. A full save dropped those it found only there, and kept +/// its references as they were under the old numbers, so nothing in the file read afterwards. +#[test] +fn a_direct_encrypt_dictionary_keeps_what_it_refers_to() { + for object_streams in [false, true] { + let mut doc = Document::open(Arc::new(fixture())).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes128, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + let mut dict = doc.get(doc.trailer().reference(b"Encrypt").unwrap()).as_dict().cloned().unwrap(); + let filter = doc.add(Object::name("StdCF")); + let filters = doc.add(dict.get(b"CF").unwrap().clone()); + dict.set(b"StmF".to_vec(), Object::Ref(filter)); + dict.set(b"CF".to_vec(), Object::Ref(filters)); + doc.trailer_mut().set(b"Encrypt".to_vec(), Object::Dict(dict)); + let options = SaveOptions { object_streams, ..SaveOptions::default() }; + let saved = write_full(&doc, &options).unwrap(); + let back = Document::open(Arc::new(saved)).unwrap(); + assert_eq!(content(&back), SECRET, "object streams: {object_streams}"); + let again = write_full(&back, &options).unwrap(); + assert_eq!(content(&Document::open(Arc::new(again)).unwrap()), SECRET, "object streams: {object_streams}, saved again"); + } +} + +/// How a stream is treated can be given by reference, and what that loads can be a stream too. A +/// file can make a ladder of those, where each rung reads the rung below, then two trees wider than +/// the object cache, which push that rung out of it, then the rung below again. Each rung would +/// take twice what the one below does, if one read could follow any number of references. +#[test] +fn selectors_that_lead_to_selectors_load_at_once() { + let mut doc = Document::open(Arc::new(protect(Algorithm::Aes128, "", "owner", -1))).unwrap(); + let stream_of = |doc: &mut Document, filters: Vec| { + let mut dict = pdfcraft_cos::Dict::new(); + dict.set(b"Filter".to_vec(), Object::Array(filters)); + doc.add(Object::Stream(pdfcraft_cos::Stream::from_raw(dict, Vec::new()))) + }; + let tree = |doc: &mut Document| { + let mut level: Vec = (0..6_561).map(|i| doc.add(Object::Int(i))).collect(); + while level.len() > 1 { + let chunks: Vec> = level.chunks(9).map(|c| c.iter().map(|r| Object::Ref(*r)).collect()).collect(); + level = chunks.into_iter().map(|c| stream_of(doc, c)).collect(); + } + Object::Ref(level[0]) + }; + let (first, second) = (tree(&mut doc), tree(&mut doc)); + let mut top = doc.add(Object::Int(0)); + // Twelve rungs of two, and the trees' five levels, stay inside the 32 that loading may nest. + for _ in 0..12 { + let (a, b) = (stream_of(&mut doc, vec![Object::Ref(top)]), stream_of(&mut doc, vec![Object::Ref(top)])); + top = stream_of(&mut doc, vec![Object::Ref(a), first.clone(), second.clone(), Object::Ref(b)]); + } + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + let (done, finished) = std::sync::mpsc::channel(); + std::thread::spawn(move || { + let doc = Document::open(Arc::new(saved)).unwrap(); + let _ = done.send(doc.get(top).as_dict().is_some()); + }); + // With the budget it takes milliseconds, and without it longer than anyone would wait, so a + // margin this wide can't fail on a busy machine. + assert_eq!(finished.recv_timeout(std::time::Duration::from_secs(120)), Ok(true), "loading must not fan out"); +} diff --git a/crates/render/src/raster/tests.rs b/crates/render/src/raster/tests.rs index 9c5d75bfbc..a5d9b57aba 100644 --- a/crates/render/src/raster/tests.rs +++ b/crates/render/src/raster/tests.rs @@ -9127,3 +9127,271 @@ fn content_after_an_inline_image_holding_ei_is_still_read() { // `EI` inside the data was not taken for the end. assert!(page.contains("AFTER"), "the content after the image was lost: {page:?}"); } + +/// `/Crypt` has nothing to undo, so a chain naming it thousands of times costs no copy of the data +/// for each. A chain of filters that does work has its own cost, which this isn't about. +#[test] +fn a_long_chain_of_crypt_filters_takes_no_time() { + use super::*; + let content = format!("1 0 0 rg 0 0 200 200 re f\n{}", " ".repeat(16 << 20)); + let chain = "/Crypt ".repeat(100_000); + let 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 0 200 200] /Contents 4 0 R >> endobj\n\ + 4 0 obj << /Filter [{chain}] /Length {} >> stream\n{content}\nendstream endobj\n\ + trailer << /Root 1 0 R >>\n%%EOF", + content.len() + ); + let (done, finished) = std::sync::mpsc::channel(); + std::thread::spawn(move || { + let page = PageRenderer::new(Arc::new(pdf.into_bytes()), RenderConfig::default()).render(RenderRequest { + page: 0, + kind: RequestKind::Pixels, + tile: None, + scale: 1.0, + tag: 0, + preview: false, + }); + let start = (100 * page.width as usize + 100) * 4; + let _ = done.send(page.rgba.get(start..start + 4).map(<[u8]>::to_vec)); + }); + // Skipped, this takes a moment; copied, 1.6 TB, which a margin this wide tells apart however + // busy the machine. + let pixel = finished.recv_timeout(std::time::Duration::from_secs(120)).expect("the chain costs no copy for each entry"); + assert_eq!(pixel, Some(vec![255, 0, 0, 255])); +} + +/// A lone filter takes its parameters as a dictionary or as an array of one, and given as an array +/// of one itself it takes either too. The form of `/Filter` decided which was read, so a predictor +/// in the other form was lost and the content came out as noise. `/StmF` and `/StrF` are optional, +/// meaning `/Identity`: such a file didn't open at all, though a stream can still name a filter. +#[test] +fn a_filters_parameters_are_read_in_either_form() { + use super::*; + use pdfcraft_cos::{Algorithm, Dict, Document, NewEncryption, Object, SaveOptions, Stream, write_full, write_incremental}; + let content = b"1 0 0 rg 0 0 200 200 re f"; + let centre = |bytes: Vec| -> [u8; 4] { + let page = PageRenderer::new(Arc::new(bytes), RenderConfig::default()).render(RenderRequest { + page: 0, + kind: RequestKind::Pixels, + tile: None, + scale: 1.0, + tag: 0, + preview: false, + }); + let start = (100 * page.width as usize + 100) * 4; + let mut pixel = [0; 4]; + pixel.copy_from_slice(page.rgba.get(start..start + 4).expect("centre pixel")); + pixel + }; + let page = |stream: &[u8]| { + 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 0 200 200] /Contents 4 0 R >> endobj\n".to_vec(); + pdf.extend_from_slice(stream); + pdf.extend_from_slice(b"\ntrailer << /Root 1 0 R >>\n%%EOF\n"); + pdf + }; + // PNG Sub over one row, which reads as noise if it isn't undone. + let mut rows = vec![1u8]; + let mut before = 0u8; + for b in content { + rows.push(b.wrapping_sub(before)); + before = *b; + } + let coded = Stream::flate(Dict::new(), &rows).raw; + let parms = format!("<< /Predictor 11 /Columns {} >>", content.len()); + for (filter, decode_parms) in [ + ("/FlateDecode", parms.clone()), + ("/FlateDecode", format!("[{parms}]")), + ("[/FlateDecode]", parms.clone()), + ("[/FlateDecode]", format!("[{parms}]")), + ] { + let mut stream = format!("4 0 obj << /Filter {filter} /DecodeParms {decode_parms} /Length {} >> stream\n", coded.len()).into_bytes(); + stream.extend_from_slice(&coded); + stream.extend_from_slice(b"\nendstream endobj"); + assert_eq!(centre(page(&stream)), [255, 0, 0, 255], "{filter} with {decode_parms}"); + } + + let plain = format!("4 0 obj << /Length {} >> stream\n{}\nendstream endobj", content.len(), String::from_utf8_lossy(content)); + let mut doc = Document::open(Arc::new(page(plain.as_bytes()))).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm: Algorithm::Aes128, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + // No object streams: they would be encrypted under the /StmF about to go. + let saved = write_full(&doc, &SaveOptions { object_streams: false, ..SaveOptions::default() }).unwrap(); + let mut doc = Document::open(Arc::new(saved)).unwrap(); + let encrypt = doc.trailer().reference(b"Encrypt").unwrap(); + doc.update_dict(encrypt, |e| { + e.remove(b"StmF"); + e.remove(b"StrF"); + }) + .unwrap(); + let mut doc = Document::open(Arc::new(write_incremental(&doc, &SaveOptions::default()).unwrap())).unwrap(); + let catalog = doc.get(doc.root().unwrap()); + let pages = doc.get(catalog.as_dict().unwrap().reference(b"Pages").unwrap()); + let first = doc.resolve(&pages.as_dict().unwrap().get(b"Kids").unwrap().as_array().unwrap()[0]); + let contents = first.as_dict().unwrap().reference(b"Contents").unwrap(); + let mut named = Dict::new(); + named.set(b"Name".to_vec(), Object::name("StdCF")); + let mut dict = Dict::new(); + dict.set(b"Filter".to_vec(), Object::name("Crypt")); + dict.set(b"DecodeParms".to_vec(), Object::Dict(named)); + doc.set(contents, Object::Stream(Stream::from_raw(dict, content.to_vec()))); + assert_eq!(centre(write_incremental(&doc, &SaveOptions::default()).unwrap()), [255, 0, 0, 255], "no /StmF, and a stream naming StdCF"); +} + +/// A stream can pick its own crypt filter: `/Crypt` in `/Filter`, with the filter's name in +/// `/DecodeParms`, or `/Identity` without one (ISO 32000-2 §7.4.10). The renderer used the +/// document's filter for every stream, and then couldn't decode `/Crypt`, so the page stayed blank. +#[test] +fn a_stream_with_a_crypt_filter_of_its_own_is_drawn() { + use super::*; + use pdfcraft_cos::{Algorithm, Dict, Document, NewEncryption, Object, SaveOptions, Stream, write_full, write_incremental}; + let content = "1 0 0 rg 0 0 200 200 re f"; + let page = 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 0 200 200] /Contents 4 0 R >> endobj\n\ + 4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\ + trailer << /Root 1 0 R >>\n%%EOF", + content.len() + ); + let centre = |bytes: Vec| -> [u8; 4] { + let page = PageRenderer::new(Arc::new(bytes), RenderConfig::default()).render(RenderRequest { + page: 0, + kind: RequestKind::Pixels, + tile: None, + scale: 1.0, + tag: 0, + preview: false, + }); + let start = (100 * page.width as usize + 100) * 4; + let mut pixel = [0; 4]; + pixel.copy_from_slice(page.rgba.get(start..start + 4).expect("centre pixel")); + pixel + }; + let name = |n: &str| { + let mut parms = Dict::new(); + parms.set(b"Name".to_vec(), Object::name(n)); + Object::Dict(parms) + }; + let hex: String = content.bytes().map(|b| format!("{b:02X}")).chain([">".to_string()]).collect(); + let crypt = || Object::name("Crypt"); + let contents_of = |doc: &Document| { + let catalog = doc.get(doc.root().unwrap()); + let pages = doc.get(catalog.as_dict().unwrap().reference(b"Pages").unwrap()); + let page = doc.resolve(&pages.as_dict().unwrap().get(b"Kids").unwrap().as_array().unwrap()[0]); + page.as_dict().unwrap().reference(b"Contents").unwrap() + }; + for (algorithm, filters) in [(Algorithm::Rc4_128, false), (Algorithm::Aes128, true), (Algorithm::Aes256, true)] { + let mut doc = Document::open(Arc::new(page.clone().into_bytes())).unwrap(); + doc.set_encryption(&NewEncryption { + algorithm, + user_password: "", + owner_password: "owner", + permissions: -1, + encrypt_metadata: true, + seed: [9; 32], + }) + .unwrap(); + let protected = write_full(&doc, &SaveOptions::default()).unwrap(); + assert_eq!(centre(protected.clone()), [255, 0, 0, 255], "{algorithm:?}: the document's own filter"); + // A filter of its own name that stores streams as they are, beside the document's. + let mut doc = Document::open(Arc::new(protected)).unwrap(); + if filters { + let encrypt = doc.trailer().reference(b"Encrypt").unwrap(); + doc.update_dict(encrypt, |e| { + let Some(Object::Dict(cf)) = e.get(b"CF") else { panic!("no /CF") }; + let mut cf = cf.clone(); + let mut stored = Dict::new(); + stored.set(b"CFM".to_vec(), Object::name("None")); + cf.set(b"Stored".to_vec(), Object::Dict(stored)); + // `/Identity` is never one of these, whatever a file says. + let mut named = Dict::new(); + named.set(b"CFM".to_vec(), Object::name("AESV2")); + cf.set(b"Identity".to_vec(), Object::Dict(named)); + e.set(b"CF".to_vec(), Object::Dict(cf)); + }) + .unwrap(); + } + let protected = write_incremental(&doc, &SaveOptions::default()).unwrap(); + let spellings = [ + "no name", + "/Identity", + "StdCF", + "Stored", + "/DecodeParms by reference", + "/Filter by reference", + "an entry of /Filter by reference", + "an entry of /DecodeParms by reference", + "/Name by reference", + "before another filter", + "a lone dictionary for an array of one", + "an array of one for a name", + "a dictionary by reference for an array of one", + "an array of one by reference for a dictionary", + "a name by reference for an array", + "a lone dictionary before another filter", + "past the longest chain", + ]; + for what in spellings { + let mut doc = Document::open(Arc::new(protected.clone())).unwrap(); + let (filter, parms) = match what { + "no name" => (crypt(), None), + "/Identity" => (crypt(), Some(name("Identity"))), + "StdCF" => (crypt(), Some(name("StdCF"))), + "Stored" => (crypt(), Some(name("Stored"))), + "/DecodeParms by reference" => (crypt(), Some(Object::Ref(doc.add(name("StdCF"))))), + "/Filter by reference" => (Object::Ref(doc.add(crypt())), Some(name("StdCF"))), + "an entry of /Filter by reference" => (Object::Array(vec![Object::Ref(doc.add(crypt()))]), Some(Object::Array(vec![name("StdCF")]))), + "an entry of /DecodeParms by reference" => { + (Object::Array(vec![crypt()]), Some(Object::Array(vec![Object::Ref(doc.add(name("StdCF")))]))) + } + "/Name by reference" => { + let mut parms = Dict::new(); + parms.set(b"Name".to_vec(), Object::Ref(doc.add(Object::name("StdCF")))); + (crypt(), Some(Object::Dict(parms))) + } + "a lone dictionary for an array of one" => (Object::Array(vec![crypt()]), Some(name("StdCF"))), + "an array of one for a name" => (crypt(), Some(Object::Array(vec![name("StdCF")]))), + "a dictionary by reference for an array of one" => (Object::Array(vec![crypt()]), Some(Object::Ref(doc.add(name("StdCF"))))), + "an array of one by reference for a dictionary" => (Object::Ref(doc.add(Object::Array(vec![crypt()]))), Some(name("StdCF"))), + "a name by reference for an array" => (Object::Ref(doc.add(crypt())), Some(Object::Array(vec![name("StdCF")]))), + "a lone dictionary before another filter" => (Object::Array(vec![crypt(), Object::name("ASCIIHexDecode")]), Some(name("StdCF"))), + // One filter more than cos decodes: it names none there, so here neither, and the + // stream is read with the document's filter, not the `Stored` it says. + "past the longest chain" => { + let mut chain = vec![crypt()]; + chain.extend(std::iter::repeat_n(Object::name("ASCIIHexDecode"), 9)); + (Object::Array(chain), Some(Object::Array(vec![name("Stored")]))) + } + _ => (Object::Array(vec![crypt(), Object::name("ASCIIHexDecode")]), Some(Object::Array(vec![name("StdCF"), Object::Null]))), + }; + let contents = contents_of(&doc); + let Object::Stream(s) = doc.get(contents).as_ref().clone() else { panic!("no content stream") }; + let mut dict = s.dict.clone(); + dict.set(b"Filter".to_vec(), filter); + if let Some(parms) = parms { + dict.set(b"DecodeParms".to_vec(), parms); + } + let raw = match what { + "before another filter" | "a lone dictionary before another filter" => hex.clone().into_bytes(), + "past the longest chain" => (0..9).fold(content.as_bytes().to_vec(), |data, _| { + let mut coded: Vec = data.iter().flat_map(|b| format!("{b:02X}").into_bytes()).collect(); + coded.push(b'>'); + coded + }), + _ => content.as_bytes().to_vec(), + }; + doc.set(contents, Object::Stream(Stream::from_raw(dict, raw))); + let saved = write_incremental(&doc, &SaveOptions::default()).unwrap(); + assert_eq!(centre(saved), [255, 0, 0, 255], "{algorithm:?}: {what}"); + } + } +} diff --git a/parity/acrobat-features.toml b/parity/acrobat-features.toml index c6c520eda7..fc5a0d3fb8 100644 --- a/parity/acrobat-features.toml +++ b/parity/acrobat-features.toml @@ -226,8 +226,8 @@ milestone = "M1" status = "shipped" commands = [] tools = [] -evidence = ["crates/crypt/src/tests.rs::identity_crypt_filters_leave_data_alone"] -notes = "" +evidence = ["crates/crypt/src/tests.rs::identity_crypt_filters_leave_data_alone", "crates/cos/tests/it/encryption.rs::a_crypt_filter_selected_by_reference_is_used", "crates/cos/tests/it/encryption.rs::an_identity_crypt_filter_selected_by_reference_leaves_a_stream_alone", "crates/render/src/raster/tests.rs::a_stream_with_a_crypt_filter_of_its_own_is_drawn"] +notes = "a stream can name its own filter (/Crypt in /Filter, its name in /DecodeParms), read and written, each part given directly or by reference, in cos and in the renderer; /StmF, /StrF, /EFF and /CFM by reference too. A /Filter given by reference selects the crypt filter, but Stream::decoded still refuses it, so such a stream is decrypted and not decoded outside the renderer." [[feature]] id = "core.attachment-only-encryption" diff --git a/vendor/README.md b/vendor/README.md index bf0f5f9129..1fc2f75135 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, 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, interpolated image 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 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); (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) | (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_samples_the_same_columns`, `raster::tests::non_integer_minification_samples_colour_and_mask_from_the_same_columns`, `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::interpolated_minified_image_edges_snap_to_whole_pixels` (`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`; (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`) | | 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 | (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` | | 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; (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 | (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`; (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) | +| 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 | (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) | | 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/crypto/mod.rs b/vendor/hayro-syntax/src/crypto/mod.rs index 15371514c1..79d182c6ac 100644 --- a/vendor/hayro-syntax/src/crypto/mod.rs +++ b/vendor/hayro-syntax/src/crypto/mod.rs @@ -74,9 +74,11 @@ pub(crate) enum Decryptor { } #[derive(Debug, Copy, Clone)] -pub(crate) enum DecryptionTarget { +pub(crate) enum DecryptionTarget<'a> { String, Stream, + /// PdfCraft patch (30): a stream that names its own crypt filter. + NamedStream(&'a str), } impl Decryptor { @@ -84,7 +86,7 @@ impl Decryptor { &self, id: ObjectIdentifier, data: &[u8], - target: DecryptionTarget, + target: DecryptionTarget<'_>, ) -> Option> { match self { Self::None => Some(data.to_vec()), @@ -93,6 +95,8 @@ impl Decryptor { let crypt_dict = match target { DecryptionTarget::String => dict.string_filter, DecryptionTarget::Stream => dict.stream_filter, + // `/Identity` isn't in the table, and a name it hasn't got reads as that. + DecryptionTarget::NamedStream(name) => dict.named.get(name).copied().unwrap_or(CryptDictionary::identity(0)), }; match crypt_dict.cfm { @@ -292,10 +296,12 @@ fn decrypt_rc_aes( with_key(final_key) } -#[derive(Debug, Copy, Clone)] +#[derive(Debug, Clone)] pub(crate) struct DecryptorData { stream_filter: CryptDictionary, string_filter: CryptDictionary, + /// PdfCraft patch (30): `/CF`, for the streams that name a filter of their own. + named: HashMap, } impl DecryptorData { @@ -312,16 +318,23 @@ impl DecryptorData { } } - let stm_f = *mappings - .get(dict.get::>(STM_F)?.as_str()) - .unwrap_or(&CryptDictionary::identity(default_length)); - let str_f = *mappings - .get(dict.get::>(STR_F)?.as_str()) - .unwrap_or(&CryptDictionary::identity(default_length)); + // PdfCraft patch (30): `/Identity` is never an entry of `/CF`, whatever a file puts there, + // and `/StmF` and `/StrF` are optional, naming it by default (ISO 32000-2 Table 20). + mappings.remove("Identity"); + let identity = CryptDictionary::identity(default_length); + let stm_f = dict + .get::>(STM_F) + .and_then(|n| mappings.get(n.as_str()).copied()) + .unwrap_or(identity); + let str_f = dict + .get::>(STR_F) + .and_then(|n| mappings.get(n.as_str()).copied()) + .unwrap_or(identity); Some(Self { stream_filter: stm_f, string_filter: str_f, + named: mappings, }) } } diff --git a/vendor/hayro-syntax/src/filter/mod.rs b/vendor/hayro-syntax/src/filter/mod.rs index e81a37df98..89d0512693 100644 --- a/vendor/hayro-syntax/src/filter/mod.rs +++ b/vendor/hayro-syntax/src/filter/mod.rs @@ -130,7 +130,8 @@ impl Filter { warn!("image decoding is not supported (enable the `images` feature)"); Err(DecodeFailure::ImageDecode) } - _ => Err(DecodeFailure::StreamDecode), + // PdfCraft patch (30): `Stream::raw_data` decrypted the data with the filter this names. + Self::Crypt => Ok(FilterResult::from_data(data.to_vec())), }; if res.is_err() { diff --git a/vendor/hayro-syntax/src/object/stream.rs b/vendor/hayro-syntax/src/object/stream.rs index a80f0730cb..4a0c08510f 100644 --- a/vendor/hayro-syntax/src/object/stream.rs +++ b/vendor/hayro-syntax/src/object/stream.rs @@ -5,7 +5,7 @@ use crate::filter::Filter; use crate::object; use crate::object::Dict; use crate::object::Name; -use crate::object::dict::keys::{DECODE_PARMS, DP, F, FILTER, LENGTH, TYPE}; +use crate::object::dict::keys::{DECODE_PARMS, DP, F, FILTER, LENGTH, NAME, TYPE}; use crate::object::{Array, ObjectIdentifier}; use crate::object::{Object, ObjectLike, ObjectRefLike}; use crate::reader::Reader; @@ -13,6 +13,7 @@ use crate::reader::{Readable, ReaderContext, ReaderExt, Skippable}; use crate::trivia::is_white_space_character; use crate::util::{OptionLog, find_needle}; use alloc::borrow::Cow; +use alloc::string::{String, ToString}; use alloc::vec::Vec; use core::fmt::{Debug, Formatter}; use smallvec::SmallVec; @@ -73,38 +74,44 @@ impl<'a> Stream<'a> { let mut collected_filters = SmallVec::new(); let mut collected_params = SmallVec::new(); + // PdfCraft patch (30): `/DecodeParms` in either form, whichever form `/Filter` has: one filter + // in an array with a lone dictionary is an everyday spelling, and a lone dictionary is the + // first filter's. Read as the chain asks, so a long array costs no more than a short one. + let parms = self + .dict + .get::>(DP) + .or_else(|| self.dict.get::>(DECODE_PARMS)); + let mut in_array = match &parms { + Some(Object::Array(a)) => Some(a.iter::>()), + _ => None, + }; + let mut lone = match parms { + Some(Object::Dict(d)) => Some(d), + _ => None, + }; + let mut next_params = || { + lone.take() + .or_else(|| in_array.as_mut()?.next()?.into_dict()) + .unwrap_or_default() + }; + if let Some(filter) = self .dict .get::>(F) .or_else(|| self.dict.get::>(FILTER)) .and_then(Filter::from_name) { - let params = self - .dict - .get::>(DP) - .or_else(|| self.dict.get::>(DECODE_PARMS)) - .unwrap_or_default(); - collected_filters.push(filter); - collected_params.push(params); + collected_params.push(next_params()); } else if let Some(filters) = self .dict .get::>(F) .or_else(|| self.dict.get::>(FILTER)) { let filters = filters.iter::>().map(Filter::from_name); - let mut params = self - .dict - .get::>(DP) - .or_else(|| self.dict.get::>(DECODE_PARMS)) - .map(|a| a.iter::>()); for filter in filters { - let params = params - .as_mut() - .and_then(|p| p.next()) - .and_then(|p| p.into_dict()) - .unwrap_or_default(); + let params = next_params(); if let Some(filter) = filter { collected_filters.push(filter); @@ -119,6 +126,20 @@ impl<'a> Stream<'a> { } } + /// PdfCraft patch (30): the crypt filter the stream names for itself: `/Crypt` in its chain, its + /// name in that filter's parameters, or `/Identity` without one (ISO 32000-2 §7.4.10). + fn crypt_filter(&self) -> Option { + let FiltersAndParams { filters, params } = self.filters_and_params(); + // A chain longer than cos decodes names none there, so it names none here either: + // `pdfcraft_filters::MAX_CHAIN`, and `/Crypt`. + if filters.len() > 9 { + return None; + } + let i = filters.iter().position(|f| matches!(f, Filter::Crypt))?; + let name = params.get(i).and_then(|p| p.get::>(NAME)); + Some(name.map_or_else(|| "Identity".to_string(), |n| n.as_str().to_string())) + } + /// Return the raw, decrypted data of the stream. /// /// Stream filters will not be applied. @@ -132,9 +153,13 @@ impl<'a> Stream<'a> { .map(|t| t.as_ref() != b"XRef") .unwrap_or(true) { + let own = self.crypt_filter(); + let target = own + .as_deref() + .map_or(DecryptionTarget::Stream, DecryptionTarget::NamedStream); Cow::Owned( ctx.xref() - .decrypt(self.obj_id(), self.data, DecryptionTarget::Stream) + .decrypt(self.obj_id(), self.data, target) // TODO: MAybe an error would be better? .unwrap_or_default(), ) @@ -241,6 +266,11 @@ impl<'a> Stream<'a> { .zip(filters_and_params.params.iter()) .enumerate() { + // PdfCraft patch (30): `raw_data` decrypted the data, so there's nothing to undo, and a + // chain can name this thousands of times. + if matches!(filter, Filter::Crypt) { + continue; + } // PdfCraft patch: see `filter::decode_limit`. let limit = crate::filter::decode_limit(image_params, i + 1 == count).min(max); let new = filter.apply( diff --git a/vendor/hayro-syntax/src/xref.rs b/vendor/hayro-syntax/src/xref.rs index d3ab7d324f..dd3ebb8b56 100644 --- a/vendor/hayro-syntax/src/xref.rs +++ b/vendor/hayro-syntax/src/xref.rs @@ -651,7 +651,7 @@ impl XRef { &self, id: ObjectIdentifier, data: &[u8], - target: DecryptionTarget, + target: DecryptionTarget<'_>, ) -> Option> { match &self.0 { Inner::Dummy => Some(data.to_vec()),