xtask pdf-deps still expected craft-pdf and the patched hayro/krilla
backends at main's 9be9a637 revision after this branch pinned them to
a499742a. Point REV at that pin (still on craft-libs main). Classify
test files in sonar-project.properties the way craft-libs does, so the
quality gate measures the product. Give the EPS operation-budget test
the same 300s cap as the other budget tests; under instrumented coverage
it overran the old 120s cap, while the Ubuntu gate finished in time.
A stored texture longer than the codec's default 2^18-pixel side still
reads back, and a gray PNG is not expanded to RGBA just to get its
levels. The AI writer and the EPS scene refuse an oversized image from
its header, then decode within MAX_IMAGE_PIXELS, keeping the pixels at
their stored depth. Flatten's see-through check uses ImageBlob's
allowance, and links.unembed decodes within MAX_RASTER_ALLOC.
A clipping object written as art cut to it (a group, an image, a symbol
instance, a live object, type, or a path inside art already cut to
another clipping object) lost its name, opacity mask and variable
bindings in the editing data: no written object carried its dictionary,
and a painted one put its name on each of its fill and stroke objects.
The unpainted path that carried its text wrap options now carries its
whole dictionary, in the form the .ai reader reads (art_dictionary in
crates/eps/src/import/ai/read/document.rs). Its fills and strokes keep
only the opacity mask they are drawn through. The path is admitted by
the object count, so at the object limit its dictionary is never left
on the object before it, and a path whose dictionary comes out empty is
taken out again.
Remove Background and Color Mask check the image's size from its header
again before decoding ("an image of W × H pixels is too large to change"),
and Crop Image's trim and Remove Background read a file that can't hold
the pixels its header declares as unreadable, as on main. The test helper
reading Remove Background's output keeps 16-bit samples.
The HEIF and JPEG XL readers and their tests go back to main's (heic-rs and
jxl-oxide): the craft-heif and craft-jxl versions failed eight of their
tests (resolution axes after a quarter turn, the container's mirror and
crop, refusing a grid of cropped tiles, 16-bit samples, the decode
allowance). Moving them stays to do.
EPS fax images decode with craft_stream_codecs::decode_ccitt_raw instead
of hayro-ccitt directly. With EndOfBlock the rows end at the end-of-block
code whatever Rows says (rows 0), as before. Needs craft-libs#1772
(rows on bytes exactly as EncodedByteAlign says).
Area type with wrap shapes was always written as outlines in the editing
data for other apps. It is now written as type when the reader works the
same wrap out again from the objects above it: each wrap object is a path
of one subpath, a compound path of visible members, or an unclipped group
of such objects, with no paint, or a solid fill, a plain solid stroke or
both, without effects, art style, opacity mask or clipping, with an
offset within -1000 to 1000 pt; the type's stored shapes match the ones
those objects give it; and the shapes of all area type stay within the
2,000,000 anchors the reader copies. Each frame of a thread is checked on
its own. When art is left out at the object limit, such type is written
as outlines again. Other wrapped area type stays outlines.
The reader recomputes the wrap with resolve_wrap after reading
(resolve_text_wrap in crates/eps/src/import/native/build.rs) and reads the
options with text_wrap in crates/eps/src/import/ai/read/document.rs; the
round-trip tests check the reopened characters, frame, wrap shapes and
wrap objects. resolve_wrap's layer walk is shared through a new
wrap_sources so the check follows the same traversal.
Graph Type has a Line Width field next to Draw Filled Lines, and
`graph.setType` takes `lineWidth` in points (0.01 to 1000; 3 when the
graph sets none). Filled lines are drawn at that width, and the width is
kept through save and open. A graph stores a width only when it is not
3 pt, and a graph with a width of its own is written to .ai as its art.
`graph.setType` with options that leave the graph as it is records no
undo step.
EMF, WMF and SVG wrote a symbol instance from its symbol's art as stored,
so live effects inside that art were missing: a square moved 100 pt by a
Transform effect stayed in place in the file while the canvas drew it
moved. Art whose own place is off the artboard was left out even when
the effect moved it onto it.
These exports now call bake_document_placing_symbols, as PDF, print, EPS
and DXF do: each instance of a symbol whose art needs baking is written
as that art placed by the instance and then baked, so an instance that
scales or turns its art does not scale what its effects do. EMF and SVG
clip 9-slice cells and WMF does not, as each places other instances. In
SVG those instances are written as groups; instances of symbols without
such effects still share one <symbol>. Preserve Editing still embeds the
document with its live effects.
WrittenLayers gains hidden_objects. SVG sets it with Hidden Layers on,
since it then writes hidden objects too, so a hidden instance is written
hidden with its effects applied. The other exporters leave it off.
An SVG instance nested in another instance's art was shared through
<use> when the def allowed its own transform, while the <use> writes
that transform after the outer instances'. A stroked symbol nested in a
scaled instance then drew its stroke at double weight. The check now
uses the transform the <use> writes.
The editing copy of a large canvas wrote every envelope as its drawn
art. It now writes warp, mesh and top object envelopes live at 1/10 of
their size with the rest of the art, under the same checks as on a
normal canvas, and the reader opens them at their size from the editing
data alone or beside the PDF part.
Read from the editing data alone, a Large Canvas copy also gets:
- an envelope's saved box size compared with its contents at the
declared scale, so a bounds margin kept with Distort Appearance is no
longer lost;
- every path's stroke weight, dashes and dash offset at the declared
scale, not at 1/10.
- EPS CCITTFaxDecode goes back to hayro-ccitt (main's filters.rs and data.rs):
craft-stream-codecs decodes /K -1 /Columns 70 data to other rows
(eps import::tests_images::ccitt_fax_images_draw). DCTDecode is main's again.
- The AI editing writer decodes through the image crate again: going through
RGBA8 lost 16-bit sources, so their 8-bit warning disappeared
(ai_editing::tests::sixteen_bit_images_are_written_at_eight_bits).
- Crop Image refuses an image over its allowance by header size again, with
main's "an image of W × H pixels is too large to draw" message.
- Test fixes: colorized_images writes its gray PNGs gray again, and the PDF
16→8-bit test expects rounding, as the 8-bit copy is made.
craft_image_codecs::probe_size only reads WebP, EXR and the GIF/BMP/TGA family;
for PNG, JPEG and TIFF it returns Unsupported. The cutover used it for header
sizes, so DXF export left every PNG out (cad tests), plugin objects lost PNG
pixels, and EPS image-mask and SVG import fell back to decoding whole images to
learn their size. vectorcraft_doc::header_size reads PNG's IHDR and a JPEG's
frame through craft-image-codecs and other formats through the image crate's
header readers, as main does, without decoding pixels. The SVG filter's JPEG
size comes from jpeg_info instead of a full decode.
- Test files whose image→craft-image-codecs rewrites didn't compile against
main's newer tests go back to main's versions (engine fileio/tests.rs,
tests_tiffbmp.rs, tests/menucmds.rs, eps tests.rs); they still use the image
crate, which keeps its format features.
- craft_jxl::encode, craft_image_codecs::encode and Image::icc at d6cd7836d.
- jp2read and pdf export pass DecodeOptions, not Limits.
- EPS DCTDecode: grey JPEGs no longer use the image after moving it, and keep
main's one-channel output.
- svg keeps its direct usvg dependency for main's image href code; pdf keeps
jpeg-encoder for CMYK JPEGs.
- clippy -D warnings and rustfmt clean.
Conflicts resolved in favour of main/#7664 in the 13 source files both touched
(doc/cmyk.rs, metafile/dib.rs, pdf images, svg import/files.rs, render paint and
encode tests, trace, engine load.rs), keeping main's newer TIFF/CMYK, DIB and PDF
image fixes. vectorcraft-doc stays on main's version until its CMYK reader moves
to craft-tiff. The image crate keeps its format features, and usvg, tiff,
zune-jpeg, hayro-ccitt and jpeg-encoder stay for the readers not yet moved.
Codec crates pinned to craft-libs d6cd7836d364177661b0ae17f48a0805e83a5f32.
Also fixes the compile errors CodeRabbit flagged (jp2read DecodeOptions,
filters.rs use-after-move) and the encode() arity at d6cd7836d.
An envelope record that lies in an outer envelope's AIDeformArtKey
block (X= ... X+) with no group around it now opens as a live envelope
inside the live outer one, and an object's own dictionary there (its
name) applies to it. Each envelope is still compared with its own drawn
art: an inner one that doesn't match keeps its drawn art as the outer
envelope's contents, and an outer one that doesn't match keeps its own.
- An envelope record in an opacity mask's art reads as part of the
mask; its block's X+ no longer ends the mask's art, which failed the
import.
- Envelopes still waiting for their XP when a dictionary ends keep
their drawn art, so a stray XP in a later container cannot take them.
The unpainted rectangle that carries the text wrap options of a symbol
instance, blend, mesh, repeat, placed document, intertwine or envelope
written as a group of its art is now the group's first member. Area type
below the object still wraps by its bounds after reopening, and area type
in the art itself is not wrapped by the rectangle, so art that holds area
type no longer keeps the options on its group and wraps by its members.
A symbol instance written as its art, and a blend, mesh, repeat, placed
document, intertwine or envelope written as a group of its art, wrapped
area type below it by its bounds before saving but by the members of
that art after reopening, because the group carried the text wrap
options. The group now ends with an unpainted rectangle at the object's
bounds that carries the options, as type written as outlines does, and
the group keeps the object's name, transparency, lock and visibility
without options of its own. Art holding shown area type written as
type keeps the options on its group; area type in an envelope does not
count. An object without finite bounds keeps the options on its group.
Without room for the rectangle within the object limit, the options are
left out with the limit's warning.
Each later placement of an image or a colorized image that was turned away at the page limit
was decoding and compressing it again. The scene now remembers what the first placement of a
turned-away encoding adds to the page and leaves later placements out with the same warning
while that still doesn't fit. Once the page has room again, after art is taken back, the image
is encoded and placed as usual.
A subpath's anchors are now an `Anchors` list held in chunks of up to
256 anchors behind `Arc`s. Copying a path copies the chunk table and
shares the chunks, and editing one anchor copies only the chunk that
holds it, so the undo step of a one-anchor edit no longer keeps a copy
of every anchor of the path. Inserting or removing an anchor copies the
chunk it is in and rebuilds the table. Reads and edits follow the Vec
and slice methods the code used; files write and read each subpath's
anchors as one list, as before, within the same anchor read budget.
The undo memory estimate counts only the chunks the other version of
the path does not hold, and the renderer's kept paths compare only the
chunks an edit copied when they look for moved anchors.
Moving one anchor of an open path of 10,000 or 200,000 anchors
(path.moveAnchor, path.setHandle, path.moveAnchors, a nudge) keeps at
most 48 KiB in its undo and redo steps, where it kept 56 bytes per
anchor.