Follow-up to #3702 (CodeRabbit): dimensions are only redrawn when their
recipe changes. Document Color Mode converted their art through the colour
profile, so black became a four-ink C67 M67 Y67 K100 again, and CMYK files
saved before #3702 kept RGB dimension art.
dimension::refresh now also redraws when the document's colour mode
changes. dimension::conform, run when a document is opened or reverted
(Session::add_document, replace_document), redraws a dimension whose art
has colours in the wrong model; one that can't be redrawn keeps its art.
dimension_colours_follow_the_document_colour_model covers RGB to CMYK and
back, and a CMYK document with RGB dimension art being opened. Without the
change it finds C0.67 M0.67 Y0.67 K1 after Document Color Mode.
With In Merge Mode, Clicking Stroke Splits the Path on, the run a click
split off could run against its path. A click on a rectangle's top side
gave a side from (100, 0) to (0, 0) on a rectangle that runs from
(0, 0) to (100, 0) there, so arrowheads and other settings that follow
a stroke's direction moved to the other end. The run now goes the way
the path goes.
The two paths of a split, and what an Alt-click or a merge leaves of a
path, had a gradient fitted to the path's bounds fitted again to their
own bounds. They now keep it where it was on the path.
A run that left 0.05 pt or less of a path still split it, into the
run and a speck. Such a path now does not split, and the tool shows
no scissors over it.
Docs: docs/mcp.md (Shape Builder Tool Options).
Follow-up to #3697 (CodeRabbit): a brush stroke is drawn as separate
pieces (draw_brush), so folding object opacity into its colour isn't exact.
A non-dashed brush stroke already counted as one paint before #3697; to
keep that change's scope, a dashed stroke counts as one paint only when it
has no brush. dashed_stroke_opacity_folds_into_its_colour checks that a
dashed brush stroke isn't folded; it fails without the change.
The Properties panel's flip buttons had been renamed Flip Horizontal and Flip Vertical, which are the Transform panel menu's names. The test lays out the Transform section for a selected rectangle and reads the screen-reader names of its buttons: Flip Along Horizontal Axis and Flip Along Vertical Axis must be there and the old names must not. With the old labels the first is missing.
ruzstd 0.9.0 let a block decompress to more than its frame's window allows (RUSTSEC-2026-0344). The test swaps a hand-built Zstandard frame with a 1 KiB window and one 100,000-byte RLE block into an Affinity container entry and checks that reading the entry fails. With ruzstd pinned back to 0.9.0 the block is decompressed and the read succeeds.
The Fill/Stroke chip's popover closed on the click that finished a drag between its proxies. The test opens a popover and checks the three outcomes: a click elsewhere closes it, a click on its own anchor that does not toggle it keeps it open, and a click elsewhere during a panel drag keeps it open. Without the fix both of the last two close it.
The rows of the panels' menus did not light up under the pointer as the menu bar's items do, because they were drawn with no frame at all. The test hovers a menu_item_name row and checks that a filled rectangle appears behind it, and that none is there at rest. With frame(false) the hovered row stays unfilled.
The theme's widget fill matched the dialog background, so a slider in a dialog showed only its handle. The test draws widgets::slider in all four brightnesses and checks that a rail in the buttons' border colour and an accent fill from the rail's start up to the handle are painted. With a plain egui slider neither is drawn, only two rectangles in the panel's button fill.
Selected text in a value field was drawn in the accent on a selection tinted with the same accent, hard to read in every brightness. The tests focus a panel number field and a dialog text field, select their text and read the colours of the glyphs drawn, in all four brightnesses: all must be the strong text colour. Without strong_selected_text the glyphs come out in the accent.
A number field with a spinner and preset chevron drew as three separate boxes, each joint with two borders. The test lays out a spin field and checks that its three boxes are the row's height and that neighbours overlap by one point so they share a border. Without the overlap (item spacing 0) the boxes touch edge to edge instead.
The toolbar's and the dock icon column's scroll bar widened to 10 points while the pointer was on it, covering the group labels and the tool icons' right edge. The test lays out a scrolling strip, rests the pointer on its bar and measures the widest filled rectangle drawn: at most 4 points. Without the slim scroll style it measures 10 points.
With the Artboard tool the active artboard's name was drawn twice: by the canvas, as for every artboard, and by the tool, a little higher. The test draws a frame with two artboards and checks that each name is painted once and that the active artboard's sits where the other's does. It fails without the canvas skipping the tool-labelled artboard (two labels for the active one) and without the tool's label position (8 points to the right and 11 higher than the other name).
The Add and Delete Anchor Point tools' icons were drawn apart from the Pen tool's, so the three didn't read as one family. The test checks that both tools' icons contain every shape of the Pen tool's icon plus a badge of their own, and that plus and minus differ. With the tool mapping pointing back at the old dc-pen-add and dc-pen-delete icons it fails: the Add tool's icon lacks the Pen's outline.
A chain of geometry effects makes at most 2 million anchors of a path (of a
type, symbol or image art's paths together) and about 8 million lines for the
renderer to draw at 1 px/pt, or what the path costs when that is more.
Roughen spaces its points further apart, Zig Zag makes fewer ridges a
segment, Free Distort, Twist and the warps split segments into longer pieces
(map_nonlinear_within) and Transform makes fewer copies where they would pass
it; a result drawn with too many lines is built again with fewer points, and
an effect that still doesn't fit is left out.
Three Roughens of a rectangle in a 25 KB file needed gigabytes of geometry,
and 2 million anchors of them flattened into 141 million lines (7 GB, past
vello's 2^27 lines a path); the file now opens in 0.4 s at about 450 MB.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found while testing crates/doc/src/lib.rs:379 (number_unit, reached from parse_measure at :374)
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A follow-up to #3677. A data URL in CSS, `style="fill:url(data:…)"` or a paint's
`url(…)`, stayed in the text the prepass searches, since only a value beginning with `data:`
was left out (CodeRabbit's review of #3677). Each `url(` in an element's quoted value whose URL
begins with `data:` is now left out up to the byte that closes it, its quote or `)`, so what
follows it, `display:none` say, stays. A `url(` that does not close inside its value is kept
whole: where the end is unclear, less is left out, never more.
The scan now jumps over a data URL with one search for the byte that ends it, instead of
stepping through it a byte at a time. The markers' capacity is a saturating sum, so no range can
panic it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Round a closed path the editing data's text document writes the end of
the type's range unwrapped, past the last segment (the start plus the
run), or wrapped the other way. The importer clamped it to the path's
end, so the type kept only the stretch from its start bracket to the
path's end and overflowed. Both brackets are now wrapped onto the loop
(non-finite values fall back to the start); an open path is clamped as
before.
The object's own path that then replaces the frame is written open,
ending on its start: such a path is now closed where its ends meet, so
the span is measured round the loop.
Hand-written tests: an unwrapped, a reversed and a negative range on a
closed circle, a range inside it, an exact full turn that fits a short
line, an open path still clamped, and a path ending on its start closed.
Review follow-up (CodeRabbit): a linear gradient expands to a mesh with a
column at each stop, and the mesh mixes linearly between columns, so a
Perceptual blue -> yellow came out as the grey RGB mix in the middle.
`breaks` now also takes a column at each stop the Perceptual expansion adds
(Gradient::expanded), so the mesh follows the OKLab curve; Classic is
unchanged. the_mesh_option_makes_a_mesh_that_paints_the_gradient gains a
Perceptual case; without the change 1812 pixels differ from the gradient.
Review follow-up (CodeRabbit): the test's doc comment quoted a phrase from
the public help page. AGENTS.md allows feature names and a cited URL, not
copied wording; the comment now describes the behaviour itself and keeps
the URL.
Illustrator 30.0's Method choice in the Gradient panel
(https://helpx.adobe.com/illustrator/desktop/paint-and-fill/create-and-edit-gradients/control-perceptual-interpolation-in-gradients.html).
- paint.editGradient takes `method: classic|perceptual` (one undo step);
paint.setFill/setStroke take it in `gradient`, and gradient params (swatches,
copied paints) carry it when Perceptual. Anything else is refused without
editing. One edit can set `method` and `dither` together.
- The Gradient panel shows a Method menu (Classic, Perceptual) for linear and
radial gradients, dimmed without a gradient; the Dither box stays under the
stops.
- i18n: "Perceptual" where a complete catalog lacked it, and the `method`
error in all 24 catalogs.
- Docs: G13 (estimate 14–24 h → 13–23 h, after Dither's 15–25 h → 14–24 h),
target-app-parity (row and evidence), ui-parity, mcp.md, plugins.md.
Illustrator 30.0's gradient Method, from the public documentation
(https://helpx.adobe.com/illustrator/desktop/paint-and-fill/create-and-edit-gradients/control-perceptual-interpolation-in-gradients.html):
Classic or Perceptual; perceptual interpolation matches colour transitions to
how the eye perceives differences between colours, and perceptual gradients
are rasterised when saved as PDF or SVG.
- Gradient gains `method` (Classic | Perceptual; serde default Classic,
written only when Perceptual, so older files open as Classic and Classic
files are written as before). The colour crate's OKLab helpers open up for it.
- Perceptual mixes neighbouring stop colours in OKLab (sample, and
sample_keep_model for the stroke meshes, Expand and blends); opacity stays
linear. Renderers that mix linearly get stops on the perceptual curve,
placed by halving each half segment where a straight line strays more than
0.4 of a colour level (a nearly straight segment needs none). Classic
gradients expand exactly as before.
- draws_as_pixels includes Perceptual gradients, so SVG and PDF write a
Perceptual fill or stroke as an image at the document's raster effects
resolution, clipped to what it paints. A gradient along or across a stroke
stays slices (PDF no longer draws it as a within-stroke image).
- With Dither: a dithered Perceptual fill is the perceptual mix with dither's
noise (the canvas's dithered grids are keyed by method too); SVG and PDF
write it as an image with the noise, and a dithered Perceptual stroke as an
image without it, as the canvas draws it (GradientPaint::pixels, SVG's
raster_paint and PDF's area are told whether they paint a fill).
Review follow-up (CodeRabbit): the refusal skipped any request carrying a
`swatch` key, so `"swatch": null` with a dithered gradient got through.
Only a named swatch (a gradient swatch keeps its own dither) is exempt now.
The dither test sends `"swatch": null`; it fails without the change.
The prepass decided what to look for in the tree by a dozen substring searches over the whole
SVG text (`:not(`, the event attributes, `script`, `transform:` and nine more), and a separate
pass counted the elements' depth before roxmltree. With an embedded picture the text is mostly
its data URL: a 1.35 MB card SVG (a 968 KB PNG) paid about 7 ms of searches and 1 ms of depth
scan per import, on a 64 ms render, and a 54-page PDF export 54 times that.
`depth::prescan` is the one pass: it counts the depth as `xml_too_deep` did (the DOCTYPE entity
count included) and, in the same walk, notes every `data:` attribute value, answering the text
with those values left out, their quotes kept. The prepass and `fx::blends` search that text;
the edits are still made in the source. A data URL can hold no tag, attribute name, selector or
CSS declaration, and everything else, a `style="display:none"` included, is kept, so what is
found is what was found before.
`vectorcraft-cli perf --deep` gains a tracked row, "import SVG, one embedded PNG", beside the
export row it had.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
With Copy Links on and Copy Fonts off, File › Package still read every
placed PDF or .ai page to find the fonts of its type, and then dropped
them. With Copy Links on and Relink off, it still read the preview and
the PDF of each placed .vectorcraft document, which only the rewritten
copy uses.
Package now reads the placed pages for their fonts only with Copy Fonts
on, and it reads the preview and the PDF only with Relink on. With Copy
Fonts off, packaging a document that places the 60 pages of a PDF took
about 390 ms in a release build and now takes about 20 ms. The package
it writes is the same.
With the Artboard tool, File › Revert kept the active artboard, but a
command.batch with a file.revert step made artboard 1 the active one.
A batch that switched to another document and back, and the Actions
panel menu's Batch… over a folder whose files it closes or over data
sets with a Revert step, did the same. Each of them starts the tool
afresh and ends in the document it started in, and the app moved the
view to the fresh tool's artboard 1. Only a plain file.revert was
handled, by its command id.
The engine now counts the times switching or closing documents, or
replacing the active one, starts the tool afresh (Session::tool_resets),
and records whether the tool's options were set after that
(Session::tool_set_since_reset). When the count changes during a
command and nothing set the tool's options after it, the app starts
the tool on the active artboard, as it does when a command chooses
another document. When a later step chose the tool's artboards (a
Paste or Cut after the revert), the artboard the tool then has active
becomes the active one.
Docs: docs/mcp.md, docs/target-app-parity.md.
Angelo Bernardini