When a .ai's layers can't be read from its editing data yet, it opens as
its PDF part, which lists the artboards as pages; those were laid out in
a row, so a grid of artboards (or any other arrangement) was lost. The
editing data's artboards are now read even then (ai::artboards, as far
as the reader gets) and, when there are as many as pages and of the same
sizes, each page goes where its artboard is, with the art on it (the art
of a page: the objects mostly on it, else nearest it). Otherwise the
pages stay in a row as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Align and Distribute Spacing move a selected compound path or group as
one object, and they found the key object only when its own id was
among the objects they move. When the key was inside a compound path
or a selected group, aligning to the key moved that compound path or
group like any other object, and the key moved with it. In that case,
Distribute Spacing with a spacing held the first object still instead
of the key.
The key object can end up inside a selected group or a compound path.
Dragging its row into a group in the Layers panel (layer.move) keeps
the selection and the key. A Selection-tool Shift-click on that group,
when it is unselected and directly in a layer, then adds it to the
selection (select.toggle) and keeps the key. Other ways of adding the
group keep the key as well: a Shift-click in the Layers panel's
selection column or on the group's target circle, a Group Selection
click that hits the key itself, select.add, and select.anchors or
select.anchorsMany in add or toggle mode. In isolation mode, a
Selection-tool click on one of several selected objects in the
isolated group makes that object the key. In an isolated compound
path, the same click on the outline of a selected member makes that
member the key.
Align to the key object and Distribute Spacing with a spacing now keep
in place the selected object that is the key or contains it. Both
commands look for that object among the key and its ancestors, so a
group around the key's group also stays in place. Aligning to the key
still lines the others up with the key's own bounds.
Docs: docs/target-app-parity.md.
arc raised its second angle by 2π in a loop until it was no longer
below the first, and arcn lowered it the same way. For an angle of
1e18 degrees that loop takes about 3e15 steps: vectorcraft-cli info
was still reading an EPS file with "newpath 100 100 50 1e18 0 arc
stroke" when "timeout 20" stopped it. For 1e308 the loop never ends:
from 2^56 radians (about 4.1e18 degrees) up, adding 2π leaves a double
unchanged. The import fuzz test eps_hostile_programs_never_panic builds
programs from tokens that include arc, arcn, 1e308 and 1e15.
arc now raises the second angle by multiples of 360 until it is at
least the first, and arcn lowers it until it is at most the first, as
the PostScript Language Reference defines them. The code works out both
sweeps from the remainders of the two angles after whole turns, and it
places the first endpoint at the remainder of the first angle. Ordinary
angles give the same arcs as before, apart from rounding in the last
digits. One case changes: when raising (arc) or lowering (arcn) the
second angle by whole turns makes it equal to the first, as in
60 -300 arc or -60 300 arcn, rounding sometimes made the old loop add
another turn and draw a full circle. Those arcs are now empty, as the
PostScript Language Reference defines them.
Tests: 1e18 0 arc, 1e308 0 arc, 0 1e18 arcn and 0 1e308 arcn draw the
arc between the angles' remainders and leave the current point at the
second angle. The test reads each program on a thread of its own and
fails when no result arrives within 20 seconds.
The sweeps `every_command_with_junk_params` and
`fuzzed_calls_keep_documents_serializable` skipped every case that set
`object.scale`'s `sx`, `object.transformEach`'s `scaleH` or `scaleV`,
`object.distributeSpacing`'s `spacing`, or `artboard.setProps`'s `x`,
`y`, `width` or `height`. The comment above `KNOWN_BUGS`, the list of
those (command, param) pairs, referred to `#[ignore]`d repro tests, but
`bug_huge_transform_makes_document_unloadable` and
`bug_artboard_huge_props_break_svg` have no `#[ignore]` and run with the
other tests. Since 42df20ca, the commit that added the list,
`Session::edit` returns an error when an edit leaves an artboard or a
selected object with coordinates that are not finite or exceed
`MAX_COORD`, and `object.transformEach` clamps `scaleH` and `scaleV`.
The list and its two checks are removed, so both sweeps run those
cases.
A malformed id gets the existing object-id message, and a listed anchor
that is not a corner the existing "give corners of the path" one, so
only the malformed-list message is new (translated in every catalog
that covers messages).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
parse_freeform and paint.freeform.addLine share point_indices, which also
keeps addLine from wrapping a huge index round on 32-bit targets (wasm).
The new messages are in the catalogs that translate every message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Text Wrap and the variable bindings share checked_ids_or_selection in
cmd/mod.rs instead of each keeping its own copy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The drawer width is worked out from the tool cell, and the new test no
longer sits between the existing test and its doc comment.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
crates/text/build.rs (craft_fonts.rs) and crates/ui-egui/build.rs (credits.rs) wrote their
output on every run. Where cargo's recorded times lose their sub-second part (an ecryptfs home
directory, as Mint and Ubuntu offer at install), a file written just before its crate compiles
reads as newer than that compile, so the next build compiled the crate and everything above it
a second time: a full app rebuild for nothing after any build-script rerun. Unchanged output is
now left alone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A .ai or PDF whose spot colours are defined in Lab opened them wrong:
- the PDF part gives such an ink an ICC Lab profile as its alternate
space ([/ICCBased profile], 3 components), which was read as RGB, so
the Lab values were clamped into wildly wrong colours (PANTONE 249 C
came in yellow). A profile whose Alternate is [/Lab ...], or whose
header names the Lab data colour space, is now read as Lab;
- the editing data writes them as `c m y k L a b (name) tint 2 Xx`,
type 2 being Lab, and only the CMYK equivalent was read. Type 2 now
gives the Lab colour, in the swatches and on the art.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
These Object commands edited the document even when nothing changed, so
the call recorded an undo step, emptied Redo and marked the document
modified:
- object.align, object.distributeSpacing and object.distribute (three
or more objects) when no object moves. With one object selected,
Object › Align aligns it to its own bounds, so it recorded a step
every time. Calls on objects that were already aligned or evenly
spaced also recorded a step, including objects that an earlier Align
or Distribute left a rounding error away from the position the next
call computes.
- Move, Rotate, Scale, Shear, Transform and Nudge by an identity
transform without a copy, Transform Again after one, Set Bounds with
the current position and size, and a Selection tool drag that ends
where it started.
- Nudge by 0 with anchors direct-selected or with only ruler guides
selected, and Transform with nothing selected or with ids that name
no object, with or without a copy.
- Bring to Front and Bring Forward on objects already in front, Send to
Back and Send Backward on objects already at the back, and Send to
Current Layer on objects already on top of the current layer.
- Ungroup, Release Compound Path and Release Clipping Mask with none of
their objects selected, and Expand Shape with no live shape selected.
- Lock with every selected object already locked, Hide with every one
already hidden, Unlock All with nothing locked and Show All with
nothing hidden.
- object.setProps with the values the objects already have, or with no
object.
- object.setLiveShape with the values the shape already has: a
polygon's sides, angle, radius or equal sides, or a side length that
gives exactly its current radius; an ellipse's pie, including Invert
Pie on a whole ellipse; and a corner radius or kind that the corners
already have.
These commands now return without editing the document, and Align and
Distribute count a move of 1e-9 pt or less on both axes as none. Such a
call records no undo step, keeps the step to redo and leaves the
document unmodified. In these cases Ungroup no longer resets the
selection, and Release Compound Path, Unlock All and Show All no longer
empty it. A Transform with a copy of no object keeps the selection.
Lock and Hide still deselect the objects. A Nudge by 0 no longer
expands a live shape whose anchors are direct-selected. The replies
have the same form as before. An identity transform still sets what
Transform Again repeats. The journal still records each call, so the
Actions panel records it as before.
When every value given to Set Bounds is the current one, it now applies
the identity transform instead of the transform computed from those
values, which can differ from the identity by a rounding error on a
rotated object or with Use Preview Bounds.
When a drag comes back to its start, the identity preview no longer
edits the document. Session::preview now advances the revision when it
undoes the previous preview, so views redraw that drag.
The undo and redo row of docs/mcp.md now notes that many calls that
change nothing record no undo step, so undo after one undoes the step
before it.
A character attribute (Character panel menu with the East Asian options;
`text.setFormat` / `text.setRangeStyle {proportionalMetrics}`), saved only
when on. In horizontal type it applies `palt`; vertical type sets as before
until the vertical shaping change (`vpal`) follows.
Line-end Punctuation Half Width takes nothing more off the punctuation
`palt` re-spaced (the flush opening bracket, consecutive punctuation, the
line-end half width): the segment is shaped once more without `palt`, and
the glyphs whose advance differs are marked. Punctuation the font leaves
full width is trimmed as before. SVG writes `"palt" 1` with the
character's other features.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After scaling 12 pt type 2x, the Character panel still showed 12 pt and
typing 10 pt gave 20 pt type. Character controls and text commands now
work in document points: the size, leading and baseline shift shown and
set are the stored ones times the text's scale along its em height, and
horizontal scale includes the ratio of the transformed em axes. Rotation,
reflection and shear stay in the transform, so text remains editable.
Groups' text counts, keyboard increments use the same units, Shrink
Text to Fit works in document tenths, and a collapsed transform rejects
dimensional edits without touching the document.
text.getRange takes effective: true for the document-space values.
Choosing a fill or stroke (Color panel, Swatches, the proxies) while the
Type tool had a few characters selected painted every run of the type
object. As in Illustrator, it now goes to the selected characters only:
the runs are split at the selection's ends and the paint is placed on
the runs in between, as one undo step; a stroke gets a 1 pt weight
where it had none. With only a caret, or with explicit ids or an item,
paint.setFill/setStroke behave as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A .ai in PostScript form went only to the PostScript interpreter, which
stopped on the first operator its missing prolog would define
(`Adobe_packedarray`, `Lb`, `Np`): files from Rhino and other CAD apps name
the procsets without including them.
- Such a program is its own editing data: when it has layers
(`%AI5_BeginLayer`) and no private data, the editing-data reader reads it,
compared with the page when the program runs, from the layers alone when
it can't be run. If the layers can't be read either, the error says why.
- The reader takes the legacy ten-operand `Lb`, ignores flatness `i`, and
uses the bounding box as the page when nothing else says where it is.
- Tests on synthetic files; fuzz fixture and hostile-program layers added.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
# docs/file-format-parity.md
# docs/gaps.md
The editing-data reader stopped on `Xk`/`XK` (global process colours) and
on `p`/`P` (pattern fills), so the file fell back to its PDF part, which has
only the top-level layers.
- `Xk`/`XK` read like `Xx`: a global swatch (not a spot ink), linked at its tint.
- `%AI3_BeginPattern` sections are read by a reader of their own into
pattern swatches (tile from `(name) llx lly urx ury`, art in pattern space);
`p`/`P` fill or stroke with them, placed by the operator's matrix. A pattern
it can't read whole still makes a fill with it unreadable, as before.
- Tests on synthetic editing data; fuzz tokens and fixture extended.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Guides on a hidden or locked layer leave the selection (Selection::prune,
run after every edit and selection change), their Layers rows neither
select nor drag them, and guide.move / guide.remove / guide.setLayer
refuse them; guide.setLayer refuses a locked target layer, as dragged
rows are refused there.
- A new guide goes where new art goes (the current layer, else the top
shown and unlocked one; refused when there is none), so it never lands
on a hidden layer and vanishes.
- Layer commands keep guides with their layers: Duplicate (and Alt-drag
copies of rows) copy a layer's guides onto the copy, Merge Selected and
Flatten Artwork put the merged layers' guides on the merged layer and
Flatten discards a hidden layer's. One shared doc predicate for "the
layer a guide is on"; the deleted-layer cleanup moves into the doc.
- Tests: engine tests for the above and their undo, a doc check that an
old guide reads with no layer and none is written, and a headless
Layers panel test of guide rows (select, drag to another layer, locked).
- Translations for the new message; docs and a ROADMAP progress line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Guide rows in the Layers panel drag (the selected guides when the row is
one of them) onto any row of another layer, which lights up, and go to
that layer in one undo step (guide.setLayer takes indexes). Guides on no
layer, from files saved before guides had layers, go on the top visible,
unlocked layer when a document opens or is reverted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
A ruler guide goes on the current layer (guide.add {layer}), shows,
hides, locks and is deleted with its layer, is listed in the Layers panel
as a row that selects it, and moves to another layer (guide.setLayer).
Guides on a hidden layer neither show nor snap; those on a locked layer
can't be picked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
A dock column's groups scroll under its top strip, each open panel at
least 160 points tall, instead of running off the bottom of the window.
Drop zones in a column only use the part of a group in view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
A group's upper half inserts the dropped groups right above it and its
lower half right below it, in floating sets and dock columns alike, and a
dock column's room under its groups puts them at its bottom; the edge
bands were too thin to hit, and the one above a dock column's top group
lay outside the column. The tabs a drop stacks panels with now light up
over the panels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
A floating set dropped on the left edge of the icon column or of a dock
column docks as a new dock column there, building a multi-column dock.
Its groups stack, take tabs and tear out as in a floating set; the strip
along its top floats the column again and its × docks its panels where
they live. Popped-out icon panels open left of the dock columns. Columns
are kept in preferences and workspaces; window.panel.dock takes column.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
Floating panel groups dropped on another group's top or bottom edge join it
in a set: the groups stack top to bottom, move together by the top group's
title bar (whose × docks them all), and a group dragged by its tab strip
leaves the set. A double-click on a tab collapses a group to its tabs.
Sets and collapsed groups are kept in preferences and workspaces, and
window.panel.float takes below, above and collapsed.
Behaviour from public documentation:
https://helpx.adobe.com/illustrator/using/workspace-basics.html
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
The sweep under Offset Path (linesweeper 0.4) never finished on some
miter offsets: on the reported glyph and on a synthetic outline it kept
allocating until the app ran out of memory. linesweeper 0.5 fixes that
infinite loop ("Fixed an infinite loop", 0.5.0 changelog), and builds
without changes.
0.5 can hand back a curve piece that doesn't follow the input curve it
was cut from. In the Shape Builder rosette test, a contour runs out along
one petal and back by such a piece, which became a 59 x 4 pt hole in the
merged shape. Tidy::repair rebuilds a piece that strays from the one input
curve through both its ends as that curve's exact sub-curve (input curves
are indexed by their bounds' left edge, so it stays cheap). The there-and-
back then encloses nothing and is dropped as a sliver. Pieces on their
curve, and pieces whose ends two input curves pass through (a real lens),
are left alone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Transform panel has both options; the Properties panel's Transform
section had no way to them. A More Options button after the flip buttons
now opens a popup with the two (the scaleCorners / scaleStrokes prefs),
open while they are toggled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With art selected the picker changed it only on OK. Each new colour now
previews paint.setFill / paint.setStroke on the selection through the
session interaction (the spec is preview: true): OK keeps the last one as
one undo step, Cancel puts the art back. Nothing previews until the colour
first changes, nor without a selection (the colour then only becomes the
default paint, on OK).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lucide names its flip icons for their mirror line: flip-horizontal-2
draws a horizontal line with triangles above and below (the top-bottom
flip). The Properties and Transform panels showed it on the left-right
flip, and the other on the top-bottom one. Swapped back, with a comment.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Beside every corner of the bounding box the curved rotate cursor bowed
up. It now bows from the box's centre toward the pointer, in 45° steps
(Cursor::RotateToward, made by Cursor::rotate_about): at the top right
it opens toward the bottom left, at the bottom right toward the top left.
The Selection tool shows it over the rotate zone and while rotating (it
follows the pointer round), and Free Transform does the same.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fill_contains promised to close open subpaths (as their fill does) but
counted kurbo's winding, which leaves the closing edge out: a point beside
it read as outside. An open path (an anchor deleted from a closed one) lost
its fill clicks on that side, and a reflection moved the edge across, so it
looked like Reflect broke selection. Open subpaths are now closed by a
straight line back to their start before the winding is counted; closed
paths are borrowed as they are.
Tests: the open path from the report, on both sides and even-odd; reflected
shapes, imported SVG hearts, a selected shape dragged by its fill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Shift-click on a guide adds it to the selected art (guide.select
toggle keeps the art), and Shift-adding or -removing art keeps the
selected guides; a new selection still clears them. With both selected,
Delete deletes the art and the guides in one undo step (not when ids are
given).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
A PDF that names a font without embedding it (here a standard Helvetica)
is drawn with a stand-in, so comparing its glyphs with the installed
family's styles picks whichever style happens to look most like the
stand-in: with a large Helvetica family installed, Helvetica became
Helvetica CE Medium (tests_import_fidelity::text_becomes_point_type
failed on such a Mac). Fonts without a font program now keep the face
their name resolves to; embedded fonts are still compared as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1
The link watch wakes the app every two seconds only while the last look
found links (068a934, #967). A file placed less than two seconds after a
look that found none asked for no more frames, so while the window sat
idle nothing stamped it; a change made meanwhile was taken for the file
as first seen and never acted on. The watch now remembers the active
document's uid and revision at each look and, when they differ on a
later frame, asks for one wake when the next look is due. There is
still no per-frame walk of the links, and no two-second wakes for a
document without links.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
$INSUNITS 22, 23 and 24 (US survey inch, yard and mile, as Autodesk's INSUNITS reference
lists them) had no size or name, so the drawing opened as unitless (millimetres, or inches
with $MEASUREMENT 0) and document.dxfInfo said Unitless.
They now derive from the US survey foot (1200/3937 m) like code 21, with their names.
The test reads each survey code's unit name and default scale and opens a survey-inch line at
7200.0144 pt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
A root such as width="96pt" height="96" took CSS pixels (0.75 pt) along x and points along y,
so a square came in 24 x 32 pt and the artboard 96 x 96 instead of 96 x 72; with a viewBox the
art was stretched too.
When either root side is absolute, both axes now take CSS pixels at 96 per inch (usvg already
parses at that resolution): the unitless or px side is the same CSS length it would be in a
browser. The root scale is one number.
The test opens width="96pt" with height 96, 96px and 72pt, and width 128 with height 72pt: a
96 x 72 artboard and a 24 x 24 square; with viewBox 0 0 32 32 a 72 x 72 square at x 12 (meet),
and with preserveAspectRatio="none" the whole artboard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
Cursor::precise listed every Pen state but PenJoin (added later, #794), so with Use Precise
Cursors on the Pen still showed its join badge over another open path's end.
PenJoin is a crosshair too under the preference.
The Pen join test now also checks the pointer over the other path's end is the crosshair with
the preference on (and PenJoin with it off, as before).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
shapes::polar_grid drew at least one radial divider (`radial.max(1)`), so radial 0, which
the command and the Polar Grid tool (Left arrow while dragging) accept, still drew a spoke.
Zero now draws none; the loop never divides by it.
The test checks 0, 1 and 2 radial dividers with no concentric ones give 1, 2 and 3 paths.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
Reversal read the midpoints after Gradient::reverse had reordered the stops, so each stop
took the wrong segment's midpoint: red-to-blue with its midpoint at 20% reversed to a 50%
midpoint instead of 80%, moving the transition.
Gradient::reverse now mirrors each segment's midpoint with it, so a reversed gradient is the
same ramp the other way round; paint.editGradient's own (wrong) pass is gone, and the SVG
importer's reflect spread, which had a correct copy of the same mirroring, uses it too.
Tests: two- and three-stop gradients with off-centre midpoints sample the same reversed (t vs
1 - t) and come back after two reversals; paint.editGradient reverse turns a 0.2 midpoint into
0.8.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
The menu item sends no artboard, so the engine fell back to the first one even with another
active in the Artboards panel.
The frontend fills in the active artboard (the view's, as the Artboards panel and navigator
show it) for select.allOnArtboard, as it already did for Paste in Place / in Front / in Back
(#693); the two now share one block, and the index is kept within the artboards. An explicit
artboard still wins; the engine is unchanged.
The test makes artboard 2 active, then 1, and invokes the menu item each time, then passes
artboard 1 explicitly with artboard 1 active.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
The desktop app's ui.render (MCP `screenshot` when connected) always rendered the first
artboard and ignored `artboard`, which the headless backend honours.
It now renders the 0-based `artboard` (the first without one, an error for one that doesn't
exist), as headless does. control-protocol.md and the control.rs list show the parameter.
The test drives ui.render through the control channel on a 64 x 64 and a 96 x 48 artboard:
no index and 0 give 64 x 64, 1 gives 96 x 48 (48 x 24 at scale 0.5), 2 is an error.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
add_text passed `color` to text.create only, and text.createInPath built its own character
style without a colour, so type on a path and path-based area type took the current fill.
text.createInPath now uses the same new-type style as text.create (size, font, style, color,
else the current fill), and add_text sends `color` on both routes.
The MCP test draws a blue rectangle (the current fill), turns it into on-path and area type
with color #ff0000, and reads red back through inspect_document.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
# ROADMAP.md
