`enveloped_type_exports_distorted_in_every_format` exports each format
twice and compares the two files without the lines that contain
`CreationDate` or `ModDate`. Two other lines of the PDF also change
with the time it was written. The XMP packet is one line that holds
`xmp:CreateDate`, `xmp:ModifyDate` and the file ID, and the trailer
holds the file ID as `/ID`. The file ID is a hash that krilla computes
over the bytes written before it, which include the dates. When the
two PDF exports were written in different seconds, these two lines
differed, and the test failed at "pdf exports are repeatable".
The comparison now also drops the lines that contain `xmp:CreateDate`,
`xmp:ModifyDate` or `/ID[`. A new check writes the PDF that
`document.export` writes, with its creation time set to Unix seconds 0
and then 1, and asserts that the comparison's filter leaves the two
files equal.
The engine's guard turns a panic inside a command into
EngineError::Internal and rolls the document back. The panic checks in
command_sweep.rs counted that error like any other error. They include
probe, which the empty-params, junk-params and structured-junk sweeps
call. The tests in import_fuzz.rs that open, place or paste a file
through a command counted it the same way. None of them reported the
panic.
Each of these checks now passes the command's result through no_panic,
which the library file tests in import_fuzz.rs already use, and the
change adds a copy of it to command_sweep.rs. A case whose command
panics now fails. The panic from no_panic reads "<command> panicked:
<message>", and the checks that print a panic's message show it. Each
check handles any other error as before.
The command palette listed select.key as Set Key Object. A pick runs
the command without an id, which cancels the key object (#541).
select.key is now labeled Cancel Key Object, the label of the Align
panel menu item that also cancels the key object. The palette lists
it once, and engine.commands reports the new label. The params text
and docs/mcp.md say that select.key {id} makes id the key object, and
that select.key {}, or an id that names no selected object, cancels
it. The Set Key Object rows are removed from the 10 catalogs that
had one.
Docs: the select.key params text, docs/mcp.md,
docs/target-app-parity.md.
# Conflicts:
# crates/engine/src/cmd/select.rs
# docs/mcp.md
# docs/target-app-parity.md
Agents could set and clear the key object but not read it:
document.inspect, the control channel's ui.pointer reply and the
headless pointer_gesture reply left it out, and select.key returned
null. select.key with an id that was not selected, or with a value
that is not an id, such as "x", cleared the key without an error. A
compound-path member or an object inside a selected group could become
the key.
select.key now checks its id as select.set, select.add and
select.toggle do (#784). It returns an error for an id that is not a
selected top-level object and leaves the key as it was. Layers,
compound-path members and objects inside another selected object are
not top-level objects. select.key {id: null} used to clear the key and
is now an error; select.key {} still clears it. select.key returns
{key}, and document.inspect and both pointer replies include the key
object as "key".
In isolation mode a Selection-tool click can pick an object that
select.key does not accept as the key: a sublayer inside an isolated
layer, a member inside an isolated compound path, or an object in an
isolated group that is also selected. The tool no longer sends
select.key {id} for such an object. When the object is already the
key, the tool still sends select.key {}, which clears the key.
A recorded action whose select.key step names an object that is not a
selected top-level object when the action plays now fails and is
rolled back, as an action with a select.set step naming a deleted
object has failed since #784.
Docs: docs/mcp.md, docs/control-protocol.md, docs/target-app-parity.md.
With two or more objects selected and no key object, a click on Align
to Key Object in the Align panel stored Key as the Align To choice and
set no key. The align buttons of the Align panel, the Control bar and
the Properties panel then ran object.align with to: "key", which fails
with "no key object". Distribute Spacing sent the panel's spacing
value and kept the leftmost or top object in place. The stored Key
stayed until Align to Selection, Align to Artboard or Cancel Key
Object was chosen.
The button now runs select.key on the topmost selected object, the
last in paint order, whatever order the objects were selected in, and
the canvas draws that object with the key-object outline. The align
buttons in all three places and Distribute Spacing then use that key,
and Cancel Key Object clears it. A key that is already set stays the
key. With fewer than two objects selected the button is disabled, as
the Distribute buttons are. The button stores no choice, so once the
key is cleared by Cancel Key Object, by a click on the key object or
by another selection, Align To shows the stored Selection or Artboard
choice again. Cancel Key Object no longer sets that choice to
Selection.
Docs: docs/target-app-parity.md.
Since #1035 moved linesweeper to 0.5, the sweep can cut the paths
through one crossing at points more than its tolerance (eps) apart.
The planar map merged a point only into a vertex within eps of it, so
the two passes of an open path could end on two different vertices,
and it kept them as two edges with different ends. Past the crossing,
shape_builder then returned a line's loose end as one piece that runs
out to the line's end and back, and an extra piece of the line it
crosses, about 1.4e-6 long. In live_paint, the face around the loose
end ran out along it and back, and FaceMerger::merge of the faces
returned a second subpath along it. pieces_cut_out_of_their_inputs
fails on main for such an input.
The map now joins the ends of pieces whose middles and ends lie within
16 eps of each other. The two passes then share their ends and become
one edge. The new test is the failing input proptest saved for the
property. Lines that slope close to -1/32 in the document are nearly
level in the sweep's frame, and the change does not cover all of them.
artboard.move with moveArt (the Artboard tool's drag of one artboard or
several and, with copy, its Alt-drag), artboard.duplicate with art and
artboard.rearrange with moveArtwork moved the art, or its copies,
without their pattern tiles, whatever General › Transform Pattern Tiles
was set to. They now follow the preference as the transforms and
artboard.setProps do: with it on, pattern fills and strokes move with
the art. Each also takes a `patterns` param for one call. When the art
uses a pattern, the journal records the choice, and a recorded action
plays back with it.
Docs: docs/mcp.md (transformPatternTiles, artboard copies),
docs/target-app-parity.md.
# Conflicts:
# crates/engine/src/cmd/menucmds.rs
These commands edited the document even when nothing changed, so the
call recorded an undo step, emptied Redo and marked the document
modified:
- All Artwork Above and Other Layers (Object › Lock and
Object › Hide) with nothing to lock or hide: no object above the
selection, no other layer, or every one already locked (Lock) or
hidden (Hide).
- Transform Each with no object to transform, and without Copy when
both scales are 100% and there is no move, rotation or reflection.
- Rearrange All Artboards when no artboard moves, or with no artboard.
- stroke.widthPoint.set and stroke.widthPoint.copy when the profile
comes out as it was, and stroke.widthProfile.set with the profile the
strokes have already.
- Liquify when the stroke reshapes no path, and Puppet Warp given the
pins the art holds already (rest: true), or without rest when every
pin stays where it is and turns the art by 0° or is free.
- Define Perspective Grid with the grid the document has, Attach to
Active Plane with no object, Release with Perspective on art that is
not in perspective, Move Plane to where the plane is, without copies,
and Move Plane to Match Object when the plane is there and active
already.
- Move, Nudge and Transform in Perspective without Copy when no object
moves (1e-9 pt or less along the plane or its normal, Snap to Grid
included, or the identity), and Transform Again after one.
These commands now return without editing the document. The replies
have the same form as before. Transform Each with Copy and no object
to transform no longer empties the selection. Puppet Warp without rest,
with every pin where it is and turning the art by 0° or free, no longer
turns a live shape into a plain path. All Artwork Above still deselects
selected objects that are hidden or locked. Width Point counts a change
of 1e-9 or less in a point's place or width factor (its width over half
the stroke weight) as none. OK in Width Point Edit with the widths it
shows, with or without Adjust Adjoining Width Points, records no undo
step. After a move in perspective that moved nothing, Transform Again
repeats the identity.
One line in the evidence list of docs/target-app-parity.md names the
commands.
Every edit of an object copied its whole appearance (its fills, strokes and effects, about
540 bytes for a basic object) into the new node, though a move leaves it as it is. Over 45
whole-selection moves of 10,000 paths, undo history grew by 589 MB; now 349 MB (41 % less), and
a nudge or a drag of the selection takes 12 ms instead of 15.
- Node::appearance is a vectorcraft_doc::Shared<Appearance>: copy-on-write. A clone shares it,
and the first change through DerefMut copies it (Arc::make_mut). Reads, writes, equality (also
against a plain Appearance) and the saved form are those of the Appearance itself, so call
sites read and change it as before. Sites that build or assign a whole appearance add .into(),
and the few that need a plain copy take (*n.appearance).clone().
- Node::transform changes the appearance only when it has gradients to transform, and pattern
paints are transformed only when there are some: a move no longer copies an appearance it
doesn't change.
The workspace suite (5,301 tests) and clippy pass with the change; the doc and engine suites
pass again after the transform guard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
At 50,000 paths unless noted, against main:
- Pathfinder Outline (300 paths): 145 → 7.5 ms. Each segment was tested against every junction
of the arrangement, and each chain against every chain kept so far. Now a segment tests only
the junctions inside its control box (junctions are sorted by x; a point within reach of a
curve is within reach of that box), with the same nearest-point test. Coincident chains are
found through a grid on their midpoints, taking the first match in the same order a scan does.
- export SVG: 267 → 176 ms. Path data is written straight into one string (`push_num`, the
same digits and trimming as `fmt_num`) instead of a string per number, point and segment.
- open .vectorcraft: 239 → 195 ms. The JSON was parsed once only to find where it ends:
- A file with no NUL has no data after it.
- Otherwise the marker (which holds the first NUL, as JSON can't) is taken when it directly
follows the JSON's last byte, as the writer puts it.
- Anything else takes the old parse.
`perf --deep` prints output hashes of an SVG export and a save → open → save round trip; main
and this branch print the same ones. pathfinder_pinned (Outline included) passes, and so do the
pathops, format and svg suites (345 tests).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
Trim, Merge and Crop walk the arrangement once per input shape, and every walk tested each edge
side with a coverage mask built from a winding vector with one slot per input. On 300 paths,
against main:
- Trim: 362 → 31 ms
- Merge: 366 → 31 ms
- Crop: 303 → 28 ms
- Unite, Intersect, Exclude, Minus Front and Minus Back: unchanged or a little faster
- Multi, the arrangement's winding, keeps only the non-zero windings as sorted (input, winding)
pairs, added in place. A face is covered by a few inputs, so a test reads a few entries
instead of one per input. The winding arithmetic is unchanged.
- Arrangement::covered lists the inputs that cover a face (the true entries of its coverage
mask). Trim, Merge, Crop, Minus Front, Minus Back, Intersect and Exclude test that list; each
test's old mask form is in a comment beside it. mask() still builds the same mask for the rest.
pathfinder_pinned hashes the output of all ten operations on 2, 5, 12 and 30 overlapping
ellipses, rectangles and stars (even-odd and non-zero, shared keys for Merge). The 40 hashes were
computed with main's dense windings, and the new code gives every one of them. A wrong Crop test
fails it. The pathops suite (103 tests) passes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
The previous commit rebuilt the cache as soon as a cached path went stale, and built it as one
Vec path per node, hashed with SipHash. An edit that removes many objects one by one
(Pathfinder Unite of 1,000 paths) went stale on every removal, and Unite took 105 ms instead of
42 ms.
- The cache is now one flat table of (parent entry, index among siblings) with a one-multiply
id hasher, so a rebuild is about one walk and allocates nothing per node.
- A stale path walks, as an id the cache doesn't have does, and rebuilds only after a few.
Either way a lookup stays within a small factor of a walk.
At 1,000 paths, Unite takes 45 ms against main's 42. At 100,000 paths the gains of the cache hold:
hit test 1.7 → 0.36 ms, handle drag 6.5 → 2.7 ms per move. The cached-lookup tests pass unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
A lookup by id walked the tree, so every edit, click and pointer move of a single object cost
time in proportion to the whole document. At 100,000 paths:
- hit test, per click: 1.85 → 0.35 ms
- handle drag, 5,000-anchor path, per move: 6.5 → 2.6 ms
- segment drag, per move: 5.8 → 2.7 ms
Document keeps every node's index path in a cache (id_index, never saved). The tree is edited
through public fields in many places, so the cache isn't kept in step with it; it is checked
instead. A cached path is used only while the node at that path still has the id, a check as deep
as the tree. A path that doesn't check out rebuilds the cache, and an id it doesn't have walks,
as before. A rebuild costs a few walks, so it happens only after a few such misses, which keeps
any edit pattern within a small factor of the walk. Ids that appear more than once are never
cached, so they keep their first place in tree order. index_paths uses the cache for small
batches and one walk for large ones.
The one case that answers differently from a walk: a second node with an existing id put in the
tree after the cache was built, earlier in tree order. The lookup then returns the original node,
also one with that id. The app doesn't make such nodes; damaged files with duplicates are cached
afresh when opened.
cached_lookups_follow_any_edit_of_the_tree checks index_path, node and index_paths against a
walk after inserting at the front, removing, swapping, moving between layers, deleting a layer,
and in a clone sharing the cache. Using cached paths without the check fails it. The doc and
engine suites (2,002 tests) pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
The automatic rasterizer thread count was one per core up to 4. On a 10-core Apple M2 Pro, 8
workers are faster (9 are slower again), and vello_cpu draws the same pixels with any number:
- fit page, 50,000 paths: 66.7 → 56.6 ms; after a pan 66.4 → 51.2 ms
- 400 % zoom, 50,000 paths: 17.8 → 15.7 ms (back under its 16 ms budget)
- fit page, 100,000 paths: 137 → 105 ms; 400 % zoom 33.0 → 25.8 ms
Machines with 5 cores or fewer keep their count, and Preferences → Performance and
VECTORCRAFT_RENDER_THREADS still override it. pixels_do_not_depend_on_the_thread_count renders
a page of live effects and 400 overlapping translucent objects with 0, 1, 4 and 8 workers and
gets the same pixels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
A node with live geometry effects (Offset Path, Outline Stroke, Warp, Transform, Roughen…) had
them computed again on every frame: Offset Path's curve offsetting and boolean sweep for every
object, every time the canvas redrew. Rendering 2,000 paths that each have Offset Path, warm:
175 → 11.8 ms.
The renderer keeps each node's effected geometry, and each appearance item's own geometry
effects on it, keyed by the node's Arc identity like its other per-node caches. The entry keeps
the Arc alive, so a hit is exact and an edited object (a new node) gets new geometry. Only the
node draw_arc is drawing uses the cache: its path comes from the node alone, so its effects give
the same result every frame. Entries not drawn for three frames are swept like the rest.
cached_effect_geometry_draws_the_same_pixels renders a page of Offset Path, Roughen + Drop
Shadow, Warp + Outline Stroke and Transform objects (with item-level Offset Path) cold, warm,
after a pan and with a fresh renderer, and gets the same pixels each time; after an edit, the
warm renderer draws what a fresh one does. Keying the cache by node id instead of identity fails
it. The render suite (216 tests, golden images included) passes unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
Every lookup by id walked the whole tree, and the edit path did one per selected id, so editing
a selection of n objects cost O(n²). Measured with the new `perf --deep` rows at 25,000 paths:
- Select All: 458 → 3.5 ms
- nudge the selection: 3,517 → 34 ms
- drag it with the Selection tool, per pointer move: 3,413 → 35 ms
- undo / redo: 407 → 3.1 ms
- vectorcraft_doc: Document::index_paths (every id's index path, in one walk; the first in tree
order for a duplicated id, as index_path), node_at, node_mut_at (copy-on-write, as node_mut)
and ancestry_at.
- apply_transform, roots_of, transform_patterns, doc_sane, Selection::prune and
Document::paint_order look their ids up with one walk instead of one each.
- Selection::set dedups with a hash set instead of Vec::contains per id.
- vectorcraft-cli perf --deep: tracked rows for editing a whole selection (select, nudge, drag,
undo, redo), live effects, every Pathfinder operation, typing in a long story, and memory.
tests_lookups checks the batched lookups against the one-id ones on nested trees with missing and
duplicated ids; storing the last duplicate instead of the first fails it. The doc and engine
suites (2,000 tests) pass unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
On a hybrid laptop whose Wayland compositor runs on the discrete GPU
(Hyprland with AQ_DRM_DEVICES naming the NVIDIA card first), the built-in
panel still hangs off the integrated GPU, so the display rule put it first,
and the window stayed empty: the compositor took the frames and showed
nothing, with no error, so neither the restart on a failed adapter (#502)
nor the first-frame watch (#964) noticed.
Which GPU drives a display stands in for which GPU the compositor draws
with. Under Wayland the compositor says so itself: the main_device of its
linux-dmabuf feedback (version 4), the device it imports client buffers on,
which Mesa's Vulkan device-select layer reads for the same purpose.
compositor_gpu.rs asks for it on a short-lived connection, maps the dev_t
to PCI ids through /sys/dev/char, and under Automatic adapter_order ranks
that adapter ahead of the display GPUs. No answer (X11, an older compositor,
a GPU no adapter matches) leaves the existing rules in charge; a user's own
choice, WGPU_ADAPTER_NAME and the restart skip list still come first.
wayland-client and wayland-protocols are already in the tree through winit;
they become direct Linux-only dependencies.
Adds the Character panel menu's Digits option (with the Middle Eastern /
Indic options, beside Paragraph Direction in the Paragraph panel) as a
character attribute: CharStyle::digits, saved only when set, with
text.setFormat {digits} and text.setRangeStyle {id, start, end, digits}
("default" | "arabic" | "hindi" | "farsi"; a bad value is an error).
Shaping draws every digit of the Western, Arabic-Indic (U+0660–U+0669)
and Eastern Arabic-Indic (U+06F0–U+06F9) sets in the chosen system,
whatever was typed, before font fallback, so a font without the
Arabic-Indic digits gets a fallback for them. The text, its clusters,
the caret, selection and search stay as typed, as with All Caps; SVG
export writes the digits as drawn.
Catalog rows for Digits, Arabic, Hindi and Farsi in the ten translated
languages. Docs: docs/mcp.md (Digits), type-parity.md (digit types
done; Character Direction, kashidas, composers and the split caret
still missing), target-app-parity.md and gaps.md.
Behaviour from the public documentation (a character-level option with
the values Arabic, Hindi and Farsi in the Character panel menu):
https://helpx.adobe.com/indesign/desktop/language-and-proofing/arabic-and-hebrew/choose-digit-types.html
Tests: doc (the mapping and serialization), text (glyphs of "page 3" in
Hindi/Farsi equal the typed Arabic-Indic digits' on the typed clusters;
skips without a font covering the three sets), engine (setFormat,
setRangeStyle, undo, error), svg (digits written as drawn), ui-egui
(the menu shows with the Indic options).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The language switcher shows every language in its own name in every UI
language (language_and_appearance_are_in_the_app_menu), so "English" stays
"English" like in the other catalogs. "List Folder" is the file.list command
that arrived with #1181 and is a menu string the complete catalogs translate.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Own entry only, as AGENTS.md describes: credited by the name on my public
GitHub profile (sync_github_name), otherwise by @username.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Changes from an independent review pass over the Czech catalog. Short keys
shared between a menu and a dialog got the wording that fits both places
(All = "Vše", Add = "Přidat", Type: = "Typ:", Center = "Střed", Edge =
"Okraj", Advanced = "Rozšířené"); sentences that built a plural from a bare
count now read "Počet ...: {count}"; "Offset Path" is "Odsadit cestu", the
New... button is named consistently in the preset hints, "resolve" is not
translated as "rozlišit", and the rendering intent is "Záměr vykreslení".
Strings that need a context in the code to be right everywhere (Scale as a
dialog title, Random/Uniform as check boxes, Subset fonts below) are left as
they are.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Object › Expand, Expand Appearance and the rest of the catalog say
"Rozložit"; three older rows said "rozšířit" (widen) for the same
command: Expand Shaper Group and the two Revolve notes about expanding.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Czech is in COMPLETE_MESSAGES, so a new status or error message needs a
cs.tsv row too. czech_reads_as_czech checks the vocabulary (Nástroj
Pracovní plocha, Vzorník, Operace s cestami, Tah, Výplň), the three
plural forms of a layer count, the axis context and a nested message;
czech_plurals_have_three_forms now also runs every @plural row through
trn for one, few and other counts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Czech catalog covered every menu label (726 rows); panels, dialogs,
tooltips and status messages fell back to English. It now has a row for
every key of the complete catalogs: every menu string and tl! literal,
the 13 plurals in one/few/other form, the axis/effect/lighting/theme
contexts and every @msg status and error message (3,404 rows added,
4,130 in all). The existing rows are kept as they were; the header says
how the new ones were written and lists the vocabulary.
With the catalog complete, Czech sets complete_menus, so the coverage
tests now hold it to every menu string and tl! literal.
Written clean-room from the meaning of the English text, drafted with AI
assistance; not yet reviewed by a native speaker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
BFlat
dotCooCoo
Ikki Nishihara
arhammusheer
Atos Lins
zakariayacine
Losminek