* M8: a large selection's overlay and Design panel values are kept for frames where nothing changed (#1571)
The selection's outlines, box, size and slots on the canvas, and the Design panel's instance checks and position values, are kept with the document snapshot (held weakly), the selection and, for the overlay, the view's mapping to the screen and the overlay green, and made again when any changes. Dragging 648 selected layers: overlays 1.99 to 0.86 ms a frame, Design panel 2.54 to 1.40 ms.
* M8: split the position fields into one function per row (Sonar S3776)
The per-address download bucket never held more than its 16 MiB burst, so a larger library or plugin was refused for good. Bucket::take_spread serves one from a full bucket and leaves the debt to refill at the rate before the next download; the rate over time still holds.
The REST import's layer walk (also the .fig import's second stage) took about 4 KB of stack a level, a megabyte at the 256-level cap, and overflowed a 1 MB thread. It now runs on a work list in the same order, fitting groups once their children are in. The .fig mapping stays recursive at about 1.8 KB a level (455 KB at the cap). Both 256-deep imports now run on a 1 MB stack; the cap stays 256 everywhere.
* M8: one depth cap for a whole render; render threads get 8 MiB stacks (#1082)
A pattern fill drew its source as a fresh render from depth 0, so patterns in patterns could stack three renders of 256 levels. The source now goes on from the depth of the layer it fills. A render level takes about 6.5 KB of stack (measured), about 1.7 MB at the cap: the canvas render, preview and save threads and a band pool of their own now get 8 MiB stacks instead of the default 2 MiB.
* M8: SVG and PDF exports carry the depth into pattern sources and embedded renders (#1082)
Each area (document, images, library, versions, branches) keeps its own cap and now also draws from one total per file open, 6 GiB (2 GiB on 32-bit targets), so a file can no longer make the app unpack about 10 GB. Read in that order, the document and the images up to their own cap fit, so a file whose images pass 4 GiB (#1313) keeps as many as before; images left out by the total stay in the file like those past the image cap. What passes the total is skipped (the document refused) with a note.
variable.import and the variables parameter of file.open and file.import read an explicit path
with the uncapped reader, while the sibling variables.json next to a REST file already went
through read_capped with MAX_VARIABLES_FILE. variables_param now uses that capped read for paths
on native builds; the browser keeps its no-file-system error. Inline objects and JSON text, and
access roots, are unchanged. The check reads the size before the file, so it does not bound a
file that grows or is replaced in between. Primary file reads and image.place stay uncapped.
variable.import's help and docs/mcp.md name the limit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6 Astra <noreply@openai.com>
Every pointer move draws about three frames, and the canvas hit-tested the hover on each: a hover where nothing changed since the last (place, modifiers, zoom, file, page, revision, selection, tool and modes) is no longer sent to the tools again. Dragging an auto layout frame laid its contents out again on every move: a frame whose only change is its transform now keeps its own layout (its parent's still runs). ui.inspect reports frameTimes. Measured on a synthetic 1,900-layer file: hover UI pass 0.91 to 0.49 ms a frame and CPU 33% to 22% of a core; normalise while dragging 648 layers 3.8 to 2.1 ms a move.
AGENTS.md (and CLAUDE.md where present) gets a CI workflows section: workflows only call
craft-common-actions at @main, with no other jobs; feature and fix PRs don't touch CI and
link or open a craft-common-actions issue (a PR only for a small fix that is safe for
every app and does not increase CI time); legacy jobs
aren't added to and move only as their own task, staying until a shared workflow runs them.
windows-arm64.yml moves from its SHA pin to @main. The pinned revision is on
craft-common-actions main and the workflow (and the actions it calls) is identical
there, so nothing changes in behaviour.
Variable and collection nodes are left out of the layer tree (variables are read from the node list), so an instance's expansion no longer walks them again for every instance outside the layer cap. A nested instance takes only the override rows addressed below it, found by range, and shares its outer instances' property values and slot contents instead of copying them. Crafted files: 13.8 s -> 21 ms and 40.5 s -> 98 ms; normal files map as before.
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M9: comment.list lists only the threads that mention a name (#1415)
comment.list takes mentions (a name, any case): only threads where a message, the first or a reply, @mentions it, by the same parser the thread view highlights with and each message's mentions reports; combined with the other filters.
* M9: comment.people lists mentioned names; comment.next and comment.prompt carry mentions (#1415)
comment.people adds every name a message @mentions to those who wrote or reacted (once each in any case). comment.next takes the mentions filter and returns the thread as comment.list lists it; comment.prompt's bounded comments carry each thread's mentions (at most 32, names cut to 80 characters). One mention helper serves them all; builds on #1585.
* M9: comment.people lists people who wrote in two cases once
The participant list is deduplicated in any case too, not only the names added from mentions, in
one pass (CodeRabbit on #1591).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M8: grid auto-placement on an occupancy bitmap (#1083)
Grid placement tested every cell of a child's span at every position it tried. The taken cells are now a bitmap: spans are checked a 64-bit word at a time, the search jumps past the blocking cell, a row's search ORs the span's rows once, and a column-anchored child looks on from where the last one with the same column and spans found room. Placements are unchanged (checked against the old search on random grids); a crafted 1000 x 1000 grid takes 0.3 ms instead of 192 ms.
* M8: grid occupancy stores rows only once a cell in them is taken
An empty or sparse grid no longer allocates the whole bitmap up front (CodeRabbit on #1593).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A group or an unclipped frame applies its layer blur, drop shadows and
texture to everything its children paint, including children reaching past
its box. The renderer sized the container's layer from the union of its
children without growing it by the container's own effects, so the blur or
shadow was cut where a child's paint ended; exports of unclipped frames
lost part of the shadow the same way.
bounds_within now grows the children's union by the node's effects reach,
and render_bounds_at grows an unclipped frame's children by its effects.
The retained partial-render test drew this very cut (61 levels in its group
child scene); with the shadow counted the damage passes the partial share
of its 300 x 220 view, so the view is widened to keep it a partial render.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A double-click with the pen drew on: its second press added a point on the last one, and a double-click meant to edit a layer drew a stray vector on it. A press on the last anchor now adds nothing; a double-click ends the path (or, when it only started one, edits the layer under it) and Enter ends the path, both going on in vector edit mode with the Move tool. In edit mode the pen's Enter and double-click take the Move tool, the double-click taking back what its first press began. Escape still keeps the pen (#760).
Desktop recovery copies saved DocState::doc, which is the version shown while a version (or a live text preview) is previewed, so a crash during a preview brought the old version back as the unsaved file. What a recovery copy holds is now one shared helper, branches::recovery_copy: the document itself (document_itself(), the same document the unsaved-changes check compares), its history and its branches. The desktop copies and the browser's (#1580) both use it.
The canvas applies layer blurs one after another and casts drop shadows from the layer they leave, cut out by it. The SVG filter blurred SourceGraphic once per blur into the same result, so only the last was drawn, and its shadows came from the unblurred alpha: blurs of 12 and 16 with a shadow came back from an export 73 levels off. The blurs now chain, the shadows are cast from the last blur's alpha and cut out by it, and the import reads that back as one blur as strong as the stack with the same shadows. Coloured copies of a blurred layer are still reported, not read as black shadows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M6: FigJam .jam files open like .fig files (#1342)
A .jam is the same archive as a .fig (its canvas header reads fig-jam.), which the .fig reader already took; File > Open and its filter, drag and drop, file.import and the command line now take .jam too, and the home screen shows its thumbnail. FigJam layers come in as the .fig reader makes them (stickies and shapes with text as frames with their generated layers, other FigJam layers as frames or vectors with a note, unknown types left out with a note). Tests use a synthetic fig-jam. canvas and a .jam made from our own export.
* M12: a FigJam .jam file opens as a board (#1342)
The .fig reader marks a canvas with FigJam's fig-jam. header as FileKind::Board, so a .jam opens with the board toolbar, the dot grid and the board panels and saves as a board; a .fig stays a design file, and importing a .jam into an open file leaves its kind alone. Its objects stay ordinary layers (frames and vectors with the existing import notes). Builds on the .jam opening branch (#1586).
A search offered engine commands that need parameters the palette can't give, such as image.place (Place image, needing data or path), which failed with Invalid parameters and sat above the menu's Place image... row with the same label. A search now leaves out a command no menu lists when its params doc names a parameter without ? (figcraft_engine::cmd::needs_params, worked out once per command), and a row no menu item gives whose label reads like a menu item's. Place image runs the menu item, which opens the file dialog.
Someone who joins with a comment link (or its join code) sees the file live and adds, replies to, reacts to and resolves comment threads, but can't change the design: their app refuses other edits and the server accepts a comment link's batch only when every change in it touches comment threads. Sharing online makes a comment link and code beside the edit and view ones; the Share popover offers Can edit, Can comment and Can view, and the commenter's toolbar shows a Can comment badge with Move, Hand and Comment.
Backward compatible: clients announce the comments word; one that doesn't is welcomed with a comment link as a viewer and gets comment links listed as view links, a level this build doesn't know reads as view, and an older server that can't make a comment link leaves the share without one.
The test only checked for a large finite margin, which the old largest-effect margin (1e308) passed as well. It now asserts the two capped reaches added up, so it fails on a return to either the old maximum or an overflowing sum.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With pref.scrollWheelZoom on, a mouse wheel zooms the canvas about the pointer and Cmd/Ctrl+wheel scrolls; Shift+wheel still scrolls sideways, and trackpads, pinches and touch screens are unchanged. A wheel is told from a trackpad by the device macOS reports, elsewhere by line versus pixel steps.
Layer blurs and texture apply one after another and drop shadows are cast from what they leave, but an export grew the node's box by each effect on its own. Stacked blurs came out cut at the image's edges (eight blurs of 10: alpha 10 at the edge), and a sharp shadow under a blurred layer lost the blur around it (alpha 126). Blurs now combine as one whose squared reach is the sum of theirs, texture adds on, shadows grow from that, and paint_reach follows the same rule. Reaches are clamped so adding them stays finite.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
comment.list takes mentions (a name, any case): only threads where a message, the first or a reply, @mentions it, by the same parser the thread view highlights with and each message's mentions reports; combined with the other filters.
wasm32 can't catch a panic, so the command and frame guards do nothing in the browser and a panic ended the page with its unsaved changes. The page now follows its files with unsaved changes after every frame; its panic hook writes them to local storage as .figcraft copies (as many as fit) and says on the page that it stopped. The next page opens them again as unsaved files: file.recover also takes the copy as base64 data. The recovery-copy helpers move from the desktop app into figcraft_engine::branches so both share them.
Each sweep takes at most 250 000 segments and 2 million possible crossings (pairs of segments whose bounds overlap, counted in O(n log n)), and a fold at most 4 million segments swept over its steps; past them the operation is refused with PathOpsError::TooComplex before it runs. A boolean group past the caps draws without its result, with a warning once, and the refusal is cached; vector.flatten refuses it instead of dropping it.
clap derive types replace the hand-written parser: typed values, usage
errors (exit 2) before any work, help from the same definitions.
Syntax changes: a repeated option is a usage error (was: last wins; the
automation roots still repeat); a value starting with `-` needs `=`; a
positional starting with `-` needs `--`; `--o` and the `version`
subcommand are gone (`--version` stays); `<sub> --help` prints that
subcommand's help; usage errors are clap's text; `--format` is checked
before the file is opened.
The browser worker's tests timed the wait after a preview with real sleeps and assumed two polls in a row land within the last full render's time, that a scene tuned to render slowly still renders slower than the preview threshold when the test runs, and that the first image after a request is its last. On a busy machine (the full suite) a test thread can be held off the CPU between polls for longer than the wait, a scene tuned while the machine was busy can render quickly once it isn't, and fonts another test loads mid-render make the worker show that render and draw it again.
Inline::poll_at takes the time (poll passes now) and, in tests, Inline::with_threshold sets when renders get a preview. The four wait tests poll at times they choose on a plain page that previews every render from scratch, the measured-time test previews once renders have taken any time at all, and the refined helpers wait for the last image of a request: no sleeps, no dependence on render speed or on other tests. The capped test now also checks the preview is not refined just before the last full render's time.
Found while testing crates/doc/src/text.rs:150 (TextCase::apply_impl, Title arm) with is_word_char at :156
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Properties were listed by name, so a rename moved one. Components now keep prop_order (the names as they were made; a rename keeps the place) and every listing follows it. Make slot opens a name field under the frame's Slot row; Enter makes the slot, Escape cancels, a taken name keeps the field open.
The web transport implements LiveLink::publish and the hosting and sharing background requests over the page's WebSocket. Managing connections run as exchanges (exchange.rs) in the browser client's driver; the requests and what their answers mean are shared with the desktop transport (requests.rs). The Share popover offers online sharing in the browser (no local network), and Publish to server opens there. Fixes#917.
Where the browser lists them (window.queryLocalFonts, Chromium), the font menu offers Use fonts installed on this computer; the fonts join the font database by name and each style's bytes are read only when text uses it, then auto-sizing text refits. Rust only (js-sys/web-sys). Fixes#1051.
The SVG parser reads neither, so fills came in black and strokes were dropped without a note. They are resolved before parsing: color(srgb|display-p3 ...) converted to the document's profile, color-mix(in srgb|oklab|oklch, ...) evaluated as CSS Color 5 defines it. Other spaces are approximated with a note; unreadable calls become a mid grey with a note. Bounded length, nesting and count.
Layer blurs and texture apply one after another, so together they reach further than any one of them, and drop shadows are cast from the blurred layer. effect_margin took the largest single effect, so a layer cut at a view's or band's edge missed content that still reached in: four blurs of 25 drawn in 64-row bands differed from a render in one piece by 8 levels, and a blurred rectangle above the view cast its shadow into it 143 levels off. The margin is now the layer effects' reaches added up, plus the largest of the others; each reach counts up to 1e12 so the sum stays finite.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The settings folder comes from craft-settings' rules: FIGCRAFT_CONFIG_DIR,
a `portable` file next to the executable (config/ beside it), then the
platform folder every earlier version used. The Windows portable zip ships
the marker, and a portable folder without preferences starts with a copy
of the platform folder's.
Direct3D 12 compiles with FXC and no device compiles wgpu's indirect-call
validation shaders, the two causes of "Parent device is lost" at start.
When the graphics still fail before the first frame, the app starts again
with the next of a fixed list of attempts (Windows: Vulkan, OpenGL; Linux
and FreeBSD: OpenGL; then lower requirements on the power-saving GPU), and
the last failure's message names what was tried.
BFlat