* M6: pasted layers keep the colours of their shared paint styles (#1638)
A clipboard links a style by the guid it has in its library, which it doesn't hold, and by its library asset, whose copy it does hold. Mapper::resource_id now takes a guid only when the file holds a node by it, else the asset, so .fig import and paste agree; a link to a style that really isn't there keeps the layer's stored colours, with a note.
* M6: Swedish catalog: keep both new import notes after the merge
* Docs: one copy of each row and status item, and a check that keeps it so
Merge resolutions that kept both sides had left progress-doc rows (the .fig row four times, co-editing, live cursors, Windows, M0.9) and status-line items (302 in all) twice or more. The rows are merged keeping every distinct fact, the newest wording winning (#1649's connector routing included); repeated status items are removed, the first kept. cargo xtask docs, run by ci, fails on a repeated table row or status item.
* Docs: remove the status items the merge of main brought back
Merging main into #1658 rebuilt the status lines from main's copies, which still listed items twice: ROADMAP 2, docs/roadmap.md 5, docs/gaps.md 43 and docs/target-app-parity.md 190, whose unclosed parenthesis also came back and hid them from cargo xtask docs. They are removed again, the first copy kept, and the parenthesis closed; every distinct item is still present against main.
Builds on #1618. Variables::get and Variables::collection take an exact id before a name; stored mode ids go through Collection::mode_by_id (expressions, the prototype's mode switch); font family variables and the colour picker, prototype editor and panels read stored ids exactly. Ported from fix/stored-ids-exact.
The board connector router moves, unchanged, from the engine into figcraft-doc (route_points, RouteEnd) so the .fig importer can use it: an elbowed or curved connector saved without its path runs in right angles or one curve from the side of each layer it joins (the file's magnet, else the edge its point is on, else the side facing the other end). Straight connectors are unchanged. Gap 1.
* M6: figcraft-cli connect publish publishes code connect mappings to a design file
figcraft-cli connect publish <file.figcraft> [--mappings figcraft.connect.json] [--dry-run] links each entry's component or set by name through codeConnect.import, warns about and skips links that don't fit, saves the file in place through the engine's file.save, and exits 1 when nothing links. Gap 16.
* M6: the generated man pages include connect and connect publish
The placeholder an import stores for an image it lacks is now striped, not flat grey; the fill keeps the image's name; .fig and paste give a fixed, translated note per image; edit.pasteDesignLayers returns missingImages, the paste's message names those layers, and a paste with notes opens the import report.
The engine notes when a drag has only moved the selected layers by one offset (Interaction::moved_by); the canvas keeps the outlines, box and slots from the drag's start document and shifts them, the selection spacing and the Design panel's values (but the position fields) come from that document, and the Export section's values are kept. Dragging 648 layers: UI pass 2.10 to 1.44 ms a frame.
The CSS export left out a colour calculation that resolves to nothing but still wrote an alias to
it as var(--name); where no block defines that property, whatever reads the alias is invalid.
CSS (and Tailwind) and Penpot now share one rule, Variables::flattener: an alias that leads to a
calculation some mode of its collection leaves out goes out as the value it shows in the mode,
or is left out when it shows nothing.
* M12: create the next board shape already connected (#1597)
A selected sticky or shape shows a + beside each side; hovering one previews where the next goes, clicking makes one of the same kind, size and colour there, joined by a connector, and starts typing in it, in one undo step. board.addConnected {id?, side?, edit?} does it from commands.
* M12: the connected-shape handle and command in every full catalog (#1597)
* M12: Shift+Cmd/Ctrl+Enter on a board shape adds the next one already connected (#1597)
inspect_fields::compact_examples_run_verbatim_from_the_docs split docs/mcp.md and docs/control-protocol.md on text ending in a newline, so on a Windows checkout with core.autocrlf (CRLF docs) it panicked and cargo xtask ci failed locally. The test now reads both with LF line ends. Found while gating on Windows.
Fonts name a family by its typographic name and by a legacy name per weight; fonts without typographic names are installed under the legacy ones only. FontDb::resolve finds either from the other, and matches names across case, spaces, hyphens and underscores; every lookup (import, paste, REST import, rendering, missing fonts, the font report) uses it. A weight-number style goes by that weight, and .fig and REST imports wait for the installed fonts to be indexed.
Why type_of gives references no type, what display shows, how resolve_memo bounds one resolve,
and what variable.create and variable.set check before anything changes.
A colour calculation that resolves to nothing in a mode is left out of that mode's token set, but
an alias to it still went out as a {name} reference to the missing token. Such an alias, and any
alias leading to one, now goes out as the colour it shows in the mode, or is left out with the
calculation's warning when it shows nothing. A calculation left out in only one mode of its
collection counts too: a reference to it from another collection would dangle whenever that
mode's theme is active.
A colour variable's value in a mode can now be {"alpha": {"variable", "opacity"}} - that
variable's colour with its alpha replaced - or {"mix": {"from", "to", "t"}}, a per-channel blend
of two colour variables or literals over the stored values (in the document's colour profile),
alpha included. Calculations resolve lazily like aliases, so bound layers, mode switches,
instances and undo follow their sources; they compose, and a loop, or a calculation that reads
more than 32 variables in its collection's modes (other collections in their default modes), is
refused when it is set. Operands are stored by id and read by exact id, as stored references
are. Files keep them; REST, CSS, Penpot and .fig exports write the colour they show in each mode;
deleting a source leaves its readers on the colour they showed; library import and swap
re-point their operands. The Variables panel shows them with their formula and can swap one for
an alias or detach it to its colour.
The .fig import gave a file one decoded value per byte of its compressed canvas (at least 20M, at
most 128M). Dense legitimate files decode past that, further since every scalar field counts, and
the whole import was refused as "not a valid FigCraft file: .fig file: message: too many values".
The budget is now four values per compressed byte; the floor and both ceilings (128M, 32M on 32-bit
targets) are unchanged, so a small file that unpacks past the floor is still refused and no file
decodes more values than the ceiling allowed before. Generated 21 MB and 15 MB canvases (220,044
and 108,012 layers) and a 49.7 MB one budgeted at the ceiling (530,106 layers) now open whole
without warnings; the 49.7 MB one's test peaks at about 9 GB.
Background blurs and glass read the pixels drawn under them, and a full
render's backdrops end where the view does. The retained renderer differed
in two ways.
render_into drew each damage rect with a margin all round, past the edge
too, so a backdrop there read content a full render never drew: with a
background blur of 60 across the left edge, moving a square beside it
patched 10 levels off a full render, and moving one entirely off the view
89. The margin now stops at the edge.
A pan re-rendered only the strips it uncovered. At the opposite edges, which
moved into the old frame, the shifted pixels kept what their backdrop had
read past the old edge: a background blur of 10 across the view was 89
levels off after a pan of 30. Those edges are now redrawn as far as a
backdrop reads.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
paint_bounds, the area a change to a layer can affect, was its paint bounds
(effects included) grown again by effect_margin, what its effects read from
outside. Drawing an area already takes that in, as culling does, so the
damage of an edit near effects was counted twice: moving a 120 x 120
rectangle with a shadow falling 40 rendered 122,844 pixels for 56,876 it can
change, and edits that would fit a partial render fell back to full ones.
The retained test that a group's shadow had pushed past the partial share
gets its 300 x 220 view back.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* M8: opening a file contacts no team library server it names until you check it (#1097)
* M8: translate the team library check strings (#1097)
* M8: the Assets panel library test checks what allows a library, without its own fake link (Sonar duplication and coverage)
* M8: cap the people a client keeps track of and the deleted ids a session remembers (#1121)
A hostile host or server could grow a client's presence, seen-peers and who-is-here lists without bound, and a session's deleted-id set only grew. Clients keep at most MAX_PEERS (4,096) others, leaving newcomers out with a warning; max_people is at most that; a session remembers at most 262,144 deleted ids, the lowest forgotten first.
* M8: a max_people above what clients track is lowered with a warning, not refused (#1121)
Refusing it would stop a server whose existing config says more from starting after an upgrade. Config::people() lowers max_people to MAX_PEERS for the core's limits, and the startup warnings name the configured and the used value.
* M8: the live welcome's handling in its own function (Sonar S3776)
UPDATE_APP (native builds; craft-update is a native-only dependency) carries our 128 px icon and accent (#02da96, its colour in ArtCraft Launcher). The updater, the self-update
helper and craft-update's startup prompt all get them through crate::UPDATE_APP; the
prompt draws the icon, the accent and "by ArtCraft" (craft-update's default developer).
craft-update is pinned to storytold/craft-libs main at 1065653e7 (#434).
local_reach gave up on transforms whose determinant was under 1e-12 and
kept the reach in page units, so a layer scaled to a ten millionth (a
billion units drawn 100 wide) had its shadow cut (245 levels). It now
converts any finite, non-zero scale, capping the result where the sums
would overflow.
Groups and stroked text spread their shadows as pixels, in page units:
their spread no longer also counts in local units.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* 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)
The renderer draws drop shadows, layer blurs and texture at their size in
page units however a layer is transformed, but sized a layer from its local
box grown by the effects' reach in local units, so under a transform that
scales down the effect was cut (245 levels on the canvas, exports holding
half to 60% of the paint).
bounds_within now takes the layer's local-to-page map and turns the reach
into local units per axis. A drop shadow's spread of a filled outline is the
exception: the outline is grown before it is transformed, so that part
follows the layer's scale. Layer bounds and own_paint_bounds take, per
shadow, the larger of the two readings, so an unscaled layer comes out as
before; exports of scaled-up spread shadows, which were cut, now hold them,
and so do the corners of a turned rectangle's spread. The mitre reach is
clamped like the other reaches, so a huge spread no longer makes bounds NaN.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A panic in a Session::grouped closure outside any command skipped the group rollback: the
documents stayed half edited, the history limit unbounded, live documents held and the group
depth raised. checkpointed now catches the unwind, runs the same rollback as a failed group and
resumes the panic, so the caller's guard still reports it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
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 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
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 <[email protected]>
* 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 <[email protected]>
---------
Co-authored-by: Brandon Thomas <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
* 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 <[email protected]>
---------
Co-authored-by: Brandon Thomas <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
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 <[email protected]>
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 <[email protected]>
* 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 <[email protected]>
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 <[email protected]>
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 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
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 <[email protected]>
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.
Session::execute_as records the agent's presence: its name, the file, and the innermost layers its last edit changed or the selection it set, for 8 seconds after its last command. The app shows it as an avatar in the people row and an outline with a name tag on the canvas, in its own colour. Display only: no undo step, never saved, never synced to a live session.
Found while testing crates/color/src/lib.rs:~173-180 (parse_css, hsl/hsla arm)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
While the interaction editor has an Open overlay action with Manual position open, the overlay frame shows dimmed and outlined where it opens, from the top-left of the layer that opens it. Dragging it sets the action's offset through proto.action.set in one gesture: the X/Y fields follow live and the drag is one undo step.
* M9: one name for comments and live sessions, shown in the Join dialog, kept in the browser, renamed live (#1380)
* M9: a reconnecting session's reset is its own helper (#1380)
ci(deps): bump storytold/craft-common-actions/.github/workflows/windows-arm64.yml from 8bb4e5ab95443171104ad88113fe1ef9b0e5ce3f to fa5ef18780cf5e7d5d3b8de068de56cffc83cac5 in the actions group
* M9: a per-address budget for server admin requests, and join codes found by hash (#1075)
* M9: the server tests share one snapshot and address helper (#1075)
These scripts write no spaces between words, so a paragraph was one word
that only broke where it overflowed, mid-word (พัฒนาซอ|ฟต์แวร์).
breaks::word_breaks finds the word starts inside each run of them with
ICU4X's line segmenter (icu_segmenter 2.3, matching the icu_* crates in
the lock), and break_lines may break there when the next glyph starts a
new cluster. Where a run meets other text the UAX #14 subset decides as
before; a word wider than its box keeps #1400's emergency break.
Desktop uses the segmenter's dictionaries; the browser uses its LSTM
models (about 0.39 MB of wasm instead of 1.86 MB, which the 24 MiB web
gate can't take). ICU4X breaks between stray marks are filtered through
safe_emergency_break, so junk input never splits a cluster.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6DyK6i9Nvvsp3tp77UP
* M4: a letter and its combining marks shape with one font (#1478)
* M4: the stacked-mark test checks positions only in fonts whose anchors stack them (#1478)
* M4: the cluster face rule is tested with stand-in faces, so it runs without system fonts (#1478)
* M4: the #1478 tests use stand-in faces only, so they run everywhere
Every message (the first and each reply) gets `mentions`: {name, known,
you} for each @mention, by the rule the thread view highlights with
(figcraft_doc::mentions, now shared with mention_ranges). Names resolve
against the file's people and you; unknown names are kept with
known: false; [email protected] mentions nobody.
The board canvas's background grid is now a preference (`board.grid`): dots
(the default), lines, graph paper (a strong line every 2-12 cells over fainter
ones, like Japanese section paper) or none, with spacing (4-512), mark size and
contrast. Preferences gets a Board grid section; right-clicking an empty board
offers Board grid settings. Zoomed out the grid still thins out (by halves, or
down to graph paper's strong lines), and dots no longer shrink to a single
pixel, so the grid stays visible at low zoom.
The renderer draws lines as well as dots, with a configurable strong-mark
rhythm; a crossing of two lines is blended once, in the stronger colour. A
hand-edited preferences file falls back to the default grid instead of losing
the other preferences.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FEkCVd8n1eC6xTHebmHMDR
A mask group's pixmaps cover only where its mask or content can paint within the view, rounded out to whole pixels, drawing the same pixels (a 40x40 mask on a 3000x2000 view: 96 MB -> 61.5 kB). Isolated layers share a 2 GiB per-thread budget; a layer past it is left out with a warning.
From the review of #1496: three background blurs stacked on each other across the band
boundaries draw the same in bands as in one piece (within anti-aliasing), so the bands'
overlap, the reach of the largest blur, is enough, as it is for the desktop's parallel bands;
pinned by a test. The docs and the `MAX_BANDS` comment no longer put the longest frame at "an
eighth of the render": a frame draws at least a band, a part of the render rather than the
whole, and the numbers measured for #1496 are cited instead.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
* M2: components can't contain their own instances (#1084)
Edits that would make a component hold an instance of itself, directly, through other components or through another variant of its set, are refused with the loop named. Files holding such a loop open with one instance per loop detached and a note; a rebuild never expands an instance inside a copy of itself, so a loop merged in a live session stops growing.
* M2: a variant may hold a sibling variant's instance; only true loops are refused (#1084)
Each variant counts as its own component: refusals and open-time repairs apply only to a component leading back to itself through instances (an instance-swap value included), so files with nested sibling variants open unchanged. sync_instance_depth is split into match_layers and copy_layers.
Found while testing crates/engine/src/cmd/props.rs:614 (also :708, :717, :718, :826 and :1007)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Once a zoom had paused, the browser drew its full render in one frame. On a heavy page that
frame took seconds, and no input was handled meanwhile: the wheel steps and drags that came
were seen only when it was over.
The renderer's `Retained::begin` hands a full render back as `Steps`, drawn `band_h` rows at a
time with the same `render_band` and `put_band` the desktop's parallel bands use (bands overlap
by what background blurs read and keep their own rows; a blur reaching past a band makes the
bands taller rather than one piece), and `finish` keeps the image as a full render. A render
that is reused or patched is done at once, as before.
The browser's `Inline` worker refines through `Renders::refine_for`, which draws at least a
band a frame, more while they fit in `STEP_MS` (24 ms), and asks for the next frame while bands
remain, so input is handled between them; a view has at most eight bands, so on a page whose
render takes seconds the longest frame is about an eighth of it. A render drawn for the very
request is done once drawn. A pan or an edit meanwhile at the same zoom keeps the render going
and is patched in once it is drawn; a pan too far to patch or an edit the patch can't take draws
it again in steps, the image just drawn showing till then (`Refined::Shown`); a zoom, another
page or file, another size or other render options drop the render for one of the new view,
and what its bands took stands as the measure the next zoom's preview and wait go by. A render
in steps counts as one render (glass keeps the shapes a later render asks for again). The
desktop, with its render thread, is unchanged.
Measured in Chrome on two synthetic pages (20 images and 60 shadowed rectangles, with and
without two background-blur panels) whose full render takes 3.6–4.0 s at a deep zoom: the
longest frame while that render runs went from 3.7–4.0 s to 0.9–1.4 s (blur) and 0.69–0.80 s
(no blur); the bands together take longer than one render (the refined image comes after
6.7–8.1 s and 5.3–5.5 s instead of 3.8–4.1 s), each band rendering its overlap again and
repeating some per-band work, shadows mostly. The zoom steps are as before.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
With map: true, edit.duplicate also returns map: each original layer's id, descendants included (nested frames, groups, instances and their sublayers), paired with its copy's id, so a batch can edit inside a copy through a step reference instead of inspecting it. Without map the result is unchanged.
The batches a session keeps for apps that reconnect were bounded only by count on a desktop host
and a private figcraft-server (`max_log_bytes` defaulted to usize::MAX). It now defaults to
64 MiB, and figcraft-server takes it from a new `max_log_mb` setting (default 64; open to the
public, also at most a quarter of `max_file_mb` as before). An app behind the trimmed log catches
up with a snapshot, as it already did past the count limits. The protocol is unchanged.
Nine 8 MiB renames with default limits: all 75.5 MB kept before, 58.7 MB after.
Fixes#1069
Tools that act on a document, and the control methods engine.execute,
document.inspect, node.get and ui.render, take an optional `document`
(an open file's id from doc.list / document.inspect, or its path). The
call acts on that file and the tab in front stays in front, so an
attached agent keeps editing its own file when the user switches tabs.
Every replica placed a created, moved or reordered layer by counting each sibling that sorts
before it, so a batch of many layers under one parent took quadratic time, on the server's
single hub thread too. `Replica::slot` now binary-searches the `(key, id)` order on each side of
the layer's old place (a position change has already given it its new key). The order is
unchanged, ties and fractional positions included.
20,000 creations and 10,000 moves under two parents: 20.05 s before, 0.33 s after (debug build).
Fixes#1065
Found while testing crates/engine/src/cmd/variables_io.rs:~92 (css_safe, used by css for String values and mode names)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Every command an MCP client runs acts as the client's name from initialize (clientInfo.title, else .name) with ' (agent)' after it, or 'AI agent' without one, carried to the app as engine.execute {agent}. Its messages, reactions and read state are its own, and its messages are marked (agent: true in the file and comment.list, an Agent label in the thread). The user's name and comments made in the app are unchanged; in live sessions the name and marker travel with the thread.
The server's fixes for a refused batch walked every deleted layer's whole subtree again and
compared each document-level fix with all the earlier ones, so a batch deleting a deep chain took
quadratic time on the hub thread. Each subtree walk now stops at a layer already restored, and
document-level fixes are looked up and sent once per target. The fixes are unchanged.
A refused batch deleting a 20,000-deep chain plus 40,000 comment puts: 27.2 s before, 0.14 s after
(debug build).
Fixes#1063
* M8: large images on the canvas draw from a smaller copy (#1483)
An image shown smaller than its pixels draws on the canvas from a copy scaled down to 4096 or 8192 px on its longer side, cached beside the image in the same budget. Two 16000 x 10000 images (640 MB each) evicted each other on every frame that drew both; as copies they take 80 MB together zoomed out. Moving one over the other: about 2.1 s a frame before, 11 ms at fit and 15 ms at 30% after. Exports, thumbnails and views at 100% and closer draw from the image itself. Tests in large_image_tests; timings in large_image_bench_tests (ignored).
* M8: drop the ignored large-image benchmark; its numbers are in the PR (#1483)
The header check from #1491 moves to figcraft_format::jpeg so the format crate and the renderer share it, and decode_thumbnail now refuses a progressive JPEG that samples a component more finely than its first (zune-jpeg 0.5.15 asserts on it; in the browser a panic aborts the page). A crafted file's thumbnail on the home screen and in recent files shows no picture instead of crashing. Regression test through decode_thumbnail and archive_preview.
Style links that name a library style by its asset key and version resolve to the file's copy of the style, as variable links already did. The linked text style beats the copies of its values below the row that links it; rows above that one still win. Override rows that restate a nested level (overrideLevel) apply at that level instead of in wire order.
* M6: .fig import keeps the file's saved line breaks while the text and its box are unedited (#1236)
The line breaker already keeps trailing spaces and the last glyph's letter spacing out of the fit test and doesn't round box widths; imported text wrapped a word early or mid-word because local fonts measure slightly differently from the file's. The import keeps the saved line starts (derivedTextData) on the text with the box width and a stable key of its layout content; layout follows them while both match and no line overflows its box by more than 3% (+1 px).
* M6: local check that imported text wraps at the line breaks its file saved (#1236)
Runs only when FIGCRAFT_FIG_SAMPLES names a folder of .fig files: each fixed-width text with saved breaks wraps there once imported, or falls back by the overflow slack; prints the counts.
A missing family the suite's font store has (Google Fonts, Noto) gets a
Download button. The first one asks the user to allow downloading fonts,
then craft-libs' craft-font-store installs pinned, checksum-verified files
into the Craft apps' shared font folder on a thread of its own. The font
scan reads that folder; once installed the text fits the real font.
font.download / font.downloads drive the store as its agent: refused while
downloading is off, never able to turn it on. Desktop only.
A press more than a few pixels from the edited shape left edit mode at
once, so a drag over another layer moved that layer. The press now
starts a marquee; released without a drag it still leaves edit mode and
replays the click to the normal tool.
The browser build's wasm was 25,982,540 bytes, past the gate raised for 0.12.1. It is now
23,915,213 bytes (22.8 MiB), 1.2 MiB under the restored 24 MiB gate:
- Catalog rows keep their plural forms behind a reference: Text is two words, not room for
four strings, across ~52,000 baked-in rows (-454 KB).
- The browser build compiles with wasm SIMD (apps/figcraft-web/.cargo/config.toml, wasm-opt
--enable-simd), so egui's glyph rasterizer leaves out its scalar SIMD fallback (-1.6 MB).
The plugin workspace is built from the repository root and stays without SIMD.
- Revert the gate raise (cb227c97); the hosting guide names the WebAssembly SIMD requirement.
* M3: shadows of images with transparent pixels follow their alpha (#1289)
A drop shadow on an image fill was hidden behind the layer's whole box, so a
transparent PNG's shadow only showed outside its bounding box. Fills that
paint their own alpha (patterns; images with transparent pixels or that leave
part of the box empty) now cast and hide drop shadows, and shade inner
shadows, by the layer's pixels, like groups. Opaque pictures keep their
outline. Each image is scanned for transparency once. Canvas, PNG, SVG and PDF.
* M3: the image shadow tests import their names explicitly (#1289)
* M1: Crop image from the Object menu, the right-click menu and the image fill row (#1464)
* M1: crop mode shows the image dimmed outside the layer and heavier crop marks (#1464)
* M1: the handle overlay draws gradients and crops in their own helpers (#1464)
Found while testing crates/doc/src/props.rs:1341 (ExportSize::label, via trim_num at crates/doc/src/text.rs:62)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Hidden containers hide their whole branch from snapping, the 5 px reach is converted through the owner's scaled axis so a zoomed frame snaps at 5 screen pixels, and a click on a guide only selects it. Tests cover all three in guide_snap_tests.rs.
Signed-off-by: Ahmed Mahmoud <[email protected]>
`ui.input` over `window.figcraft` (the page's control channel, `?control`) never did anything in
the browser: the batch was queued for the next frame as on the desktop, but the browser shell
has no `raw_input_hook`, the eframe callback that hands a queued batch to the frame about to
draw (`windows::take_input`). The events stayed in their queue and the request's promise never
settled, so an agent driving the page could run commands and take screenshots but not point,
click or type.
The browser shell now has the hook the desktop app has: the batch goes into the frame's raw
input, is applied by it and is acknowledged once the frame is presented, as on the desktop.
Checked in Chrome on the release web build: `ui.input` with a key (O, R, F: the ellipse,
rectangle and frame tools) and with a pointer gesture (a drag with the rectangle tool: a
rectangle drawn) resolves and takes effect, where on `main` the same requests time out and
change nothing.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Dragging a guide, or moving one, now snaps to nearby layer edges and centres on the guide's own axis: within 5 screen pixels, gated by Snap to Objects and suspended with Cmd/Ctrl. Page guides see the page's top-level layers; frame guides see the frame's box, its layout grids and every visible layer inside it, in the frame's own space.
Snapping sits in the interaction path (canvas/guides.rs, not guide.add/guide.move) so the commands stay deterministic for the control channel and MCP. Tests: guide_snap_tests.rs in the engine and ui-egui crates; docs/ui-parity.md updated.
Signed-off-by: Ahmed Mahmoud <[email protected]>
- An image a layer or style still shows can't be removed: removing it and
putting other bytes under its key would swap the picture that layer
shows, which giving the key other bytes directly can't. A batch made by
the diff stops using an image (node changes, then styles) before it
removes it; an undo keeps the changes that apply, so its image may stay.
- Restore and replay leave out the image entries history written under
older rules can't apply, using one method shared with the undo history
(Replica::apply_kept, extracted from history.rs). It is exact for every
image rule, including removals in use and replacements, and leaves out
only image entries: any other change that doesn't apply is still an
error. The image-only filter it replaces is gone.
- A server whose restore or replay left image entries out, or whose saved
log puts images, catches every resuming client up with its snapshot,
including one at the same sequence number.
- Tests: an_image_in_use_cannot_be_removed,
an_image_no_longer_in_use_can_be_removed, undoing_an_image_layer_still_works,
a_restart_replays_past_a_logged_removal_of_an_image_in_use,
a_resume_after_history_was_left_out_catches_up_with_the_snapshot,
a_resume_after_a_clean_restart_catches_up_with_the_log,
a_snapshot_with_a_broken_layer_change_still_fails_to_open.
- Docs: collab-protocol.md, collaboration-parity.md.
From the review of #1491: patching a component's sampling in the header of a baseline JPEG
`image` wrote leaves its entropy data for 1 × 1 components, so decoding it could only fail and
the test proved nothing past the frame not being refused. The fixture is now a baseline CMYK
JPEG as jpeg-encoder writes it with 2 × 2 sampling (its black sampled finer than its cyan), and
the test asserts it is left to the decoder and decodes.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
- without_refused_images judges each image entry against the document it
applies to and the batch's earlier entries (image_for, shared with
apply): a logged put of other bytes under a key the replay state holds
is left out, so a restart no longer stops there and drops the rest of
the log.
- Server::restore keeps its saved log only when the log puts no images:
a resuming client may already hold other bytes under a logged key, which
the restored state can't check, so clients catch up with the checked
snapshot instead.
- Tests: a_restart_replays_past_a_logged_image_replacement,
a_client_resuming_past_a_refused_image_in_the_saved_log_catches_up,
a_client_resuming_past_a_replaced_image_in_the_saved_log_catches_up.
- Docs: collab-protocol.md, collaboration-parity.md.
zune-jpeg 0.5.15, the decoder behind `image`, asserts in its upsampler on a progressive frame
whose later component is sampled more finely than its first (etemesi254/zune-image#236; an 8 × 8
Y 1×1, Cb 2×2 file is enough). The panic reached everywhere such an image is decoded: the CLI died
writing a thumbnail, the desktop canvas's guard caught it on every render of the page, and in
the browser it aborted the tab. The fix in zune-jpeg (36007420) is in 0.5.16 pre-releases only,
which `image`'s requirement doesn't take.
Such a JPEG is now refused from its frame header before the decoder sees it, as an image past
the byte budget is. The decoder refuses the larger ones of this shape itself, so nothing that
decoded is lost; baseline frames of the same shape decode and are left to it. The check can go
once zune-jpeg 0.5.16 ships.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
- The entry size cap is figcraft_doc::ImageData::MAX_BYTES (256 MiB), the
cap the .fig, .sketch and REST importers already read images with (now
one constant), instead of 32 MiB: an imported image larger than 32 MiB
made the document unjoinable and its server snapshot unrestorable.
- History written before the checks (a restored or joined snapshot, a
replayed server log, a file published to a server) applies without the
image entries they refuse (docprops::without_refused_images), instead of
the file not opening or a restart dropping every edit after the image.
- A key that already holds an image can't be given other bytes: producers
key images by their bytes (the REST importer's placeholder keys always
hold the same picture), and merge treats a key's image as never
changing.
- REST import: an unresolved imageRef shaped like a content hash shares
the rest-image-missing placeholder (the placeholder's bytes don't hash to
it); ImageData::claims_content_hash is shared with collab.
- Tests: collab image_tests (4 new, size test uses a small cap), format
image_tests (1 new).
- Docs: collab-protocol.md, collaboration-parity.md, file-format-parity.md.
The app's logic runs while the main window is minimised, without the
frame that acts on a close. Two quits there queued the main window's close
twice: the first let it take over another window, and the second then
quit the app over that window's unsaved changes, as answered.
- close_current queues a window's close only once.
- Test: quitting_twice_before_the_window_draws_closes_it_once.
- docs/hardware-parity.md revision history.
- finish(): a viewer-fired link opens directly (notice skipped) only while
no notice shows; any other link that fires while one shows is dropped.
- finish(): raising a notice ends the press under way, so its release
doesn't fire the prototype's mouse-up actions (a link, a navigation)
behind the notice, which takes no input while it shows.
- The mail notice says the link opens the mail app to write a message
(opening a mailto: link doesn't send anything).
- Tests: nothing_opens_while_a_notice_shows_with_the_notice_skipped,
a_press_that_raised_a_notice_doesnt_release_behind_it.
- docs/prototyping-parity.md row and revision history.
Image entries applied bytes to any key with no checks (audit COL-9): a
collaborator could poison the shared content-addressed image store, claim
absurd dimensions, balloon memory, or name an image `../../x`, which a
saved file then wrote as the archive member `images/../../x`.
- Every key must be 1–128 ASCII letters, digits, `-` or `_`:
`figcraft_doc::ImageData::is_valid_key`, promoted from the format
crate's private `is_valid_hash` so the file reader and collab share one
rule. The reader now also takes `-` and `_`, so the REST importer's
`rest-image-missing` placeholder survives a save and load.
- The REST importer's side of the same attack: an unresolved `imageRef`
that can't legitimately be an archive member name (path text, an
over-long ref) shares the `rest-image-missing` placeholder instead of
being keyed by its raw name
(`image_tests::a_path_like_rest_image_ref_never_names_an_archive_member`).
- A key with the shape of a content hash (64 hex chars) must be exactly
the lowercase sha256 of the bytes, the form `image_hash` writes; an
uppercase look-alike is refused. Other keys (svg FNV-128, REST
placeholders) claim no content address and only meet the rules above.
- A side is at most 65 535 px; decoded bytes at most 32 MiB, with the
base64 text capped before decoding. 0×0 stays valid: the format crate
stores it for sizes it can't probe.
- Refusals are Error::Invalid and reject their batch, so every replica's
store stays clean; catch-up snapshots apply through the same check.
- Docs: collab-protocol.md, collaboration-parity.md and
file-format-parity.md (rows, status lines, histories).
- Tests: collab image_tests (8), format image_tests: round-trip and the
REST placeholder gate.
winit's default macOS app menu ends the process through AppKit's
`terminate:` at once, so ⌘Q and Quit from the menu bar never asked about
unsaved changes (the recovery copies were all that survived).
- Install craft-libs' app menu (craft-menu, same rev as the other
craft-libs pins; macOS only). Its Quit sends the main window a close
request, the same path as the window's close button, so
windows::on_close_request cancels the close and the unsaved-changes
question asks first. Failure to install keeps winit's menu and logs it.
- windows_tests::a_quit_with_unsaved_changes_waits_for_the_answer pins
the contract the menu relies on: a close request with unsaved changes
sends CancelClose and asks; Don't Save then closes.
- Like the close button it closes the main window only: with several
windows open, each quit closes one (documented).
- docs/hardware-parity.md macOS row, status line and revision history.
- Cargo.lock: craft-menu, craft-quit, craft-keymap, muda and their deps;
cargo's own minimal resolution also moves a few windows-sys/getrandom
edges to versions already in the lock.
An Open link action carries a URL from the document and fires from whatever
trigger the prototype sets, so an unattended presentation could open web links
and mail messages one after another: a mail link bypassed the leaving-the-
prototype notice entirely, and a skipped notice let any trigger open any
number of links.
- finish() routes every openable URL, mail included, through the notice.
- Skipping the notice ("Don't show this again") opens at once only a link the
viewer's own click, press, drag or key fired, one per frame. A delay, the
pointer passing over a layer (enter, leave, hover) and a timed mouse
trigger that fires after the gesture is over still show the notice, which
then offers no checkbox: it couldn't skip that notice anyway.
- The player tags each queued link with whether the viewer fired it
(Presenter::fire_from); an OpenUrl outside an interaction defaults to not.
- The notice covers mail links (new locale string, mail wording).
- Engine prefs/command doc text follows the rule; behavior untouched there.
- present_links_tests: shared setup helpers for the link tests; new
tests for the skip rule (delay, hover, delayed mouse-up, key, one link per
click, checkbox only for clicked links); the scheme-filter test now clicks
with the notice skipped so the filter alone keeps those URLs out.
- docs/prototyping-parity.md row, status line and revision history updated.
A page kept its ruler guides in `NodeKind::Page`, and since frame guides
(#70) every layer also has its own `guides` field. Both are written as
`guides`, and on reading the layer's own field takes the key, so:
- a reopened page's guides landed in the layer field and were not drawn;
- saving that page wrote `guides` twice (the page kind's list, always
written, then the layer's), and the file no longer opened ("duplicate
field `guides`").
Every layer, pages included, now keeps its guides in the one layer field;
the key in the file is unchanged, so existing files open with their guides.
The field reads every `guides` list a layer has, so files already saved
with the key twice open again with both lists, in file order. This also
covers collab `Create` payloads from older peers.
Regressions: JSON round trips of page and frame guides, a page with two
lists, a bad list still refused (doc); save, reopen, save, reopen through
`.figcraft` and a document.json with the key twice (format). Checked in the
app over the control channel: guides drawn after two save/reopen cycles, and
a file with the key twice opens with all its guides.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6Dyzof21BCrUsByhx2w
Images carried inside a `.fig` message (`image.dataBlob` indexing the
message's `blobs`) were copied out before the message is dropped, one copy
per image hash. Nothing stops a file from pointing many hashes at one
blob: 100 hashes on a 250 MB blob made 25 GB of copies, an out-of-memory
abort from a small file.
The walk now records each hash's blob index, and `embedded_copies` makes
one shared copy per blob for every hash that names it, so the copies are
bounded by the message's own size. Each resolved image is still copied
once on its way into the document, where images are kept by content hash.
Regression: fifty hashes on one blob share one copy; a hash naming a
missing blob is left out.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6Dyzof21BCrUsByhx2w
The Scale tool (and `transform.scale`) scales a layer's children whatever
their constraints. After the edit, overrides are recorded by diffing the
changed layers, and a size or place that changed while an ancestor resized
is read as what constraints give, so it isn't an override. For a scaled
instance that left its layers' scaled size and points unrecorded: the next
rebuild of the instance (a layer added to its main, say) put a 14 px
icon's triangle back at the main's 24 px size. Strokes already kept their
scaled weight, as a separate override group.
`scale_node` now marks `size` and `position` overrides on the layers it
scales below the scaled layer when they are layers of an instance and the
value changed. Undo takes them back with the rest of the edit.
The report's own sequence (clearing effects on the component holding the
icon) no longer reproduces on main; this covers the rebuild path that
still did. Regressions: a scaled instance through value edits and a
rebuild of its main; one scaled with the tool, with undo; a scaled icon
nested in a component and in an instance of that component, through edits
of both mains.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6Dyzof21BCrUsByhx2w
Session::execute ran a command under guard::run but normalised what it
changed afterwards, outside it: a panic in normalisation escaped execute
instead of becoming EngineError::Internal with the edit rolled back. Inside
a batch it unwound through Session::grouped, which then never restored the
undo limit it had lifted or ended the group. Normalising is now the last
step of the guarded result chain, so a panic there fails the command like
one in the command itself (and a batch step fails like any other failed
step); live text previews normalise under the guard too.
The other half of #1085, restoring grouped()'s state when its closure
panics, is left alone: #1419 rewrites that function.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6E2qDnjtMzfPVuEP7iH
The stroke's position dropdown took half the row and the weight field what
was left after the per-side and Advanced buttons: about 40 pt at the
panel's default width, so the variable button sat on the weight's icon and
hid its value, and the row ran 3 pt past the panel's edge. The position
and the weight now share what the buttons leave: the panel's two columns
for layers without per-side strokes, and about 66 pt each with them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6E2qDnjtMzfPVuEP7iH
Repair re-listed unlisted pages with a scan of the page list per page, and
reading a published library kept manifest styles, collections and variables
with a scan of the snapshot's lists per entry: a file with many pages or a
long manifest made both quadratic. Both now look ids up in a set built once.
validate() already indexed its parent/child edges and page list.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01LvG6E2qDnjtMzfPVuEP7iH
Resolve host-tuple to the host triple, pass --target to cargo build when
CARGO_BUILD_TARGET is set, and skip the --version smoke run for cross
builds (they cannot run on the x86_64 VM); cross builds get an ELF check
instead. Archive names carry the arch from the target triple.
Co-authored-by: Claude <[email protected]>
The pixel cap reads only the PNG's header, so a small image followed by a
large tail of data passed it and was copied and base64-encoded whole.
MAX_CLIPBOARD_PNG_BYTES (256 MB, the same cap as design data read from the
clipboard) is checked on the NSData length before the copy on macOS, and
again in the paste path so every Services implementation is held to it.
A PNG past the cap is treated like an unreadable one: paste falls back to
the decoded image.
The 0.12.0 release's Web job failed at 25,319,327 bytes against the 24 MiB (25,165,824-byte) gate. opt-level z, fat LTO and wasm-opt -Oz are already in place and extra wasm-opt passes save only ~20 KB, so the gate moves to 25,690,112 bytes, still under Cloudflare's 25 MiB per-file cap. The wasm grew ~2 MB in three days (egui 0.37, new features); bringing it back down is follow-up work.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The context menu looked only at what is under the pointer on the page, so a name label drawn over
its section selected the section. Copy link to selection then gave every frame in that section the
section's link. Labels now win, as they do on a press.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_015wxxFg3gana4zaUKnyirTC
With a fixed 120 ms wait, a page whose full renders take 15–100 ms was held coarse for 120 ms
after a key zoom or a page switch, where before it was refined in the next frame. The wait is
now the last full render's time when that is shorter than `SETTLE_MS`: a wait never costs more
than the render it may save, and on a heavy page (seconds a render) it is 120 ms as before.
Also from review: the test that the wait counts from the preview no longer times a preview
render (a margin of a few milliseconds under load); a sleep between the request and the poll
stands in for the render, with the whole settle as margin. `Inline::in_flight` says that a
preview's wait counts.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Comments become the task queue for an attached or headless MCP client: comment.prompt returns a bounded Markdown brief, the Comments panel gains Copy for AI, and the MCP server exposes fix_comments, watch_comments, figcraft://comments and wait_for_changes. Mutating comment.* commands append to a bounded in-memory event log; the control channel's events.poll long-polls it without the UI thread, and a successful batch commits its events only if the whole batch lands. No model, network dependency, asset or file-format change.
From the review of #1468: a pan or an edit at the new zoom before its full render has run
renders from scratch too, so it shows a preview and restarts the wait instead of running a
blocking full render between it and the next step; once refined, requests at that zoom build
on the kept image and are refined at once. The new test pins both down. The status line of
`docs/gaps.md` now carries the date of the gap 7 entry.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The browser renders on the UI thread. A zoom showed its coarse preview and then, in the very
next frame, ran the full render, so every wheel step of a continuous zoom had a blocking full
render between it and the next: on a heavy page each one took seconds, and the zoom stuttered
for as long as it went on.
The browser's worker (`Inline`) now refines a preview only once no newer request has come for
`SETTLE_MS` (120 ms) after the preview was shown: the zoom has paused. Each wheel step still
shows its preview at once, and the worker asks for a repaint when the wait is over. Pans and
edits have no preview and are refined at once, as before. The desktop has a render thread of
its own and is unchanged. A zoom passing through a scale also no longer spends a full render
there.
The wait counts from the preview being handed over, not from the request: rendering the
preview can itself take longer than the wait (the page is blocked meanwhile, so no newer
request could have come), and counted from the request the full render would follow in the
very next frame, as before.
Measured in Chrome on a synthetic page of 20 images, 60 shadowed rectangles and two
background blurs (a full render takes about 3.7 s, a preview about 250 ms), 20 wheel steps in
and 20 out, 50 ms apart, three runs each: the frame interval's p95 went from 3.7 s to 240 ms,
the frames over 100 ms from 79–80 (one full render per step) to 39–41 (one preview per step),
and the zoom took 10–12 s instead of 141 s. The one full render after the pause still takes as
long as before.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Found while testing crates/engine/src/cmd/props.rs:623-634 (the "paddingHorizontal" and "paddingVertical" arms)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Translate 2508 interface entries from English, keep count-neutral plurals, and gate the web catalog on Arabic font readiness. Record AI assistance and pending native review.
AI-assisted-by: OpenAI Codex
Merging every region of one stroked shape is a no-op again: only fill
parts count, so a vector with several painted regions still merges.
The region paint import budget now covers the whole file instead of
resetting for each layer, so instances of a main with large region paint
stacks cannot multiply the work.
The Home tab started at x=8 and was 44 wide, so its divider sat at x=52 while the icon rail below is 56 wide. Without inset window buttons the Home tab now ends at the rail's width (shared RAIL_WIDTH constant), so the divider continues the rail's edge. The inset (macOS) title bar keeps its 44-wide Home tab.
"(as file.export)" read as all of them, but the PDF and XML options are
refused there, so it now says they are file.export's SVG options.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
file.placeImage runs image.place's code and file.exportSvg file.export's,
so their docs now list the names those read (width, height, target, mode;
scale or size and the export options). MCP's create_node passes points,
closed and select to node.create, and pointer and key read ctrl, so their
schemas list them too.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`file.open` on a path already open in a tab switched to it and returned
its index, even when another process had changed the file since, so the
caller kept looking at the old copy with nothing saying so (#997). The
tab's disk stamp from #1330 (`DocState::on_disk`) now tells:
- the result says `stale: true` when the file on disk differs from what
the tab read or last saved;
- `reload: true` reads the file again into the tab, as Revert to saved
does (refused while the tab has unsaved changes or is in a live
session), and the result says `reloaded: true`.
MCP's `open_file` takes `reload` and passes it on.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`run_command {"command": "inspect.code", "args": {...}}` ran the command
without its params, on the selection. MCP tool arguments are now checked
against each tool's input schema (and the objects it describes inside,
such as a batch's steps; open objects like `params` take anything), and
a step key other than `command`, `params` and `as` is refused in
command.batch, figcraft-cli scripts and the MCP batch tool.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A misspelt parameter name was ignored and the command reported success,
answering as if the parameter hadn't been given (`comment.list
{"filterr": "open"}` listed resolved threads). `Session::execute` now
refuses names the command's params doc doesn't list, naming them and
what the command takes, before the command runs, so every path into a
command (CLI, batches, control channel, MCP, plugins, the UI) gets the
check.
The names come from the existing docs, parsed once with the registry:
each top-level item's first word and the words joined to it by `|`, `/`
or `+`, bracketed text and what follows `:`, `=` or `→` being
description, alternative groups (`{x, y} | {points}`) included. `ids`
and `id` are always taken (`params::targets` reads either); node.set,
whose doc is its props list, keeps its own check.
The junk sweep runs each junk object as given and cut down to the names
a command takes, so the junk values still reach the parsers. Tests that
passed names their command never read (comment.resolve's author,
collab.host's link, node.get's depth) pass only what it takes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Before commands can refuse names their docs don't list (#1413), the docs
have to list what the code reads. These are the top-level names the run
code (and every helper it hands the params to) reads that its doc left
out, all checked against the code; nothing changes behaviour:
- creation: node.create (characters, points, closed), frame.create,
shape.*, text.create and vector.create (index, name, select),
image.place (name)
- aliases the code already took: comment.* (id | link), annotation.add
and annotation.edit (label | text), text.set (characters | text),
text.listStyle (style | list), fill.add and stroke.add (paint | color),
proto.interaction.add/set (durationMs | duration_ms, delayMs | delay_ms)
- options: file.export and export.update (contentsOnly,
includeBoundingBox, svgOutlineText, svgSimplifyStroke), comment.next
(frame, resolved), layout.suggest (direction), variable.create
(description), vector.edit.widthPoint (side), collab.share (key),
collab.manageFile (access, label, link, name), fill.pattern (visible,
blend)
- docs whose shape hid a name: proto.interaction.add (a stray `}` closed
the object before delayMs; dest was only in a trailer),
proto.interaction.set (patch fields in a second object), collab.join
(a top-level comma inside the link examples), pref.uiScale (commas in
the value's description)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`{"layoutMode": "NONE", "wrap": true}` was refused with a message
naming `layout`, a key the call didn't have. The error now names the
key the call used.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`system-ui`, `-apple-system`, `BlinkMacSystemFont` and
`.AppleSystemUIFont` join the names `apple_alias` resolves to the
installed System Font, as SF Pro and SF UI already do. SVG import keeps
a family the SVG lists when a font answers to its name, so on a Mac
`font-family="-apple-system, system-ui, sans-serif"` keeps
`-apple-system` and draws in System Font; elsewhere the generic
stand-ins apply as before.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`apply_order` gave `layout` the rank of every other key, so the props
map's alphabetical order decided: `gap` applied before an object
`layout`, which then replaced the whole auto layout (the gap went back
to its default), and `padding` applied after `layout: null` and turned
auto layout back on.
`layout`/`layoutMode` now apply right after the geometry keys and
before the keys that tune them, and auto layout keys beside a `layout`
of null or "none" are refused, naming them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The pre-signing check matched any folder named Program Files, so it rejected
a valid MSI when Windows resolves Program Files elsewhere, accepted the 64-bit
root for an x86 package and accepted a different app name. Build the expected
folder from the package architecture's resolved root and the policy name.
Shadows and layer blurs dominated PNG/JPG/WebP export time (gdb stack
sampling of the example files: about 45% in fx::box_blur_h, 13% in
fx::morph). Both now produce the same bytes, faster:
- fx::blur runs its three box passes per row while the row is in cache,
works on whole pixels instead of bounds-checked bytes, skips fully
transparent rows, transposes in cache-friendly column blocks, and
splits rows and transposes across rayon on desktop for pixmaps of
64K pixels and more, only when not already on a rayon worker (a band
of render_view stays on its thread: no nested parallelism).
- fx::morph (shadow spread on layers that aren't one outline) uses the
van Herk / Gil-Werman running max/min: constant work per pixel instead
of 2r + 1.
Integer arithmetic and pass order are unchanged; new tests compare both
against the previous implementations byte for byte, including sizes on
the parallel path.
An ignored timing test exports every top-level frame of examples/ as
PNG, JPG and WebP at 2x through export_file:
cargo test --release -p figcraft-engine --lib export_timings -- --ignored --nocapture
Ryzen, 16 threads, Linux; median of 7 (min):
PNG 2195 ms (2182) -> 992 ms (989) -55%
JPG 2880 ms (2560) -> 1550 ms (1535) -46%
WebP 1964 ms (1901) -> 1022 ms (1018) -48%
total 7035 ms (6677) -> 3567 ms (3555) -49%
All 80 exported files (PNG, JPG, WebP, PDF, SVG) are byte-identical.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Real msiexec runs (per-user and all-users, silent and wizard) found two defects
that the metadata fixtures cannot see:
- "Only for me" published ALLUSERS=2, so FindRelatedProducts treated the run as
per-machine and skipped the existing per-user copy. An interactive upgrade
left two registered products and uninstall left files behind. The page now
leaves ALLUSERS unset for a per-user choice.
- Windows Installer resolves Program Files once, for the context it started in,
so choosing "Everyone" in the wizard installed into the user profile with a
machine-wide registration. The wizard now sets the new-install folder itself
after the choice (Program Files, Program Files (x86) for 32-bit, or the
per-user Programs folder). Silent installs are unchanged.
check-install-policy.ps1 now fails if either regresses.
* feat(ui-egui): font family menu keeps its search field and grows back after a filter clears
Repin craft-libs to the shared dropdown (craft-ui), move the font family menu to
craft_ui::combo::Menu with the search in a fixed header, and use craft_ui::ComboBox
for the credits sort. clippy bans egui::ComboBox.
* chore(deps): craft-libs 60d60d48 (dropdown fixes)
* chore(deps): craft-libs 4d7b82f7 (dropdown fixes)
* chore(deps): craft-libs 92d392cf (#387 merged)
* fix(ui-egui): craft-ui on every target (wasm uses the shared ComboBox too)
Found while testing crates/engine/src/automation/params.rs:170 (color_of, Value::Array arm)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/engine/src/cmd/props.rs:476-479 (the "opacity" arm), with crates/engine/src/automation/params.rs:14-20 (f64_of)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A viewer moving the pointer sends presence about 30 times a second, and every message sent put
the next ping off by the keep-alive interval, so no ping went out while presence did. The host
answers pings and nothing else to a client alone in a session (presence goes to the other peers),
so nothing was heard for the idle timeout (45 s): the client dropped a healthy connection and
reconnected, over and over. Seen in the browser build; the desktop app's guest has the same
timing (its reader times out the same way) and gets the same change.
Pings are now timed from the last ping, not from the last message sent: one goes out every 15 s
whatever else is on the wire. Regression test: 90 s of presence at 10 a second with the host
answering only the pings keeps the connection (it closed before the change).
Co-Authored-By: Claude Fable 5.1 <[email protected]>
node.create and node.set take {"$var": variable id or name} for fill and stroke: an exact id
first, then an unambiguous name or Collection/name. The binding goes through the same code as
variable.bind, is applied after literal values and style links, replaces other bindings of that
paint's colour and clears its style link, all in one undo step.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
Session::grouped saved only the undo stack's length, so a failed group could drop the entries it
added but not bring back the ones an edit.undo inside it popped. It now keeps each document's undo
entries on entry, as it already did for redo, and restores them when the group fails.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
node.get, the control channel's inspect methods and the MCP get_node / inspect_document tools
take fields: [node keys], applied at every level, or outline: true (id, name, type, characters
cut to 120 characters, mainComponentName, visible). Ids and structure are always kept, properties
that aren't selected aren't computed, and the default output is unchanged. Part of #996.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
Found while testing crates/doc/src/lib.rs:64-71 (impl FromStr for NodeId), reached through crates/engine/src/automation/params.rs:78 (id_of)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/vectordrawable/src/xml.rs:22 (num), reached via stroke_attrs at :189
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A density now counts only within 0.005 of a whole 2x-4x multiple of
72 dpi. That covers the rounding of pixels per metre (144 dpi is stored
as 5669 or 5670) and rejects nearby resolutions: 141, 143 and 145 dpi
were read as 2x under the old 0.05.
Paste falls back to the decoded clipboard image when the clipboard's PNG
can't be read, instead of pasting nothing.
Also the line on what the objc2 crates are for, and the ROADMAP row for
the new placing default.
A placed image was always fitted into 1000 px on its longer side, so a
screenshot of a 1440 px wide page came in at 1000 px and stopped lining
up with the design it was taken from. Figma places an image at its pixel
size and only scales down a side past 4096 px.
Placing now does the same, and reads the screen density a PNG records:
macOS screenshots carry 144 dpi (pHYs), so an 800 x 600 px capture of a
400 x 300 pt area comes in at 400 x 300, as in Figma. Only whole 2x-4x
multiples of 72 dpi count as a screen density; 300 dpi print scans and
other values keep their pixel size. The image data stays at full
resolution, so exports at 2x and up stay sharp.
On macOS, pasting reads the clipboard's public.png as is instead of the
decoded pixels arboard returns, which drop the density (and the colour
profile). Other platforms keep the decoded image.
Files dropped, Place image, paste and image.place all go through the
same path; an explicit width/height still wins.
Lowercasing a name ending in Σ gives ς, while a query Σ on its own lowercases to σ, so the
substring check missed it (CodeRabbit). Both sides now fold ς to σ.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
command.batch and the MCP batch tool take onError: "abort" (the default, unchanged) or
"continue". In continue mode each step runs inside a rollback-only checkpoint split out of
Session::grouped: a failed step is rolled back on its own, later steps still run, and the result
is one {step, command, ok, result|error} object per step. What was applied is one undo step; a
reference to a failed step fails too; malformed batches are still rejected before any step runs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
node.find {root?, name?, type?, text?, component?, limit?} returns {ids, total} for the strict
descendants of root (or the current page) that match every filter given: name, text and main
component name as case-insensitive substrings, type as an exact type name. It changes neither the
selection, the page nor the undo history. Part of #996; fields/outline are left for later.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
The attached backend read each reply from the app's control port with read_line, which has no
limit. It now reads up to 512 MiB before the newline (an app window capture is at most 256 MiB as
base64) and stops at the first byte past that. An oversized or non-UTF-8 reply drops the
connection and is reported like a lost reply; the next call reconnects.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Co-Authored-By: GPT-6.1 Sol <[email protected]>
Co-Authored-By: GPT-6 Astra <[email protected]>
Brings #1210 onto main as it stands after #1019 and #1209, and keeps both:
- #1019's flyout caret keeps its 16 pt target, 2 pt clear of its tool (#988).
- #1209's narrow-window handling stays as it landed: the pill is never wider
than the canvas less a margin and its buttons scroll inside it under a drag
or the wheel, the help button lifts above it when they'd overlap, and the
pill still clips to itself rather than floating over the panels.
The design itself, as #1210 proposed it:
- One grid for every state of the pill: 32 pt cells, 8 pt padding, 8 pt between
top-level items (the divider included), a 1 pt divider as tall as the content,
16 pt carets, 14 pt mode marks, and a 32 pt mode strip whose current mode is a
plain inset with no outline.
- The Resources button goes; the left rail's Assets tab, Shift+I, Alt+2 and
View > Panels > Assets remain the ways into the Assets panel (and the locale
key, its test and the parity notes go with it).
- The vector edit and shape builder pills move onto the same grid, so a pill
never changes its spacing as the tool changes.
main_width is now the same walk the pill draws, and a test walks it for the
design and board toolbars: the last item ends exactly PAD from the right edge.
That caught the board pill asking for 8 pt more than it drew.
Split out of #932 (the 2 px outlines and missing frame radius handles there
move the canvas away from the reference tool, but the constraint-guides part
is good on its own — bflatastic, #issuecomment-6420566791). When a layer inside
a non-auto-layout frame is selected, blue dashed guides show how it is pinned:
a left/right/top/bottom constraint draws a line from the layer's edge midpoint
to the frame edge, and both-axis constraints draw both. A centre constraint
marks the layer itself with a short solid bar through its middle — horizontal,
vertical, or a cross when centred on both axes — rather than a guide reaching
the frame's centre.
main split `codegen.rs` into `dev/codegen.rs` (the shell: `CodeLang`,
`code_for`, `Ctx`, the shared helpers) and one module per native target
(`css`, `html`, `swift`, `compose`, `arkui`), and added Vue and ArkUI from
other PRs. Flutter arrives as the tenth target, in its own module:
`dev/codegen/flutter.rs`, wired from `codegen.rs` (`CodeLang::Flutter`, `dart`
as an alias, rem left to the web targets), with `flutter_definition` /
`flutter_type` / `flutter_default` beside main's `arkui_definition`.
Every language list in the tree is a union of the two sides now: the
`inspect.code` and `inspect.componentCode` specs, `codeConnect` (parse, import
line, expression, template, path checks, messages), the playground CSS/HTML
assets, the 21 UI catalogues (main added eight after this branch: bn, fi, it,
ka, nl, pl, th, vi, each with Flutter in its own conjunction) and the docs.
Flutter is an extra code target; nothing main gained is dropped.
The reviewer's absolute-child bug (#1382-1403: a child that ignores its auto
layout parent fell through to `Place::Flow`) is fixed in `flutter_children`:
flow and absolute children are split, the absolute ones go in a `Stack` around
the Row/Column at the parent's own box (so their offsets stay the parent's and
the padding moves onto the line instead of insetting them), the line fills that
Stack (`Positioned.fill`, so `Expanded` still has a bounded width), and only
flow children now decide `CrossAxisAlignment.stretch` and `Expanded` — the
absolute child no longer joins the line's spacing or loses its offsets.
`flutter_lifts_a_child_that_ignores_its_auto_layout_parent_out_of_the_line`
covers both kinds of child in one auto-layout frame: a 50×20 flow child and a
30×30 absolute one asking for `fill` on both axes, in a parent with 4 pt
padding. It checks the offset, that the absolute child is out of the Row, that
no `Expanded`/stretch comes from it, and that the padding insets the line only.
ROADMAP and dev-mode-parity keep main's header (2026-10-26) and take this
branch's history row in date order (2026-10-16).
`cargo test`: engine 1391 pass, ui-egui 814 + 8 + 1 pass; fmt and clippy clean.
The one engine failure, `board_tables_tests::a_double_click…`, fails the same
way on a pristine `origin/main` here: this sandbox has no `craft-fonts`
checkout, so its text doesn't wrap and the row can't grow.
Found while testing crates/engine/src/cmd/props.rs:165-175 (four, used at :506, :540 and :613)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/svg/src/import/mod.rs:723-740 (looks_like_svg, the `<!` arm using s.find('>'))
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
main moved the colour picker into `widgets/colorpicker.rs` and gave it a fixed
area id (#976: `window_id()`), which the style picker looks up to tell a press
inside the picker from one outside. This merge keeps that id as the picker's
own area: the popover takes it as-is (`Popover::with_id`), so the lookup and the
#237 regression test keep working — the test now finds the picker by the id
instead of by its z-order.
The rest is the popover arriving on top of current main:
- the colour picker's `egui::Window` is the shared card: its 40 pt header —
title 16 in, X 8 in from the right — is the popover's, and its body is
unchanged, eyedropper included;
- the style picker's own area, frame and border are the same card on the dark
surface; its title moved into the header, its tools row stays;
- presses in the Design panel and on the control that opened the picker are
held out of the closing, so that control's own click toggles the picker;
- while the eyedropper samples (still picking, or a colour just picked) a press
on the canvas is the pick, so it does not close the picker it picks for.
`right.rs` was split by main: its four swatch call sites are now in
`panels/right.rs` (page background), `panels/right/paints.rs` (a text range's
fill, a paint row) and `panels/right/effects.rs` (an effect's colour). egui 0.37
deprecates `Id::new` for `Id::unique`, which the new module uses.
`cargo test -p figcraft-ui-egui`: 824 + 8 + 1 pass; fmt and clippy clean.
main split `right.rs`, `typography.rs` and `layout_panel.rs` into `panels/`
(the right panel's sections in `panels/right/{paints,stroke,effects,inspect,menu}`,
the Type section in `panels/typography/`). This merges that refactor with the
panel grid, so the grid's numbers now live where the code does:
- rows inset 16 from the left and 8 from the right (`block`, `sublabels`);
- two 88 pt columns with 8 pt gutters and a 24 pt button column at a row's end
(`col_width`, the stroke's align/weight pair, the effect and grid rows);
- a row's two buttons are one group on the right edge, 24 each, 4 apart
(`widgets::tight_end` in the paint, effect and grid rows);
- section headers: the last icon ends 8 pt from the edge, icons 28 apart;
- the right panel's frame keeps no stroke (a stroke counts as margin and took
1 pt off each side); its edge is drawn in `right_panel`.
Widget text stays at 11 pt, the size the app we match uses: the merge keeps
main's type throughout (section titles 11 pt, now regular; captions 10 pt) and
changes geometry only. The Type section's size field follows the new column.
`cargo test -p figcraft-ui-egui`: 814 + 8 + 1 pass; fmt and clippy clean.
The backward search for a safe break started at line_start + 2, so in a
three-letter run like กเม it skipped the only safe boundary and left the
line ending in a leading vowel.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Thai, Lao, Khmer and Burmese write no spaces, so a paragraph has no break
opportunity and wraps where it overflows. That emergency break could start a
line with a letter's trailing mark or end one with a leading vowel, splitting
a syllable (and, when the break fell between the glyphs of one cluster, drawing
it twice).
- The break backs up to a boundary that keeps a mark with its letter and a
leading vowel with the consonant after it; when the line's first character is
in one, it carries the pair over to the next line instead. Lao leading vowels
count too.
- It only breaks between grapheme clusters now, never between glyphs of one.
- Tests: a font-free unit test for the boundary rule, and a Thai wrapping test
over 39 widths (skips without a Thai face).
Co-Authored-By: DeepSeek V4.1 Flash (opencode) <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/format/src/sketch/mod.rs:241-249 (Archive::image, MAX_IMAGE at :69), called once per distinct imageRef from crates/format/src/rest/mod.rs:951 (Ctx::image_key)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Retain regional paint stacks and coordinate frames through simplify, offset, flatten, bucket normalization and Shape Builder. Preserve legacy fill semantics and synchronize importer warning catalogs. Verified with cargo xtask ci on Windows and Ubuntu/WSL, native UI screenshots, and CLI round-trip/export checks.
From a code review, each behaviour question checked on Sketch's own renders of
files built for it:
- background and motion blur radii are standard deviations too (doubled, like
Gaussian ones); zoom blurs keep half their radius
- colour adjustments clamp FigCraft's contrast before working out the exposure,
so the slope stays exact past FigCraft's ranges (brightness below -0.5 bent
the whole curve)
- an image override applies after a layer style override on the same shape
- a bitmap's fills (its image under the style's fills) aren't linked to the
shared fill style any more
- one read of a gradient's colorInterpolation; a dead stack condition removed
Checked and unchanged: symbols of a Sketch 63 file don't clip either, and
Sketch mixes OKLab gradients' alpha as the importer does.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Each rule was checked against Sketch's own renders (sketchtool) of files built
for the case, and on real Sketch 50, 63 and 2026 files:
- instances show a master's background only when it is included in both
instances and exports
- symbols clip their layers only with clippingBehavior 1
- group and instance fills are tints only from Sketch 64 (format > 123)
- an outline mask's opacity fades the mask and what it clips together
- a bitmap's image lies under its style's fills
- image overrides reach a shape's top image fill
- stretching stacks fill their items across
- a layer blur's radius is Sketch's standard deviation (FigCraft radius x2)
- colour adjustments: contrast, then brightness added, mapped exactly onto
exposure and contrast within FigCraft's ranges
- gradients mixed in OKLab or OKLCH become sRGB stops along the same colours
(with an import note)
- auto-width texts an override made longer grow from their aligned edge
- shape group operands lose their own borders (since #1184 a stroke-only
operand gives its stroke band, not its area)
Known limitations are listed in the file-format parity row.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- The last line shown runs on into the word wrapped after it, so the ellipsis
fills the line ("Lorem consec…") instead of following the last whole word.
- An explicit height (node.set, transform.resize, a resize handle, fixed
sizing) on truncating text keeps it truncating (TextData::fix_size).
- Max lines on auto-height text hugs up to that many lines and ends in an
ellipsis; text.maxLines keeps auto height, Type settings show it as
truncating, .fig export writes textTruncation and import keeps auto height.
- node.get reports maxLines; text.fit measures every line.
Co-authored-by: Cursor <[email protected]>
Open and Revert to saved recorded the file's size and modification time after reading it: a change written in between would be stamped as already loaded and a later save would replace it. Taking the stamp first makes such a change refuse the next save instead (CodeRabbit on #1374).
With a file open in the desktop app, figcraft-cli could change it on disk;
the next Save (or auto-save) wrote the app's copy over it and said
"Saved", losing the CLI's changes.
Each tab now remembers its file's size and modification time when it
opens, reverts or saves it (the Stamp library files already use), and
Save and auto-save refuse while the file on disk differs, pointing to
Revert to saved or Save as. Scripts pass `file.save {overwrite: true}`.
Save as and saves to an explicit path are unchanged.
On a main component's Properties row, the default of an instance-swap
property opened a menu of every component in the file at full height, so
in a file with many components (an icon library) the bottom ones ran off
the screen and couldn't be picked. The menu now scrolls past 320 pt like
the instance's swap menu, is at least 200 pt wide, ticks the current
default and is named for assistive technology.
Right-clicking an instance always showed "Create component" (and never "Go to
main component"), so an imported instance looked disconnected from its main
even though the link was intact. When the selection is an instance, replace
"Create component" with "Go to main component" in the context menu, matching
Figma. A regression test covers the menu contents for an instance.
Co-Authored-By: Claude <[email protected]>
Zoomed in on an image, the canvas smoothed it with bicubic sampling, so
a screenshot under inspection went soft at every edge and its pixels
could not be told apart. Checking a screenshot of a live page against
its design up close needs the pixels as they are.
The canvas now samples an image paint at the nearest pixel once each
image pixel covers two or more device pixels. Below that the image stays
smooth: between 1x and 2x nearest sampling draws pixels one and two
device pixels wide side by side, which looks jagged rather than sharp.
Exports, thumbnails, PDF and SVG are unchanged: the switch is a render
option, magnified_image_pixels, that only the canvas sets, and the
retained renderer reuses a render only when it matches. Pixel preview
was not a fit: it rasterises the whole canvas at 1x or 2x, so vectors
and text go blocky along with the image.
Branches of a file in a live session: create, edit, review, update from
main and merge while the file is in a session. The branch stays with its
author and makes ids in their own client space far above the main file's;
merging is one synced edit and one undo step; peers' edits merge against
the branch's own base and conflicts resolve as offline; a view link still
refuses merging; a branch made outside the session that adds layers is
refused with a reason. Branch history entries for creation and updates
from main. Fuzz and junk tests extended to export settings, guides and the
live case. Docs: gap 17, collaboration parity, ROADMAP, mcp, protocol.
* M8: widgets take text input and a property menu
Plugin API 4 (gap 6). A widget's text layer can be an editable field:
a double-click edits it like any text, and when editing ends the value
(cleaned, at most 1,000 characters) goes into the widget's state under
the field's name and its plugin redraws, as one undo step. A widget can
also carry a property menu of up to 8 actions, toggles and colour
choices, shown in a small bar over it while it's selected.
- widget.setText and widget.menu drive the same paths from scripts,
the CLI and MCP; widget.insert/list also report fields and menu
- descriptions, values and stored menus are validated as untrusted;
an API 3 plugin drawing a field or menu is refused, older hosts
refuse API 4 manifests; no new host imports
- SDK reads the new events; the Counter example gains a label field
and a menu (reset, steps of ten, a colour)
- tests for fields, menus, canvas editing, rollback, forged data and
live sync; junk sweep extended; labels in every command catalog
- docs: plugins, mcp, gaps, target-app-parity, ROADMAP, roadmap
* M8: widget property menu on egui 0.37
Id::new and Ui::id are deprecated in egui 0.37: the widget bar's area uses Id::unique and its buttons' interaction ids come from ui.scope_id().
* M11: a double-click a board object or a widget field takes, in a function of its own (Sonar)
Found while testing crates/engine/src/codegen.rs:122-150 (lengths_to_rem calls num, which rounds to 2 decimals, at :112)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The Share popover's server key was stored without its server, so changing the
server in Preferences would send the previous server's key to the new one.
Remember the key with the server it was typed for and only send it when the
addresses match. A regression test covers the mismatch.
Co-Authored-By: Claude <[email protected]>
A publish uploads a whole file; a large project on a slow link takes the
server a while to receive and answer. But the server's one-shot request
connections (manage, hosting, library, plugins) used the short request_idle
as their read timeout, so a slow upload could be cut off mid-processing. The
manage/hosting connections that carry uploads now wait the longer
request_reply_timeout (as the client already did), keeping the short idle
for the other one-shot requests and sessions.
Co-Authored-By: Claude <[email protected]>
* M12: board templates and tables
Five built-in board templates (retrospective, brainstorm, weekly planner,
flow chart, kanban board) placed as ordinary board objects in one undo step
(board.template, board.templates) from a toolbar picker with pictures, and
board tables made of auto layout frames (board.table, board.tableEdit; up to
50 x 20), typed into with a double-click and edited by row and column from
the bar over the selection. Docs: boards-parity, gap 24, ROADMAP, mcp.
* M12: board templates picker on egui 0.37
Id::new and Ui::id are deprecated in egui 0.37: the templates picker's area uses Id::unique, its buttons' interaction ids come from ui.scope_id(), and its test reads the board bar's area by Id::unique.
* M12: board timer and voting
A countdown everyone on a board sees (board.timer, board.timerState), synced as its end time or the
time left while paused, and voting sessions (board.vote, board.voteResults): votes per person, on
everything, stickies or the selection; clicks vote, your badge takes a vote back; your own votes
show while it runs, everyone's with a ranked list once it ends. Timer and vote sync as document
properties and ballots as one entry per voter; neither is an undo step. A pill at the top right
of the board, new Timer, Pause and Vote icons, labels and strings in every catalog, docs.
* M12: board timer and voting on egui 0.37
Id::new is deprecated in egui 0.37: the timer and voting area and its memory ids use Id::unique.
* M12: board.vote's start and ballot changes in functions of their own (Sonar)
* M12: boards dot grid, re-attaching connector ends, curved connectors
Board pages draw a dot grid in the canvas render, after the page colour
and under the layers (figcraft_render::Dots), thinned out when zoomed out
and coloured from the theme's ink against the page. A selected connector
shows its ends as handles; dragging one onto an object re-attaches it,
onto empty canvas leaves it free (board.reattach, one undo step, also
over MCP). Connectors can be curved, and elbows go around the objects
they join when the short way would cross them.
* M12: boards dot grid: blending a dot in a function of its own (Sonar)
Found while testing crates/color/src/lib.rs:198 (Color::to_css)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Review on #1309 pointed out that core-text 20.1's get_descriptors retains
the array it gets under the create rule a second time, so every scan kept
one array of font descriptors alive. Checking that turned up a larger leak:
Core Text autoreleases objects while it builds the collection, and the scan
runs where no autorelease pool is active (the app's background scan
thread, the CLI, the MCP server), so nothing ever drained them.
Measured with `figcraft-cli run` of N `text.rescanFonts` steps on macOS 15
(peak memory footprint):
| | 10 rescans | 200 rescans |
|---|---|---|
| main, without this PR | 221 MB | 220 MB |
| this PR before the fix | 228 MB | 796 MB |
| with the fix | 226 MB | 225 MB |
A standalone Swift loop over the same Core Text calls grows from 9 MB to
759 MB over 200 scans without a pool and stays at 20 MB inside one, so the
pool is what matters; the core-text bump removes the over-retain on top.
- The query runs inside objc2::rc::autoreleasepool. objc2 0.6 is already
in the lockfile (winit's), and the function is safe, so the workspace's
unsafe_code = "forbid" still holds.
- core-text goes to =22.0.0, the first release that takes ownership of the
array correctly. It brings core-foundation 0.10, core-graphics 0.25 and
core-graphics-types 0.2 beside the older versions other crates use. 22.0
moved framework linking behind a `link` feature that is on by default,
so with default features off it is named explicitly.
A rescan takes about 0.2 s instead of 0.08 s on that Mac (266 more font
files are indexed); it only runs when asked for.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
(cherry picked from commit 7f6a3a4e46922d737b8053c7cd5968a02a558ae3)
Font managers such as RightFont, FontBase and Typeface keep fonts in their
own library folders and activate them through Core Text, without copying
them into /Library/Fonts or ~/Library/Fonts. Every other Mac app lists
these fonts; FigCraft walks only the font folders, so it didn't, and a
file using one of them drew in a fallback font.
On macOS the scan now also asks Core Text for the file of every available
font and indexes the files that lie outside the walked folders. All three
scans (the first, the on-demand one and text.rescanFonts) share this, so a
font activated while the app runs is found by a rescan like a newly
installed one.
Core Text needs unsafe FFI, which the workspace forbids, so the query goes
through the safe wrappers of the core-text crate, pinned at =20.1.0. That
version builds on the core-foundation 0.9, core-graphics 0.23 and
foreign-types 0.5 the lockfile already has: core-text is the only new
package. The call sits inside catch_unwind, so anything unexpected from
Core Text leaves the scan with the font folders alone. Other platforms are
unchanged.
On a Mac with RightFont, 15 families from its library are now listed
(Geist, JetBrains Mono, Clash Display and others) that weren't before. A
full font listing went from 0.06 s to 0.20 s on that machine.
Tests: the filter that keeps files outside the font folders once, indexing
a font file found outside them, and (macOS only) that Core Text lists the
system's font files. cargo test --workspace, cargo clippy -p figcraft-text
--all-targets -D warnings and cargo fmt --check pass on macOS 15. CI
builds on Linux only, so the macOS path is checked locally.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
(cherry picked from commit 5699e7abe6e5994e526051774ec92aadddb0550c)
class_name runs layer names through deunicode (new dependency: pure Rust
tables, zero deps, BSD-3-Clause, the slug crate's base) before its ASCII
filter, so Korean, Cyrillic, Latin-diacritic and CJK names become
readable ASCII classes in HTML/CSS/React/Leptos output, component roots
and data-* variant attributes: 로그인 화면 → rogeuin-hwamyeon, кнопка →
knopka, über → uber, 按钮 → an-niu (before, every one came out layer-,
layer--2, ...). Symbols-only names take the node's kind (.text, .frame)
instead of layer-, the 40-character cut lands on a word boundary, and a
leading digit keeps the layer- prefix. SwiftUI/Compose type names are
unchanged (their own sanitizer); ASCII-only names are byte-identical.
html_codegen_tests updated and extended.
Existing text keeps the named instance's optical size, so documents render as before; the
Sketch import stores the size as the text's opsz value instead.
The paragraph above the ROADMAP stage line moves into the revision history; docs/server-memory-retention.md
is folded into the compaction bullet of docs/server.md.
One side_strokes helper (square or rounded, as #1044 paints them) serves stroke_area and the boolean
operand outline, instead of a second, square-only copy; test for a rounded per-side border.
CI unpacks a coverage archive containing a build-generated locales.rs (the
i18n table from crates/ui-egui/build.rs) into coverage-archive/. Scanning it
made almost all new code duplicated (new_duplicated_lines_density 67.8%).
Each logical section of the text style editor (sample, font menu, style and
size row, line height and letter spacing, and the two menus) is its own
helper. UI calls keep their order; no behaviour change.
A controller button pressed while the player has no frame to show,
for example after the frame is deleted, no longer plays the
prototype's gamepad trigger once the frame comes back.
Two gaps CodeRabbit found in the guard:
- `same_file` compared paths, plain and resolved. A second hard link to
the file is a different path to the same data, so `convert`,
`run --out` and Save As to the link passed the guard and truncated
the file. It now asks the file system (same-file: device and inode on
Unix, volume and file index on Windows). same-file was already in the
lock through walkdir.
- `run --in X --out X` checked the tab the script left active. A script
that ends on `file.duplicate` or `file.new` leaves a tab with no count
of images left in X, so X was written over. The count is now taken
when X is opened and checked against `--out` whatever tab is active;
the active tab's own file is still checked as before.
The script check is a separate function so it can be tested without a
file past the 4 GiB budget.
A native file whose images together pass IMAGE_BUDGET (4 GiB, or 20,000
images) opens without the images past it. They are intact in the file, but
saving it again wrote the archive from the images in memory, so every
save, autosave or `convert` onto itself deleted them for good. A 7.6 GiB
`.fig` converted to a 7.3 GiB `.figcraft` reopened with 1,312 of its 2,337
images: one save would have dropped the other 1,025.
Loading now tells images left out for the budget apart from damaged ones
(`Loaded::images_past_budget`). A tab opened or reverted from such a file
remembers the file, and saving over it is refused with a message that says
how many images would be lost and to save under another name; saving a copy
elsewhere works, and the original keeps its images. Auto-save skips the
tab instead of retrying and failing every minute. `figcraft-cli convert`
and `run --in X --out X` refuse to write over their input the same way.
The same file is matched by its resolved path, so `./a.figcraft` and
`a.figcraft` count as one (`file_access::same_file`).
The note for each image left out for the budget now says it stays in the
file, which can't be saved over without losing it. Damaged and missing
images keep the old wording: they can't be kept, and saving drops them as
before. The doc comment on report_lost_images said saving always drops
the images it reports, which is no longer true for budget ones, so it went.
This stops the data loss; it doesn't make such files editable in place.
Loading images on demand would, and is discussed in the issue this links.
Tests: the count and the note for both budgets; saving over the file
refused with the file unchanged, Save As then working, and saving back over
the original refused again; a snapshot encoded in the background refused
too; a tab without images left out saving as before. Checked by hand on the
7.3 GiB file with file.save, file.saveAs, convert and run --out.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The PNG row of docs/file-format-parity.md listed the export settings as
"contents only" and "include bounding box (the layer's own box)". Since
#1307 the Export section labels the first option Ignore overlapping
layers and shows Include bounding box for text layers. The row now uses
those names, says that ignore overlapping layers is on by default, and
gives include bounding box for text layers, as the help page lists
them. The status line gets the new date and topic, and the revision
history gets a row; the percentages and hours are unchanged.
Picks up the startup prompt without the glyphs a font may lack (storytold/craft-libs#343, #344): the close mark is painted and the version pill is ASCII, so apps that panic on a missing glyph no longer take the prompt down.
Part of tracking issue storytold/craft-libs#319
Picks up the update prompt fix that draws its close mark and version pill
without glyphs a font may lack, so the prompt no longer takes the app down
when a bundled face has no ✕ or →.
Part of tracking issue storytold/craft-libs#319🤖 Generated with Codebuff
Co-authored-by: Codebuff <[email protected]>
A controller button pressed while the notice before a website link
shows no longer plays the prototype's gamepad triggers, as keys do not.
Buttons play again once the notice closes.
* feat: startup update prompt modal and direct self-updating
The update flow (check at start or on request, the startup prompt, the self-update worker) is craft-update's Updater; the app only supplies the suite, its close path and its status window. The check at start is opt-in for the desktop app, an install holds back while documents are unsaved, and the Windows update helper runs before argument parsing.
Ref: storytold/craft-libs#319
* fix(updates): re-add the update hooks to the 0.37 frame loop
* fix(updates): address review feedback
Wake the UI when a manual check is queued from a menu callback; the update prompt owns the keyboard while it is open (LightCraft); Download opens through the desktop URL hook (FilmCraft); the Help menu dispatches through menus::invoke (FigCraft); print jobs and comment drafts hold an install back (WordCraft); Check for Updates is a registered command (GridCraft); an install holds back while documents are unsaved.
* fix(updates): show the prompt on Home and presenter frames and close the application, not one window
* fix(updates): add the update strings to the English catalog
* E21: require release font coverage and stabilize capture deadline tests
* E21: separate capture, IME and field-save regression phases
* M0: correct the migration qualification reference
* M-egui037: document input lifecycle and UI regression contracts
* M0: correct egui 0.37 status docs and stabilize join-code regression
---------
Co-authored-by: Dylan M. Taylor <[email protected]>
Co-authored-by: Codebuff <[email protected]>
Reconcile upstream: keep both roadmap entries, main's Copy as submenu marker and
copy-link toast, and the migration's explicit join-order switch in the live
session test.
Generated with Codebuff
Co-Authored-By: Codebuff <[email protected]>
Found while testing crates/engine/src/cmd/arrange.rs:~560 (scale_node, the Vector branch calling VectorNetwork::scale); crates/doc/src/vector.rs:631 (VectorNetwork::scale)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The 0.11.1 release's Web job failed: the wasm was 25,248,590 bytes, over packaging/web/package.sh's 25,165,824-byte limit. opt-level z brings it to 23,306,645 bytes (measured locally with the release's craft-fonts ref, wasm-bindgen 0.2.129 and wasm-opt -Oz).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
import-cert.sh deleted its whole private temp directory on exit, and the keychain lives there, so codesign failed with "The specified item could not be found in the keychain" in the release build.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/ui-egui/src/field_math.rs:105 (Parser::factor, unit loop; end() at :109)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A .sketch file opened from disk joins the recent files, where it had no
picture and a generic icon. FileKind::Sketch reads the preview.png Sketch
saves under previews/ (within the thumbnail caps) and gets the design-file
mark, like .fig. Test: sketch_files_show_their_preview_png.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The command parameter docs (CLI `commands`, MCP list_commands, the Actions
palette) and the MCP open_file tool description listed .fig but not
.sketch; docs/mcp.md and docs/cli.md already did.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Linux: the shared-mime-info package declares application/x-sketch (a zip,
*.sketch) and the desktop entry and AppStream metadata list it. Windows:
the installer registers a Figcraft.Sketch ProgId under .sketch's
OpenWithProgids and .sketch in figcraft.exe's SupportedTypes, never as
the default. Both hand the path to file.open, as macOS does.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Info.plist declares Sketch's public type identifier (imported, so it is
known on a Mac without Sketch) and lists it as an Alternate, Viewer
document type: Sketch stays the default, FigCraft is offered. The app
already opened .sketch from an open-documents Apple event via file.open.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Conflicts were only in the Change notes of docs/file-format-parity.md and
docs/target-app-parity.md: main's notes kept, the `.sketch` note added.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/pathops/src/points.rs:89-92 (Styled::arrowheads)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/engine/src/component_codegen.rs:~226 (RESERVED list used by ident at :346)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/svg/src/lib.rs:97 (num, used by matrix at :129 for the image pattern's patternTransform at :243-254)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
* Prototype player: no clicks reach the prototype behind the link notice
* Selection colors: scopes filter the variable menu; the command binds as variable.bind does
* Style picker: color variables scoped to every selected layer's fill or stroke
* Corner button: named Corner smoothing on polygons and stars
* docs/gaps.md: known gaps from the #1311 review
* cargo fmt
* Style picker: clippy
Found while testing crates/engine/src/cmd/arrange.rs:~495 (scale_node, the Text branch). Compare crates/layout/src/lib.rs:136, which calls TextPath::scale on resize.
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
On Windows a socket's read timeout counts whole milliseconds, so a read
bounded by the hello deadline can time out a moment before it.
`next_message` then took the timeout for a closed connection: the client
got no "no hello in time" refusal, and the socket closed only after the
2 s linger. `Duplex::read` now reads again until the deadline has passed.
Tests:
- `connections_without_a_hello_are_told_why_they_end` (new): twelve
connections that never say hello each get the refusal. Failed 5 of 10
runs without the fix.
- `pre_hello_lifetime` notes the end of the stream on a thread of its own
instead of reading one message per 300 ms poke, where pongs queued
ahead of the close delayed it.
- `quiet_guests_stay_and_silent_connections_are_dropped`: the guest takes
only the host's keepalive. Its 600 ms idle timeout also bounded the wait
for the welcome.
- `two_files_are_hosted_and_joined_independently`: the two joins race, so
the file whose welcome arrived last is in front. The test checks that,
then brings B to the front.
- `a_local_share_is_joined_by_its_code_and_edits_sync`: once the share
stops, another program's host can take the freed port and refuse the
code.
* security: linear vector import, stroking and rounding, validation and a cycle check for dev snapshots
Closing a subpath, rounding corners and chaining stroke runs each rescanned the whole
network per step, and Document::validate searched a parent's children per child. They now
update only the current subpath's last segment, rebuild each region loop in one pass, walk
per-vertex adjacency lists, and check edges against a once-built index. A dev snapshot whose
layers contain a cycle under the root is refused instead of reconstructed.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
* fix(doc): address review findings for snapshot children and 2-segment rounded loops
---------
Co-authored-by: Claude Sonnet 5.5 <[email protected]>
The export setting's advanced options name the contentsOnly option
"Ignore overlapping layers" and offer Include bounding box only when
every selected layer is a text layer. A setting that already has Include
bounding box on still shows the option, so it can be turned off, and the
export still applies it. The saved keys and the contentsOnly and
includeBoundingBox params of export.add, export.update and file.export
are unchanged.
Every catalog's "Contents only" row becomes an "Ignore overlapping
layers" row.
A slice exported every layer of its page that showed in its area, whatever
its Contents only setting said. With Contents only on (the default), a slice
inside a frame or group now exports only that frame or group and the layers
in it. With it off, or for a slice on the page or in a section, the slice
exports everything in its area as before. The Export section's preview of a
slice follows its first export setting in the same way.
The player opened a prototype's Open link at once, also from an After
delay trigger with no click. The help center's Prototype actions page
says that the viewer is told they are leaving before an external
website opens, with a checkbox that skips this page on the current
device.
An Open link to an http or https address now shows a notice with the
link, Open link, Cancel and "Don't show this again". Open link opens
it; Cancel, the close button or Escape drops it. While the notice
shows, the prototype takes no keys, and the presenter's C, Shift+C and
F keys do nothing. When Open link is pressed with the box ticked, the
desktop app saves the choice with its preferences; the web build keeps
it until the page reloads. proto.linkNotice {show} turns the notice on
or off. A mail link opens at once, as before, also while the notice
shows. A second link to a website that fires while the notice shows is
dropped.
The bindings that move with the path's fills and effects are now found with list_binding, which reads `fills` as the first fill's color, as style_sets in cmd/styles.rs does.
The hover search now recurses into each child first and looks up its mask only when the child added a candidate, as layers_at in crates/doc/src/geometry.rs does.
A path linked to an effect style with no effects gave the text an empty effect list and the style link. The text now takes the path's effects only when there are any.
text.onPath, and the Design panel's Put text on path button that runs
it, kept the text layer's own fill and effects when the text went onto
a path. The help center's Guide to text says that the vector object's
fill and effects are transferred to the text layer.
The text now takes the path layer's fills, with their fill style and
variable bindings, when one of them shows, and the path layer's effects,
with their effect style and bindings, when it has any. The text's own
bindings on those fields are dropped. When the path has no visible fill
or no effects, the text keeps its own.
A server with a key rejects publishes without one, but Preferences' server
address never holds a key. Add a "Server key" field to the Share popover's
Advanced section and pass it to collab.share so a keyed server can be used
from the popover without the Publish dialog.
Co-Authored-By: Claude <[email protected]>
Publishing uploads a whole file, and a large project on a slow link takes
the server several minutes to receive and answer. The one-shot request
connections (hosting, the publish upload, a library, a plugin directory)
waited the session idle timeout (45s), so a large publish timed out with
"no answer from the server" before the server finished. Add a dedicated
request_reply_timeout (5m) for these connections, keeping the session idle
timeout unchanged.
Co-Authored-By: Claude <[email protected]>
Found while testing crates/pathops/src/lib.rs:352 (path_end, used by arrowhead)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Text went onto a path only with the Design panel's Put text on path
button (a text layer and a path selected together) or with text.onPath.
The Text on a path tool, in the Text flyout of the toolbar
(tool.textOnPath), shows the I-beam over a shape's or vector's path, and
a click there puts new, empty text on the path, starting at the click,
and starts typing. The text takes the path layer's place, transform,
box, constraints and effects, and its fills when one of them shows; the
path layer is removed. Ending the edit with nothing typed puts the path
back with no undo step.
The tool measures the distance on the page, so a layer scaled more one
way than the other, or shown smaller, is reached where it shows. Shapes
in front whose outline is away from the click do not hide a path behind
them, and paths that a mask or a clipping frame hides at the click are
skipped, as are locked layers and layers inside instances.
A double-click into a text layer that was just made no longer makes it
count as an existing layer, so leaving it empty records no undo step,
with this tool and with the Text tool.
Found while testing crates/engine/src/codegen.rs:112 (num)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The status line of docs/file-format-parity.md repeated eight topics,
from SVG/PDF export depth refusal (#208) to "last major". Each topic is
now listed once, and the boards entry closes its parenthesis.
The Files paragraph of the CLI usage text now names .sketch (Sketch 48 or
newer, best effort), as docs/cli.md already does (PR review).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/doc/src/text.rs:592 (TextData::replace_range, with normalize at :573)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Conflicts resolved keeping both sides: `.sketch` stays in `is_document`
(now via `OsStr::to_str`); ROADMAP and the parity, roadmap and
file-format docs keep main's headers and rows with the `.sketch` notes
and history rows added.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Release builds embed some families with only a few styles (Poppins
Regular, Italic and Bold), and the first font scan dropped every
installed face of such a family, so an installed Poppins SemiBold or
Medium was never offered. The first scan now keeps the embedded faces,
which still win for the styles they cover, and offers the installed
styles they lack beside them, as a rescan already did.
Closes#1246
Three parity rows still listed work that has landed on main. The
prototyping Triggers row listed delays on mouse enter, leave, down and
up as missing; they landed on 2026-10-13. Canvas & navigation's
remaining work in target-app-parity listed outline mode's Include
hidden layers and Include object bounds, which View > Outlines has.
The file-format Fonts row said missing fonts give warnings only; the
desktop app lists them in the Missing fonts window, and replacing them
is still missing.
Each file gets its status line and a revision-history row. The
percentages and hours are unchanged.
Copy as CSS, SVG, PNG and Copy link to selection move into a Copy as
submenu, placed under Paste to replace, instead of four flat rows near
the bottom of the menu.
Edit > Copy as > Copy link to selection (edit.copyLink) was only in the main
menu. Add it to the right-click menu on a layer, after Copy as PNG, labelled
with the command's own translated label.
The command now says "Link copied" when it copies, and copies nothing when
no layer is selected (it used to put figcraft://node/0 on the clipboard).
Page rows and layer rows take drags in Dev Mode again. A drag on either leaves the order as it was and shows the Dev Mode note, as a double click or the row menu already does.
samples() reversed the whole script list for a Korean UI, which put the
Bengali and Thai entries ahead of Hangul. Swap Hangul and Kana instead,
so the order is Hangul, Han, Kana, then Bengali and Thai. A test covers
the mixed case.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Cs5R9irCMt9rwfhpkwxvSn
The Corner smoothing row showed "iOS: 60%" as plain text. The help
page on corner radius and smoothing says to click iOS to set corner
smoothing to 60%. The row now has an iOS button that sets
cornerSmoothing to 0.6 on the selection in one undo step, with the
hover text "Set corner smoothing to 60%".
Layer, page and frame names in Thai drew as tofu squares in the panels and
on the canvas labels: the UI loaded system faces for CJK and Bengali only.
- cjk.rs: Thai is a fifth script (U+0E00..U+0E7F). When a name, a typed text
or the UI language needs it, the desktop app adds an installed Thai face
(Noto Sans Thai, Leelawadee UI, Tahoma, Sukhumvit Set, Thonburi, Ayuthaya,
Sarabun, Loma, Garuda, Waree). English users still load nothing; the web
build leaves Thai out, like Bengali.
- locales/th.toml: machine-drafted Thai catalog (every UI string, 589
command labels, the menu wordings), marked as a draft needing native
review. i18n.rs: Thai has no plural forms.
- docs/translating.md and docs/localization-parity.md updated.
cargo test -p figcraft-ui-egui passes (662); cargo xtask i18n-report,
assets and layers are ok. cargo xtask wasm and clippy were not run
(no wasm32 target here; clippy -D warnings already fails on other files
with Rust 1.95).
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Cs5R9irCMt9rwfhpkwxvSn
Polygon and star outlines did not use their corner smoothing: the value
was stored, imported and exported, but `shape_path` drew their corners
as plain circular arcs, and the Design panel showed the Corner
smoothing field only for rectangles and frames. Rounded polygon and
star corners now take smoothing the way rectangle corners do. Each
corner keeps a circular arc over the unsmoothed part of its turn and
eases into both edges, and it starts (1 + smoothing) times the plain
corner's length from the vertex. When a smoothed corner does not fit
in half of its shorter edge, smoothing is reduced first, then the
radius, as for rectangles. The Individual corners button opens the
panel for polygons and stars too, with the Corner smoothing row and no
field per corner.
The Design panel's font family and font style menus offered no way to
bind a string variable; these bindings could be made only by commands
such as variable.bind. Each menu now starts with Apply variable, which
opens the variable picker on string variables for the selected text
layers. A bound family or style shows its variable's chip in place of
the menu. A click on the chip picks another variable, and its detach
button unbinds it. While a text is being edited (a caret or selected
characters), the menus have no Apply variable item and show the plain
dropdown. A pick applies to the selected characters, or to the whole
layer when no characters are selected, as before.
The variable picker opened from a style's field listed every variable of
the field's type, and the Fill and Stroke style picker's Color variables
listed every color variable, whatever their scopes. Both now offer only
variables scoped to the property, as the Design panel's pickers do. A
paint style's colors take variables scoped to a fill or to strokes, an
effect style's fields take effect variables, and a text style's fields
take the variables of their font property. Layout grid fields still take
any number variable. style.bind still binds a variable outside its
scopes, as variable.bind does.
Found while testing crates/engine/src/cmd/arrange.rs:210 (distribute, also used by arrange.packH/packV)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The style editor bound variables to paint colors, shadow colors, blur
radii and grid sizes, but a text style's fields took variables only
through style.bind. The font and font style menus now start with Apply
variable, which opens the picker on string variables, and the size, line
height and letter spacing fields show a variable button on hover for
number variables, as the Design panel's number fields do. A bound field
shows its variable's chip in place of the value, as in the Design
panel. A click on the chip picks another variable, and its detach
button unbinds it.
kurbo restarts the dash pattern at every MoveTo, so many short subpaths
with a pattern of tiny dashes before a long gap made ~1M dashes (5M path
elements) while the length-based estimate allowed them. stroke_center
now adds one pattern's length per subpath to the estimate: an upper
bound on the dashes kurbo makes.
Test: each_subpath_restarting_the_dash_pattern_counts_toward_the_limit
(5M elements before).
Reported by CodeRabbit on #1278.
stroke_center handed any dash pattern to kurbo::stroke, which makes one
outline per dash: a 0.01/0.01 pattern on a 10,000-unit line made 2.5M
path elements, and a long line with a tiny pattern ran out of memory.
The variable width stroke already drew solid past MAX_DASHES (20,000);
that estimate is now width::dashes_fit and stroke_center uses it too, so
both strokes follow one rule. The path length is only measured for
dashed strokes.
Test: a_dash_pattern_too_fine_for_its_path_strokes_solid (2.5M elements
on the old code).
In Dev Mode the menus, shortcuts, clipboard, canvas and layers panel
refuse design edits, but an image or SVG dropped on the canvas was
still imported. The Pages list still added, renamed, reordered,
duplicated and deleted pages. The Assets panel still placed instances,
swapped them on an Alt-drop, applied styles, changed Hide when
publishing, accepted library updates and added, removed or swapped
libraries.
All of these now go through dev_mode_refuses with the command they
would run, and show its note. A dropped document still opens as a file
of its own, and the drop overlay shows the note instead of "Drop to
import". A click in the Pages list still opens a page, and page rows
take no drags in Dev Mode.
boards() listed a section before its contents, so the label of a frame
near the section's top-left was painted over the section's title pill,
and label_at (last label wins) gave clicks on the pill to that frame.
A section now comes after everything inside it: its pill draws on top
and takes the click.
Test: a_sections_pill_is_above_the_labels_of_frames_inside_it (fails on
the old order).
world_render_bounds and children_render_bounds recursed once per level
with no limit, so a crafted chain of ~20,000 nested groups overflowed the
stack and aborted the app on PNG/JPG/WebP export, rasterize, copy as PNG
and thumbnails. A stack overflow is an abort, not a panic, so the guard
can't catch it.
The walk now carries a depth and stops descending at 256, the depth
draw_layer draws to (MAX_EXPORT_DEPTH): bounds match what is drawn and
shallower trees are unchanged.
Test: render_bounds_stop_where_drawing_stops_on_very_deep_chains (aborts
on the old code).
Found while testing crates/pathops/src/lib.rs:326-333 (stroke_outline_of, StrokeAlign::Inside and Outside arms calling binary(), which uses FillRule::NonZero at :82)
Co-authored-by: Claude Opus 5.5 <[email protected]>
Found while testing crates/doc/src/colors.rs:30 (TextStyle::for_each_color_mut, used by TextData, Style and Document::map_colors / set_color_profile)
Co-authored-by: Claude Haiku 5.5 <[email protected]>
Found while testing crates/engine/src/cmd/props.rs:145 (line_height) and :153 (letter_spacing)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A Selection colors row could take a color style but not a color
variable, and a row linked to a style or variable could not be
unlinked from there.
selection.bindColorVariable binds a color variable to every solid fill
and stroke slot a row stands for. It replaces a color style on that
list and any earlier binding of the slot. A text range with fills of
its own takes the variable's value in the layer's modes. Gradients and
slots outside the variable's scopes are refused.
selection.detachColor unlinks the style or variable of a row's slots
and keeps their colors.
In the Design panel, the Styles menu of an unlinked row lists the
document's color variables after its color styles, and a linked row
shows a Detach button in the Styles button's place.
Review follow-up. The overlay kept the screen rect its trigger had when it
opened, so after a resize, a size option change or a pan a trigger-relative
(or manual) overlay stayed where the trigger used to be. It now keeps that
rect in page units of the presented frame and places it again on every
layout. The dropdown test switches to Fill while the menu is open. The
anchor's is_screen gets a doc comment.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the Figma forum request "Set overlay position relative to element,
not to page": an open overlay action takes anchor: screen (default) |
triggerInside | triggerOutside. With a trigger anchor the position preset
is measured on the layer that opens the overlay instead of the screen.
Outside puts top presets above it and bottom presets below it with the
edges lined up (a dropdown is Bottom left + Outside); Center centres it
on that layer. Manual is unchanged.
The interaction editor gets Relative to (Whole screen / Trigger layer)
and Inside / Outside. The player, the command (and so MCP and the
control channel) carry it; the command's params doc lists it. Figma has
no such setting, so .fig export writes it as a manual offset. Old files
and JSON without anchor read as screen and write unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* M6: Dev Mode measurements kept on the canvas, a read-only layers panel
Measurements placed between two layers, along a side or to the layer
around one (devMode.addMeasurement / removeMeasurement / measurements),
stored on the layer they start from (saved, synced, undoable; old files
open without them), drawn in Dev Mode as labelled redlines that follow
the layers and listed, placed and removed in the Inspect tab. In Dev Mode
the layers panel only selects and expands.
* M6: Dev Mode measurements: explicit imports, a measurement row of its own (Sonar)
Thai, Lao, Khmer and Burmese are written without spaces between words.
Tools that know the words insert U+200B ZERO WIDTH SPACE so lines can
break between them, and line breaking already treats it as a space. But
justification counted every space-class char as a gap, so a justified
paragraph got extra room at every U+200B: visible gaps between all its
words, which is not how these scripts are justified.
Item::stretches() (a space other than U+200B) is used for the justify
gap count, the trailing width and the pen advance. Line breaking is
unchanged. Test: on a justified line of words joined by U+200B and one
real space, the extra room goes to the real space only; with no real
space the line is laid out as if left-aligned.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Duplex::read_until gives a reader a deadline: each socket read waits at most what's left of it.
The server reads the handshake and hello under the hello deadline, so a hello trickled in
fragments can't outlast it; manage's close drains under CLOSE_DRAIN, so a silent server can't
hold it for the idle timeout.
A click on Save takes the keyboard from a focused panel field, which commits next frame: the files
save then, as the field-save shortcut does. OsStr::to_str as a method reference (SonarCloud).
auto_layout's grid branch moves to grid_layout and grid_sizing (SonarCloud cognitive complexity).
unset() clears gridColumnsSizing and gridRowsSizing, so an instance layer with equal tracks no
longer reads back its main layer's track sizes.
color_alpha is the part of a solid paint's opacity that is its typed colour's alpha, not a flag:
binding a colour variable takes only that part out, unbinding puts the variable's alpha back in,
and REST import records it. A 50% paint bound to a 50% variable stays at 25% bound or not.
* chore: resolve SonarCloud maintainability issues
- Define SHA256_FMT constant in packaging script to remove duplicated string literal (shelldre:S1192)
- Remove redundant 'static lifetime annotations from const items (rust:S8863)
- Replace closures with method and function pointers (rust:S1612)
- Extract helper functions in synthetic document generation, bezier polyline flattening, and component restoration to reduce cognitive complexity (rust:S3776)
* fix(packaging): use verify_sha256 helper function to avoid shellcheck SC2059
* fix(engine): remove duplicate closing brace in large_document
* fix(engine): take &mut Document in restore_id_map; keep 'static on associated consts
restore_id_map allocates ids, so it needs a mutable document. Elided
lifetimes in associated consts trip elided_lifetimes_in_associated_constant,
so those annotations are restored.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Assign positional parameters to locals in sha256 verification
---------
Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: Dylan Taylor <[email protected]>
UiState::open_left_panel / close_left_panel set the narrow window's panel along with left_collapsed:
the Assets panel (Shift+I), libraries, file and layer rename, page select, version history, the rail
tabs, the panel's Collapse button and panel.leftSidebar (which toggles what a narrow window shows).
Test: view.assets opens the left panel in a phone-sized window.
Resolves the package.sh conflicts with main's SonarCloud cleanup: the cross-build changes are kept
and use main's style, with [[ ]] tests and release_arch taking its argument as a named local and
ending in an explicit return.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Tailwind v4 drops theme variables no utility uses, but page code also
reads tokens through var() and (--name) utilities, so the export keeps
them all. Two test assertions mixed && and || without parentheses.
- Text that already fits keeps its exact size; shrunk sizes round to a
tenth but never below minSize (per run, its share of it).
- A last line after a hard break is not an orphan, and the no-break space
is only placed inside the last paragraph.
- docs: data.fill documents formats? and errors?.
Found while testing crates/svg/src/lib.rs:549 (unique_id) with :162 (id) and :387
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
On a device, a presented frame bigger than the screen now scrolls through
its whole size whatever its overflow, as Figma documents for top-level
frames. Before, "No scrolling" frames only clipped, so the landing page
template's long Phone frame couldn't be scrolled in Present. Content past
a frame's own edges still scrolls only on the axes its overflow allows.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Board files (File > New board) with sticky notes, shapes with text, connectors that follow what they join (straight or elbow, arrowheads, labels), marker, highlighter and eraser, and stamps; a board toolbar and a contextual bar; the Design panel hidden on boards, Prototype and Dev Mode refused there. Boards are ordinary FigCraft files built from ordinary layers, so saving, undo, export, MCP and live sessions work unchanged. File tabs, recent files and Open files show a pen nib for design files and a note for boards (original icons drawn in code).
Import lines for connected components inside generated layer trees; Compare changes against any saved version; a Dev Mode behaviour suite with an inspect-only canvas (menus, shortcuts, paste and Revert to saved refuse design edits; a drag under way is dropped on entering), Changed in the ready list, statuses only on top-level frames and sections; read-only plugins' panels from the Inspect tab.
Found while testing crates/doc/src/geometry.rs:75 (hits_node_itself, the `_` arm, using path.winding(lp) != 0); compare VectorNetwork::fill_contains at crates/doc/src/vector.rs:321
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
* CI: run the gate through craft-common-actions' shared gate.yml; FreeBSD nightly and on release
ci.yml calls the shared workflow (cached apt packages, nextest, main-only cache saves, Cargo.lock check). FreeBSD runs nightly and on pushes to release instead of on pull requests. Depends on storytold/craft-common-actions#69.
* CI: build the shared gate without craft-fonts, as before
* CI: build the shared gate with release.yml's craft-fonts pin
The set wrote one definition per name, so a variant whose property of
that name had another type named a definition the file didn't hold
(CodeRabbit review on #1152).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Packaging: pin the Windows signing client and check its hash
* Packaging: pin appimagetool and check its hash
* Reject AppImage download for architectures without a pinned appimagetool
* Pin the AppImage runtime
* Dev profile: line tables for workspace crates, no debug info for dependencies
* Tests: one integration-test binary per crate with several test files
* xtask ci: nextest, doctests, tests without examples, --keep-going; add xtask test
SonarCloud flags every bare [ conditional in bash scripts under packaging/ as a reliability smell (rule S7688). Convert the flagged [ ... ] tests to [[ ... ]]: safer word-splitting and no glob surprises, same semantics for these string and file tests.
- Use [[ ]] instead of [ ] for conditional tests in the bash packaging scripts (S7688). Every
converted test compares quoted values to literals, so no pattern-matching change.
- Name positional parameters with locals inside functions (S7679).
- End the env.sh helpers with an explicit return that keeps their exit status (S7682).
- Give case statements a default branch (S131).
- Send the check-app-specific.sh error to stderr (S7677).
packaging/freebsd/package.sh is left alone while #1158 rewrites it.
Adds crates/ui-egui/locales/it.toml: all 1747 UI strings (42 plural entries,
one/other), 574 command labels and 31 menu wordings, marked as a draft awaiting
native review like the other machine-drafted catalogs. Adds a resolve/plural
test for Italian locales and lists Italian in docs/translating.md.
Not compiled locally; the catalog was validated with a script (same keys as
en.toml, identical placeholders and ellipses per entry).
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Fr2xzM8A8RiUwZWddELBCt
read() returns Result<Message>, not a lock guard; these test loops
deliberately ignore what the client sent first.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <[email protected]>
Review: the doc test serialised TextStyle directly. The engine test makes
text with text.create (no fontFamily) and with a Text tool click, and
checks each starts in figcraft_doc::DEFAULT_FONT and writes it with the
text. Passes with FIGCRAFT_DEFAULT_FONT unset and set.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review: the test chose its Thai face, and checked each glyph's face, by
U+0E01, which the sample doesn't contain. It now asks for a face that has
every Thai character of the sample, and checks that each glyph's face has
the characters of its cluster.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review: a saved family renders only where that family is available. The
docs and the test's comment now say the variable changes only what new
text starts in, and that a file keeps its fonts in a build with another
default, falling back like any missing family elsewhere.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The accepting thread read listener.incoming().flatten(), which drops a
failed accept and asks again at once. While accepting keeps failing, for
example when the process has run out of file descriptors, that thread
kept a core busy. It now waits 100 ms after a failed accept. The loop goes
through a small accepted helper so the wait can be tested.
With points selected only on another edited layer, a press inside the
combined box over an unselected filled region of the active layer picked
the region instead of dragging the box. Other layers' selections now count
in the same priority check; a press on an already selected region still
drags the regions.
TextStyle::default() hard-coded Inter, which has no Thai (nor Hebrew,
Devanagari, ...). In a build for users who mostly write such a script,
every new text layer started in a font that can't draw what they type:
fallback faces hide the boxes, but the layer still says Inter and gets
Inter's metrics and line height.
Like FIGCRAFT_DEFAULT_SERVER, a build variable read with option_env!
names another family; unset or empty means Inter. font_family is always
written, so files open in the same font whichever build reads them.
docs/releasing.md gets a "The default font" section.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
In the web build, Thai layer names, frame labels on the canvas, text in
property fields and the layer search were drawn as missing-glyph boxes,
even when the build embedded Noto Sans Thai: the UI's built-in fallback
list (BUILT_IN in theme::install_fonts) didn't name it, and the browser
has no system fonts to fall back on.
Noto Sans Thai joins BUILT_IN. It is used only when the face is
embedded, so builds without it are unchanged. Test: with it embedded,
the UI fonts draw Thai, marks included (skips without it).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Thai typed into a layer whose font has no Thai glyphs (Inter, which new
text starts in) was drawn as missing-glyph boxes, on the desktop and on
the web, even with a Thai face embedded or installed: FALLBACK_FAMILIES
listed Latin, CJK, Hangul, Arabic and symbol families but no Thai one.
Noto Sans Thai, Noto Sans Thai UI, Leelawadee UI, Thonburi and Tahoma go
after Noto Sans Arabic; they are only consulted for characters the
layer's own font lacks. Tests: the list has a Thai face for each
platform, and with one installed or embedded every Thai cluster in a
Latin layer, marks included, is drawn from it (skips without one).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A submenu with no room beside its row (a phone) opens over its parent menu instead of off the
right edge. The toolbar pill is not drawn while a narrow window shows a side panel, where it was
squeezed into the strip left beside the panel.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Windows narrower than 760 pt show the Layers or Design panel over the canvas, opened from two
touch-sized buttons. Two fingers pan as well as pinch. The main toolbar pill is capped to the
canvas width and scrolls sideways (drag or wheel) when its buttons don't fit; the help button
moves above it when they would overlap. The web page's viewport covers the safe area and stops
touch scrolling and overscroll.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A Vue component's instance-swap prop now defaults to its component reference in withDefaults, so the dynamic component renders the design default. Attribute detection only applies inside an unfinished opening tag, so text like `Total = {title}` stays text instead of becoming a quoted attribute value.
The format keeps one default per property of a set; the first variant's
is written (CodeRabbit review on #1152).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Grids wrote their tracks as equal ones, with a warning. Track sizes
(fixed px, hug, fill shares) are now written as gridColumnsSizing /
gridRowsSizing, the fields the import already reads.
Checked on a synthetic grid in the incumbent (fixed 320/140/112/34/34 px
columns): written as equal tracks it came in with five equal columns and
its layers overflowing their cells; now its columns keep their sizes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A quoted Vue attribute stays open across placeholders, so `title="{a} {b}"` escapes both values as attribute text. Plain Vue attribute values in instance calls now HTML-escape `&`, `<` and `>`, so a literal `&` survives the round trip. Tests cover both.
The "On this page" swatches were indented by a single add_space, so once they wrapped the later rows started at the picker's left edge and ran past its right edge. The test fills a page with 40 distinct colours, opens the picker and checks every painted swatch lies at least 16 pt inside both sides of the picker window. Without the fix the rightmost swatch ends 3 pt inside the edge instead of 16.
With points selected in more than one edited layer the transform box now
wraps them all, in the page's axes since the layers can be turned
differently. Its corners and edges scale them, outside a corner rotates
them and inside drags them; each layer's points are mapped in its own
coordinates and the whole drag is one undo step. A click inside clears
every layer's points.
Review follow-up. CARGO_BUILD_TARGET must be aarch64-unknown-freebsd or x86_64-unknown-freebsd,
the two the release builds; anything else, including made-up triples such as
aarch64-foo-unknown-freebsd, stops with exit 2.
Part of storytold/craft-common-actions#39.
Review: after a Save as, closing called request_close, which asked the
question again while another file still had changes. It now carries on
with save_then_close. Save is the default answer: Enter picks it when no
button has the keyboard.
Review follow-up. With CARGO_BUILD_TARGET set on a host that is not FreeBSD (say an x86_64 Mac
building x86_64-unknown-freebsd), the architectures match but the binary still cannot run, so
the --version run is skipped there and the file(1) check is used instead.
Part of storytold/craft-common-actions#39.
Closing a window with unsaved changes asked "Close the window and lose
them?" with only Close without saving and Cancel. It now asks
Don't Save / Cancel / Save, as PhotoCraft and VectorCraft do. Save saves
each file of the window with unsaved changes and closes; a file without
a path (or saved as .fig) gets Save as first, and closing goes on once
that saved it. A failed save or a cancelled Save as keeps the window
open with its changes. The two changed strings are updated in every
catalog.
Review follow-up. The file(1) check for binaries that cannot run here now needs an executable
(file says "executable" or "pie executable") or an ELF with an interpreter, which is how an older
file(1) still shows a PIE it calls a shared object. FreeBSD libraries, object files and core files
for the target architecture are refused.
Part of storytold/craft-common-actions#39.
With the keyboard in a Design panel field (X, W, a colour…), the
shortcut handler returned before looking at any key, so Cmd/Ctrl+S and
Shift+Cmd/Ctrl+S did nothing: no save, no message, the tab kept its
unsaved dot. Those two now work from a field: the field lets go of the
keyboard, which commits what was typed as clicking away does, and the
save runs the next frame.
Review: `mcp --bogus --help` stopped at --bogus with exit 2. The parser
now keeps the first bad option and reports it only when no --help or -h
follows.
Review: " . .name. . " kept inner dots at the edges; trim spaces and dots
in one pass. docs/cli.md and safe_file_name's comment give the 100 and
200 character limits.
Review follow-up. CARGO_BUILD_TARGET must now be a *-unknown-freebsd triple, so a Linux or macOS
target can no longer be packaged under a FreeBSD name, and it is read before uname -m, so an
unsupported host does not block a valid cross build. When the binaries are for another
architecture and cannot run --version, file(1) must report each as a FreeBSD ELF executable
for the target architecture.
Part of storytold/craft-common-actions#39.
Signs such as U+2316 and U+2912 are in no text face. The fallback list
had the macOS and Windows symbol fonts (Apple Symbols, Segoe UI Symbol)
but nothing a Linux install has, so they drew as empty boxes there even
with Noto Sans Symbols 2 or Noto Sans Math installed. Adds Noto Sans
Symbols, Noto Sans Symbols 2 (and the Symbols2 spelling some packages
use) and Noto Sans Math after them.
`figcraft-cli mcp --help` failed with "unknown option --help" and
exit code 2; only a leading --help worked. The argument parser now
returns CliError::Help for --help or -h among a subcommand's options
(not as an option's value), and main prints the usage on stdout with
exit code 0. docs/cli.md says so.
export into a folder replaced every character outside letters, digits,
'-', '_', space and '.' with '_', so a frame called
"Homepage - desktop (live + verbeteringen)" became
"Homepage - desktop _live _ verbeteringen_.png". The app's
file.exportLayers keeps those names and only replaces what file systems
refuse. The CLI now uses the engine's safe_file_name, and docs/cli.md
says which characters are replaced.
Also adds my own entry to contributors/people.toml.
Like an unset axis, a cleared maximum was written as 0 and could
collapse the layer (CodeRabbit review on #1152).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The release workflow now cross-compiles FreeBSD aarch64 inside the x86_64 VM
(storytold/craft-common-actions#50) and sets CARGO_BUILD_TARGET=aarch64-unknown-freebsd.
package.sh now reads the binaries from target/<triple>/release when CARGO_BUILD_TARGET
is set, names the archive after the target instead of uname -m, and skips the --version
run when the binary is for another architecture than the machine. x86_64 packages are
unchanged.
Part of storytold/craft-common-actions#39.
Checked by importing synthetic files into the incumbent (Drafts > Import,
black-box) and against a "Save local copy" of a synthetic document.
- Images are named and referenced by the SHA-1 of their bytes. With
SHA-256 names, any file with an image failed to import ("Internal error
during import").
- Component sets hold their property definitions, variant properties
included (stateGroupPropertyValueOrders); each variant has its own copy
of each definition (parentPropDefId) and its variantPropSpecs, and its
layers and instances name that copy. Definitions and values also travel
as varValue; text values carry their lines; layers get componentPropRefs
next to the parameter consumption map. Before, sets came in without
variant properties and with their other properties unbound.
- An axis without a maximum size is written as infinite: 0 collapsed
auto layout frames with only a maximum width.
- Compound weights are spelt the way font families name their faces
("SemiBold"; "Semi Bold" in Inter): text in Libre Franklin Semi Bold was
a missing font there and wasn't drawn.
- The kiwi encoder writes infinite floats.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/text/src/layout.rs:1070 (shape_segment calls TextCase::apply_after one char at a time; crates/doc/src/text.rs:124 apply_after)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
node.create, node.set and the other prop appliers map fill/stroke
"none" (either case) and null to an empty fills/strokes list, as
fills: [] already did (undocumented), so an outline-only shape is one
call - agents write "none" for no paint. KEYS and docs/mcp.md
document it; the error names the accepted forms. fill_none_tests
covers create, set, undo, text and the error.
Change to rebuilt the instance from the new variant, and set variable
and set variable mode re-applied bindings, but the player never laid the
result out: no text sizing, no auto layout. A hugging button with a
longer label than its main showed the other variant at the main's width,
its label clipped, and a text bound to a variable kept its old size.
These actions now run the engine's normalize on the player's document
afterwards, the pass that follows every edit on the canvas: text resizes
to what it shows, instances follow and auto layout reflows.
Validated formatting, workspace Clippy, workspace tests excluding the two pre-existing localized startup assertion failures tracked in #1017, layers, assets, WASM and plugins. The startup tests are unchanged in this independent fix.
Found while testing crates/doc/src/text.rs:138 (TextCase::apply_after; is_word_char at :153 counts ' and U+2019 as word characters)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Keep single-quoted attribute delimiters, render text placeholders as text (not quoted strings), skip a > inside an attribute value when placing v-if (and insert before a self-closing slash), give slot content a per-slot class prefix so repeated layer names stay distinct, use a binding name no generated prop can shadow, and add Vue to the five translations that still listed only the older languages.
Vue connected templates keep a string's JS quotes inside bound attributes, so `:title={title}` becomes `:title=""Buy""` rather than a bare identifier; static attributes take plain text. Nested instances no longer lose a boolean `v-if` guard, instance swaps render through `<component :is>`, slot names stay distinct when their idents lowercase alike, slot content CSS reaches the parent stylesheet, and text defaults escape `</script>` in the script block. The five locale catalogs the Vue key rename missed (bn, fi, ka, nl, pl) are updated, and the docs list Leptos/Vue consistently.
A file or library saved before "modes" was an override group has a mode
chosen on a layer inside an instance without its marker, so the next
rebuild or library import still put the main's mode back.
Copies follow their main layer's modes when the main changes them, so a
layer inside an instance on other modes chose them itself.
mark_layer_modes marks those when a file opens and when a published
library is read from a file or a server. Layers on their main's modes
and top-level instances stay unmarked.
A variable mode chosen on a layer inside an instance (the icon inside a
Close button, recoloured through an icon colour collection) wasn't an
override group. A full rebuild of the instance starts each layer over
from its main and carries only override groups, so the mode fell back
to the main's. Library import always rebuilds, as do main edits such as
a new auto layout direction, so a modal's recoloured close icon came
into other files in its default colour.
"modes" is now an override group: setting a mode inside an instance
records it, rebuilds carry it, rebasing a nested instance finds it, the
component panel lists it, and resetting it brings back the main's mode.
A top-level instance's own modes stay its own and record no override,
like its resizing.
Narrows the range-extension wording: the date library computes the
calendar fields, and splitting seconds into hours, minutes and seconds
is ordinary arithmetic, which is what the code does.
The rule said never to hand-roll calendar math, but the range fix
(shifting by whole 400-year Gregorian cycles so dates past chrono's year
range keep their old output) is cycle arithmetic, so the rule and the code
disagreed. The rule now names that exception: the library still computes
month, day and time, and only whole cycles are carried in the year.
`utc_date` (collab) and `date` (ui-egui) took `u64` seconds, which the old
hand-rolled civil-from-days code converted for the whole `u64` range. The
chrono rewrite used `from_timestamp(..).unwrap_or_default()`, so any instant
past chrono's own year limit (262143) silently became the epoch — changing
existing output. `u64::MAX` seconds, for example, used to read
`584554051223-11-09 07:00 UTC` / `Nov 9, 584554051223`, not 1970.
Fix: add a `utc_parts(days, sod, nanos)` helper per crate that moves the
instant into chrono's range by whole 400-year Gregorian cycles (146_097 days),
lets chrono do the month/day/time, and restores the real year afterwards.
chrono stays the only calendar logic; the cycle count just shifts the year, and
every `u64` day count fits in `i64` so the conversion is always exact (no epoch
fallback). Split seconds with `u64` arithmetic first to avoid the `i64` seconds
overflow near `u64::MAX`, matching the old code.
Tests now pin the old output instead of the epoch:
- collab utc_date(u64::MAX) == "584554051223-11-09 07:00 UTC"
(and utc_date(253_402_300_800) == "10000-01-01 00:00 UTC", already pinned)
- ui-egui date(253_402_300_800) == "Jan 1, 10000"
date(u64::MAX) == "Nov 9, 584554051223"
chrono's %Y writes a '+' before years past 9999, whereas the old hand-rolled
{y:04} did not. utc_date now writes the year with {:04} to keep the exact old
output. from_timestamp already reads a clock past chrono's range as the epoch
rather than panicking. Tests cover the 5-digit-year and out-of-range cases.
Replaces the two hand-rolled Howard Hinnant days-from-civil copies with chrono
as a workspace dependency:
- crates/collab/src/message.rs: utc_date (YYYY-MM-DD HH:MM UTC)
- crates/ui-egui/src/home_recents.rs: date (Mon D, YYYY, UTC; keeps the
translated month names and trf formatting)
Output is unchanged. Neither function had a test, so each gains one pinning the
exact strings the old code produced (leap days included). chrono uses only std,
so wasm stays safe. AGENTS.md/CLAUDE.md now say time goes through chrono.
Follow-up to storytold/cadcraft#1785
A Button main's override on its icon instance's glyph (a red fill) was
lost when an instance swapped the icon: the outer rebuild re-copies the
glyph from the main's layer with no overrides of its own, so the nested
rebuild matched it by name and carried nothing. Before that rebuild, each
old layer of the swapped instance now takes the overrides held along the
source chain it had before this rebuild, through layers inside instances
in the mains that place the swapped instance. Geometry, slot content and
groups the layer overrides itself are left out. Matched by name, the
values carry onto the new main's layer as the instance's own: inherited
at the swap, frozen after, and back to the icon main's value on reset.
An icon main's own overrides stay with it.
stroke_area used rect_side_strokes only for square corners; a rounded
rect-like layer with individual weights fell through to a uniform ring at
stroke.weight (canvas, SVG, PDF, shadows, shape builder, outline stroke).
pathops::rounded_side_strokes now draws them as CSS does: the outer edge is
the rounded rect (the fill outline for inside strokes), the inner edge is
inset by each side's weight with elliptical corners max(0, R - weight), so
border-l-4 rounded-r-lg is a straight bar and a weight that differs across
a rounded corner tapers through it. Center and outside grow the outer box
and each non-zero radius by half or all of the weight (CSS has only
inside). The inner edge runs against the outer one, so the ring is a ring
under non-zero too; where it could leave the outer edge (radii shrunk to
fit a narrow inner box, smoothed outer corners) it is cut out instead.
Found while testing crates/doc/src/proto_expr.rs:172 (display writes `not {x}` with a bare var_ref) with crates/doc/src/proto_expr_parse.rs:262 (not() checks variable names before the `not` keyword)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
apply_order gave fillStyle, strokeStyle, effectStyle, gridStyle and
textStyle the rank of plain values, so the JSON map's byte order decided:
fills, strokes, effects, grids, layoutGrids and variationSettings sort
after their link and overrode it (node.set then unlinked the style;
node.create kept a link its values contradicted). Style links now apply
after the values and x/y, just before boundVariables. A deviation set in a
later call still unlinks.
Found while testing crates/doc/src/proto_expr_parse.rs:462 (variable lookup before the true/false literals at :466) with crates/doc/src/proto_expr.rs:151/186 (display writes the literal as bare `true` and var_ref writes such a variable bare)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A third row in open_subpaths.svg: a filled and stroked path that is only
0.001 short stays one layer, and an open triangle with
paint-order="stroke" puts its stroke layer under its fill layer.
A file whose font family variables alias each other in several modes made
the missing-fonts walk branch in every mode, billions of steps before the
depth limit stopped it, freezing the app on open and `figcraft-cli info`.
The walk now visits each variable once.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
SVG fills every subpath as if it were closed; the import filled only
subpaths that close themselves (a z, or an end within 1e-6 of the
start). Logos exported from PDF round relative coordinates to three
decimals, so their outlines end about 0.001 short of where they started
and most of their letters came in empty.
A filled path now has its open subpaths closed before it becomes a
vector network: an end within 1/1000 of the path's size snaps onto the
start, any other open subpath of two or more segments gets a closing
segment. Stroke-only paths are unchanged. A filled and stroked path
with a really open subpath becomes a fill layer (closed) under a stroke
layer (open), so the stroke stays open as SVG draws it.
SaveFiles named each test's folder after the process id and the
clock in nanoseconds. The tests run in parallel and the macOS clock
counts microseconds, so two of them often got the same folder and one
failed with AlreadyExists (14 of 30 runs here). A counter now keeps
the names apart.
Found while testing crates/text/src/breaks.rs:42
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Switching an instance to another variant copied the old root's sizing
and min/max back over the new variant's on the rebuild, so an instance
switched from a hugging variant to a fixed-width one whose children
fill kept hugging at the old width. The next edit then turned those
children Fixed, as a child can't fill an axis its parent hugs.
Resizing still equal to the old variant's was never the instance's own,
so on a sibling-variant switch it now follows the new variant; sizing or
min/max the instance set itself stays.
- --no-system-fonts holds for .sketch imports too: the CLI opts the
process out, so the import's font matching doesn't scan.
- Shared text styles get the installed family and platform tracking
their linked layers get.
- Each text run writes its letter spacing, line height and font axes,
neutral when it has none, instead of taking the first run's.
- The unsupported-drop hint and its catalogs list .sketch.
- A test sends damaged .sketch data through file.open and file.import.
- Stack items are matched to their source layers through a map.
- Docs: .sketch is a convert input; the Carbon kit's Sketch 47 edition
is refused (the 68 and 83 ones open).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- usage: a non-plain Vue argument escapes " as " in the attribute value (same rule as the connected-code path)
- usage: Vue pads with two spaces like the markup, not four
- test: an unconnected instance with a quoted title asserts inspect.componentCode and inspect.code
Dev Mode generates React, Leptos, SwiftUI and Compose; a Vue project had to translate the HTML
instead. `inspect.code {language: "vue"}` now writes a single-file component: a `<template>`, a
`<script setup>` with `defineProps` and its defaults, and a `<style scoped>`.
A Vue template holds no arbitrary expressions, so what the web targets write as a braced
expression has its own spelling. Text is `{{ prop }}`, and braces in text are escaped; a boolean
property's layer carries `v-if="prop"` rather than a wrapper element; a slot becomes Vue's own
`<slot name="…">` with its fallback content, which takes no prop, and an instance's content in one
is a `<template #name>`. Variants switch on the root's `data-*` attributes, as on the web.
`codeConnect` links a component to its code as a single-file component: a default import, and a
quoted binding, so `:size={size}` gives `:size="Large"` and a component swap passes the component
itself. A value that needs escapes keeps them inside the quotes, where the attribute's own quotes
become `"`.
`inspect.componentCode` kept a second copy of the code-language list, which is how Vue was missed
when `inspect.code` learned it; it now asks `code_connect::language`, the one list, as
`codeConnect` does.
Eleven tests in `vue_codegen_tests` and a `codeConnect` case; the Inspect panel's language picker
picks Vue up from `CodeLang::ALL`, and the language lists in the docs, the Dev Mode hint and its
thirteen catalogs name it.
A Shift-click with the Pen already put the next point at a 45° step
from the last one, but the line drawn to the pointer while moving
ignored Shift and followed the pointer, so holding Shift looked like it
did nothing. The line now ends where the click will put the point. Both
use one helper, at_45_degrees.
Pin third-party actions to full-length commit SHAs (with version comments)
instead of mutable tags/branches to prevent supply-chain tampering via
retagged releases. Resolves SonarCloud rule githubactions:S7637.
Figma imports Sketch files from its file browser (desktop app too); the
area table hadn't listed it. Import/export breadth 86% → 87%, file
formats 60% → 61%; headline numbers unchanged after rounding.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Texts an auto layout places keep the layout's position (the next layout
pass would take the baseline move back). A faded layer's group takes the
layer's id, so stacks, links and overrides still find it; faded masks
stay masks. figcraft-format no longer depends on figcraft-layout.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Tinting copies a group's or master's layers; the copy is charged to the
import's work budget (past it, tints are skipped with a note). An
override handed to a nested instance costs its value's size, so deep
chains of large overrides can't be copied without end.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A tint or fill that isn't a JSON object is left alone instead of being
indexed. A structural fuzz replaces each value of a rich file with values
of other types and imports every variant.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sketch draws an artboard's layout and square grids only while its layout
view is on; they now import hidden (still there, shown with the eye in
the Layout grid section).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Alpha masks only mask (no drawn copy of their fill); a progressive
opacity (Fade) becomes an alpha mask of its gradient over the layer;
linear gradients run square in pixels as Sketch draws them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The font database knows each face's PostScript name; the Sketch import
finds the installed face a file's font names (CourierNewPSMT is Courier
New Regular) and reads the family from the name only when it isn't
installed. Text without a fixed line height takes the line Sketch draws:
ascent and descent rounded to whole pixels, the line gap left out, a
fifth more for classic Mac fonts, the baseline where Sketch puts it
(measured on sketchtool renders of 80 fonts).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sketch keys a font's variation axes by the tag read as a number; they map
onto the text style's variations (a fontVariations key in the REST
mapping). Font references without font data only name the fonts a
document uses and no longer raise the embedded-font note.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Groups as frames and graphics, stacks as auto layout (instances laid out
by their stack), corners and per-side borders from the style, pins and
sizing, tints (layeringType, fillColor overrides), text colour and
visibility overrides, blur lists (progressive, glass), inner shadows in
the shadow list, angular gradients starting on the right, text sizing,
and texts an override changed measured again on open.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Compared with Sketch's renders of a Sketch 64 file. Resized instances lay
out their layers the way Sketch does (measured: pins share the change in
proportion, a fixed size keeps its centre's share, and files before format
110 just scale); masters' backgrounds show in instances only when Sketch
includes them; swaps to a missing symbol keep the original; artboards
don't draw a style they carry. Apple's system fonts are found under the
names designs use (SF Pro, SF UI Text and Display, SF Compact), the
optical size follows the text size for named styles too, and imported
text keeps the platform's trak tracking as letter spacing.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Compared artboard by artboard with Sketch's renders of a real Sketch 50
file (sketchtool, black-box). Shape groups whose children have no
operation are one compound path filled by the group's winding rule (read
from the layer), not a union: rings, map streets and icons keep their
holes. Filled open paths close, as Sketch fills them. Flips apply after
rotation. Text with a fixed line height sits where Sketch's line spacing
behaviour puts it. Edited ovals and rectangles keep their points. Bitmap
colour adjustments become image filters. scale 0 means unscaled. A local
test compares sample files with sketchtool's renders.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the branch review: overrides that point at junk (shared styles that
aren't objects, text runs that aren't) no longer panic; symbol expansion
spends a work budget (JSON copied, overrides handed on to nested
instances, grouped once by their first path segment), so small files
can't expand into gigabytes; far path points stay in range; notes are
capped and de-duplicated with a set; shared styles are looked up by
index; a styled second mask's drawn copy isn't clipped by the first;
JSON members are refused past 100 times their packed size or 2 GB in
all, and an unreadable document.json keeps the pages.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
FIGCRAFT_SKETCH_SAMPLES runs every .sketch file of a folder through
file.open (opens with linked instances and finite geometry, or is
refused as too old). File format parity gains a .sketch row; CLI, MCP,
roadmap and ROADMAP mention the import.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Layers linked to artboards navigate there on click (back links go back,
slide animations slide), hotspots become transparent frames with their
link, flow-home artboards start a flow; export options become export
settings (scale, width or height, suffix); artboard layout and square
grids become layout grids, which instances show like their masters.
Smart Layout, stacks and locked overrides are noted once per file.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Shared layer styles become a paint style for their fills, a paint style
for their border colours (linked to strokes) and an effect style; shared
text styles become text styles (colour stays on the layers); local and
library copies alike, unused ones listed too. Layers link to them, keeping
their own values where they differ. Color Variables (swatches, also
library ones) become colour variables in a Color Variables collection,
with fills, borders and text colours bound to them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Symbol masters become components, also those copied from libraries (on a
page of their own); instances carry their master's layers with their
overrides applied (text, nested symbol swaps and hiding, images, layer
and text styles, nested paths) and resize them by their constraints, so
editing a master still updates every instance. A symbol inside itself
stops with a note; instances of missing symbols are empty frames.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Text layers keep their string and styled runs (font, size, colour, line
height, letter spacing, decorations, text transform; ranges in UTF-16
units, clamped), paragraph alignment and spacing, vertical alignment and
how the box sizes. Fonts named by PostScript name become family and
style; the engine then matches families to installed ones ignoring
spaces (PingFangSC → PingFang SC). Fonts embedded in newer files are
named in the import report.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sketch draws a clipping mask with its fills, borders and effects under
what it clips; a FigCraft mask only clips. A plain copy of such a mask
goes underneath it, so a light avatar circle with a white icon in it no
longer vanishes on a white artboard.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sketch 51 and older files make masks from shape groups; the mask flag was
lost when the group became its shape or boolean operation, so a masked map
spilled out of its card.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Solid, gradient (linear, radial with its ellipse, angular) and image
pattern fills (tile, fill, stretch, fit); borders as strokes with the
first enabled border's width and position, dashes, caps, joins and arrow
ends; shadows, inner shadows, layer and background blurs; layer and paint
opacity and blend modes; clipping masks, wrapped in a group of their own
where Sketch breaks the mask chain; Display P3 documents. Shape groups
pass their style to the shape or boolean operation they become. Noise
fills, motion and zoom blurs are listed in the import report.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Paths become editable vector networks (handles, per-point corner radius,
open and closed, winding rule). A shape group with one shape becomes that
shape in the group's place; several combine as boolean operations, mixed
operations nesting from the left as Sketch applies them. Stars and
polygons stay stars and polygons while their points are the generated
ones, edited ones and triangles become vectors.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Artboards (and Sketch 2025 frames) become frames with their background
(white by default, as Sketch shows them), groups and graphics groups,
rectangles with their corner radii, ovals ellipses, bitmaps rectangles
with the image, slices slices. Positions accumulate through groups,
rotation and flips keep their direction, resizingConstraint bits become
constraints. Not yet mapped layer kinds are listed in the import report.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
.sketch archives (document.json, meta.json, pages/*.json) open through
file.open, file.import, drag and drop, the CLI and MCP; files saved by a
Sketch older than 48 are refused with a message saying how to convert
them. Layers come in the next commits.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`file.saveAs {data: true}` marked the file saved before the UI handed the
bytes to `Services::download`. When the download was refused, the toast
said "Couldn't download" but the document was clean, so closing it no
longer warned and the edits could be lost.
The data result now leaves the save state alone. A new `file.markSaved`
(the bookkeeping a native save already did: the file and its branch tabs)
runs from `save_as_dialog` only once the download was delivered. It ends a
version preview like the other commands that act on the document itself.
Tests: the engine save test checks the file stays dirty until
`file.markSaved`; `a_refused_download_says_so` checks the document is
still dirty.
Clash detection compared every variable with every other, and suffixes
rescanned the taken list, so a file with very many variables made the
export quadratic. Counts in a HashMap, taken names in a HashSet and a next
suffix per base keep it linear.
Calls storytold/craft-common-actions/.github/workflows/sonarcloud.yml@main
with secrets: inherit. Adds SonarCloud badge to README where missing and
removes the redundant sonar-project.properties (config now passed by the
reusable workflow).
Instance swap defaults that name an in-file main skipped by the layer cap
were reported as "isn't in the file": only instance targets fed the
capped-mains list. Defaults now feed it too.
- inspect.code {language: "tailwind"}: a layer tree as a React component
styled with Tailwind CSS v4 utilities, no stylesheet. Auto layout maps to
flex utilities on the spacing scale; fixed-width auto layout frames, text
and wide layers become fluid (w-full max-w-[…]); instances are inlined
with their overrides, except those connected to code (codeConnect), which
are their snippet with one import each; bound variables are theme
utilities (bg-accent, rounded-card).
- mobile: <frame id> (same layer names): mobile-first code, desktop classes
that differ behind lg:, resets for desktop defaults, padding compared per
side; layers only on desktop are hidden below lg, those only on mobile are
listed in the header.
- variable.export {format: "tailwind"}: @import "tailwindcss" and an @theme
block (--color-*, --radius-*, --text-*, --spacing-*, --font-*), names
unique per file and shared with page code.
Builds on the token roles of the CSS export px fix (#1008).
- data.fill {id, rows, gap?, formats?, dir?, format?, scale?}: repeats a
frame once per row in one undo step; each row key names layers in the
copy (text, or {text?, fill?, visible?, fit?}). `formats` makes each row
once per size (feed, story, square), the copy resized so its auto layout
reflows. With `dir`, each copy is exported under a unique file name; a
failed export is reported in `errors` and keeps the copies.
- text.fit {ids?, maxLines?, minSize?, orphans?}: the largest font size,
down to a floor, at which fixed-width text stays within N lines and its
box; avoids a one-word last line; `fits` comes from the final text.
- Plugins can't call data.fill: it runs other commands inside, past the
plugin policy, like command.batch.
A radius exported as `--radius: 12` was invalid where CSS reads it as a
length (`border-radius: var(--radius)` ignores a bare number). Number
values now carry px unless the layers bound to them, their scopes, or the
variables aliasing them make them an opacity or a font weight, which stay
bare. The roles come from a small token_roles module.
Updates the variables_io_tests expectation for the radius (4 → 4px,
6.5 → 6.5px), and adds a test for a bound radius, a bound opacity and an
opacity alias's target.
Cause: in the browser, exporting several layers started one download
per file. Chrome holds back every download after the first until the
site is allowed automatic downloads, so the other files didn't arrive.
Fix: file.exportLayers takes `zip: true` (without `dir`) and returns the
files together as one base64 zip `data`, written by a new
figcraft_format::zip_files (the workspace's existing zip dependency).
The UI asks for it when the platform saves by downloads and hands over
one `<document name>.zip`.
Test: ui-egui several_layers_download_as_one_zip (one download, a zip
holding both layers). It fails on the previous commit (two downloads).
component.addProperty {type: "boolean", default: false, layer} hid the
bound layer in new instances but left it showing in the main component,
so the main didn't show its own default.
On the main, a boolean property's value is now the visibility of the
layers bound to it:
- addProperty, bindProperty {field: "visible"} and editProperty {default}
show or hide the main's bound layers to match the default (every
variant of a set for editProperty). Without a default, addProperty
takes the bound layer's current visibility instead of always `true`.
- Showing or hiding a bound layer on the main sets the default, and the
property's other bound layers (in every variant) follow. Edits inside
instances stay overrides.
- Existing instances keep their stored values; documents open as saved
(no normalisation on load).
Fixes#993
Co-Authored-By: Claude Opus 5.5 <[email protected]>
node.set took no `path`, so redrawing a vector meant deleting and recreating it and
losing its id, place in the layer list, paints, styles, bindings and step references.
`path` now takes SVG path data for vector layers and replaces the geometry through
the same reader node.create uses (path_network, shared with create). width/height in
the same call scale it as create does. The path's own box goes at the layer's current
origin, so rotation and flips stay; with x/y in the same call it goes at x/y plus the
box's offset, which is where create puts it.
On several layers, layers that aren't vectors are skipped with a warning, as auto
layout keys already are (the skip logic is shared now). A single non-vector is
refused, and so is a vector inside an instance: there is no geometry override, so
the next sync would undo it. Data that doesn't parse, draws nothing or has
non-finite bounds is refused cleanly; create now refuses non-finite bounds too.
Tests cover keeping the layer, placement and scaling against create, rotation,
undo/redo, invalid data, non-vectors, instances, boolean operands, edit mode, and a
junk-path sweep. The engine-wide junk sweep has node.set path entries.
Fixes#1001
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A batch step can name itself with "as" and later steps refer to it with
{"$ref": "name"}, alongside positional {"$step": n}. "field" now takes a
slash path into the result as well as a key, so {"$ref": "copy",
"field": "ids/0"} reaches the first id edit.duplicate returned (the
convention variable.bind uses for fills/0/color; a leading "/" as in
JSON Pointer also works, and a key holding "/" still matches first).
Resolution moves to crates/engine/src/batch.rs (BatchStep, BatchScope),
shared by command.batch, the CLI script runner, the control channel and
the MCP batch tool, which now passes "as" through. Duplicate names fail
before any step runs; unknown, same-step or later names, bad paths and
out-of-range indexes fail the batch with the step, name and full path.
$ref is reserved like $step; nested batches keep their own names.
Fixes#994.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`node.reparent {ids, parent, index}` read `index` against the parent's
children as they were before the move, so within the same parent moving a
layer up one place did nothing and asking for two places moved it one.
`index` now counts the parent's children after the move, the way
`node.reorder` already does, in the same parent or a new one; several
layers still land together in their stacking order.
The layers panel was the only caller passing an index other than 0. It
drops on a gap between rows counted before the move, so it now subtracts
the dragged layers below that gap and drags behave as before.
Documents `index` in the command's params and in docs/mcp.md.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Cause: in the web build no file has a path, so Save goes to Save as, and
Save as and every Export go through Services::pick_save / pick_folder.
WebServices keeps the trait's defaults there, which return no path, so
the callback never ran: no download, no message. Only Variables ▸ Export
used the platform's downloads (#368).
Fix: where services.downloads() is true, those routes run their command
without a path and hand the bytes to services.download():
- engine: file.saveAs takes `data: true` and returns the .figcraft file
as base64 `data` with its `name`, marking it saved (no path, nothing
remembered). Programmatic Save as without either still asks for a path.
- ui: FigApp::download decodes a result's `data` and says when nothing
arrived; Save as, one-off export (PNG/JPG/WebP/SVG/PDF), .fig export,
frames to PDF and multi-layer export (one download per file) use it.
Tests: engine save_as_data_hands_back_the_file_and_marks_it_saved;
ui-egui web_save_tests (save, export, refused download) with the fake
browser Services from variables_export_ui_tests. The UI tests fail
without the fix.
JIS X 4051's line-start (行頭禁則) and line-end (行末禁則) sets were mostly
there — closing punctuation and brackets, small kana, the prolonged sound
mark, iteration marks and opening brackets already never start or end a line —
but a few categories fell through:
- the halfwidth small kana ｧ–ｯ (U+FF67..U+FF6F) and the halfwidth voicing
marks ﾞﾟ (U+FF9E/U+FF9F) were not line-start prohibited, so a wrapped line
could begin with them;
- the spacing (semi-)voiced sound marks ゛゜ (U+309B/U+309C) and the kana
iteration marks 〱〲〳〴〵 (U+3031..U+3035) likewise;
- the halfwidth parentheses ｟｠ were missing from the line-end/line-start
sets.
Add them (ranges in `class`, the two parentheses in the OPEN and CLOSE
strings), and pin every JIS category in the behaviour suite: the existing 22
cases stay and 9 join them (halfwidth small kana and small tsu, halfwidth and
spacing voicing marks, kana and vertical kana iteration marks, both halfwidth
parentheses), plus a wrap case where a halfwidth small kana moves down with
its base (キｬ) and a case showing a manual newline may still start a line with
a prohibited character (禁則 is for automatic wrapping only).
`break_opportunities` fails on the old sets at the first new halfwidth
small-kana case.
Latin wrapping, ideograph breaks and the rest of the behaviour suite are
unchanged, and the headline numbers stay as they are.
Verified: cargo test -p figcraft-text --lib 92 passed; cargo clippy
-p figcraft-text --all-targets -- -D warnings clean; rustfmt clean on the
touched files.
Docs: docs/target-app-parity.md text behaviour row and ROADMAP progress log,
revision history and header.
A design-system .fig file (54 pages) stopped at 244 248 layers, the rest
of its pages skipped, and most of its instances came in as frames "of
components that aren't in the file", with the cap note shown twice.
- The layer budget follows the compressed canvas at 8 bytes a layer, not
32: design systems are mostly instances, which pack tightly. Files under
about 1.6 MB keep the 200 000 floor and the 2 million ceiling stays
(measured on synthetic files: about 5.4 KB and 19 µs a layer).
- The library copies the pages use (the file's hidden internal canvas) are
mapped before the pages: at the cap the pages' last layers are skipped,
not the mains their instances link to. The REST stage keeps building
pages first (its ids are first-wins) and allows for the wrapper frame of
each library copy, the one layer the mapper doesn't count; that layer
tripped the REST stage's own cap and repeated the note.
- Instances and swap defaults whose main is in the file but past the cap
say so instead of "aren't in the file" (en and es catalogs).
- The .fig import gives a note the mapper already gave only once.
Tests in fig/layer_cap_tests.rs: the budget for a #981-sized canvas, library
mains ahead of the pages at the cap, the note for a capped local main, and
a whole 210 000-layer file with one cap note and its library links kept.
All four fail on the previous code.
Plain `-` zoomed out but plain `=` (the `+` key without Shift) did nothing,
so a canvas zoomed out with `-` couldn't be zoomed back in the same way;
held down, `-` reaches the 2% floor and `=` leaves you there. Ctrl+Shift+=
zoomed in while Ctrl+Shift+- did nothing either.
Found with a trace of the input on GNOME Wayland: Ctrl+wheel, Ctrl+=,
Ctrl+-, Ctrl+Shift+= and `-` all reach the canvas and zoom; `=` and
Ctrl+Shift+- arrive too but no binding takes them.
`view.zoomIn` gains `=` and `view.zoomOut` gains `Shift+Cmd+-`, pinned in
the shortcut parity table; neither key was bound elsewhere and the menus
still show Ctrl/Cmd and `+`/`-`. A keymap test checks each zoom-in key
has its zoom-out twin on both platforms.
Panel dropdowns (prototype device, Preferences, blend modes, versions…)
draw their rows with `right::menu_row` on the theme's popup fill, which is
white in the light theme. Since 5a4b7c27 those rows painted their labels in
`menu_text`, white in both themes because it belongs to the dark menu bar,
so the options were invisible until hovered.
Rows now take the theme's text (`t.text`, shortcuts `t.text_secondary`) and
white only on the green hover highlight, as before that commit. The menu
bar, tool flyout and style/variable pickers paint their own rows on
`t.menu` and are unchanged.
Regression test: a dropdown opened headless in both themes keeps its labels
at 4.5:1 and shortcuts at 3:1 against the popup fill.
The colour picker's Window area is keyed by egui as Id::new(Some(title)), but
the style picker looked it up as Id::new("color_picker"), so the lookup always
missed and any click in the picker counted as an outside click that closed the
editor (and the picker with it). Give the window a fixed id and query it through
colorpicker::window_id(), with a regression test.
* feat(window): adopt shared craft-window for Linux decorations
Replace FigCraft's bespoke Wayland xdg-decoration probe
(apps/figcraft/src/linux_window.rs) with the shared craft-window crate
from craft-libs, so every Craft app agrees on who draws the window frame
per desktop.
The decision is now made once, before the viewport is built, by
craft_window::Choice::probe(Prefix::new("FIGCRAFT")) and mapped onto
FigCraft's existing TitleBar enum:
Frame::Desktop => TitleBar::Native (TitleBar::Inset on macOS, unchanged)
Frame::App(_) => TitleBar::Custom
This drops the old runtime send_viewport_cmd(Decorations(..)) switch and
the linux_window probe. macOS keeps its inset traffic lights and Windows
keeps the system frame exactly as before. The only behaviour change: on
KDE and other Wayland compositors with server-side decorations (Sway,
Hyprland, COSMIC) FigCraft now keeps the native frame instead of always
drawing its own custom title bar, matching the rest of the suite.
NOT MERGEABLE YET: craft-window is pinned to the unmerged feature branch
from storytold/craft-libs#217 and must be repointed to a merged rev
before this can land.
* fix(window): preserve Windows titlebar; pin craft-window to merged rev
- Keep FigCraft's own custom title bar on Windows. craft-window returns
Frame::Desktop on Windows, and the mapper was turning that into
TitleBar::Native, flipping Windows from its existing undecorated custom
bar to the system frame. Map Frame::Desktop to Custom on Windows (as
macOS maps to Inset); only Linux/BSD Frame::Desktop becomes Native. This
keeps macOS and Windows unchanged; the sole change stays KDE/Wayland.
- Repoint craft-window from the (now-merged) feature branch to the squash
commit on storytold/craft-libs (craft-libs#217), and commit Cargo.lock.
Verified: cargo build -p figcraft and cargo test -p figcraft (53 tests)
pass against the merged dependency; rustfmt clean.
Use only root-supplied Korean translations; overlapping unmerged PRs are not dependencies. Preserve existing main translations. Static catalog, placeholder and source-lookup checks passed. Rust and native/web UI tests were not run because Rust installation remains awaiting approval.
Preserve existing Korean translations and exclude unused keys and keys already proposed in open PRs.
Static catalog, key, placeholder, ellipsis, shortcut and whitespace checks passed. Cargo fmt, clippy, tests, xtasks and native/web UI validation were not run: Rust tools are unavailable and installation awaits approval. See docs/korean-translation-validation.md.
Found while testing crates/geom/src/lib.rs:287
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Include document UID in pages_seen_page to correctly handle page selection and scroll tracking when switching between documents. Record page as seen only after the active row renders so that pages filtered out by search scroll into view when the filter is cleared.
Found while testing crates/format/src/lib.rs:256 (to_json/write_json/save_bytes) and :323 (parse_json)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/svg/src/lib.rs:589 (also :604 and :611)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The threaded test asserted a fixed glass scene renders in over 50 ms,
which would flake on a fast machine or once the glass speedups land.
It now grows the scene until the full render takes at least twice the
threshold plus the preview's own time, so the watchdog's preview comes
first with a margin wherever it runs.
Refs #295
Whether a zoom got a preview rested on the previous render from
scratch, so the first render, another page or file, and a zoom from a
quick view to a heavy one came without one: the page colour showed
until the full render was done.
Rendering every such preview up front would cost 8-56% of a full render
(bench pages of 1k-50k layers at 2880x1800; per-layer costs dominate
large pages). Instead the render thread arms a watchdog thread as it
starts a render from scratch without a preview; once that render has
run past 12 ms the watchdog renders the preview beside it, on one core
while the full render keeps the others, and sends it unless the full
image is done. Quick renders cost nothing and never flash a preview.
Warm slow zooms keep the preview first, which keeps up with a gesture.
With two threads sending, finished images carry their request's
number: a poll takes the newest request's, and of one request the
refined image; the canvas never shows a preview over the refined image
of its own or a newer request.
Refs #295
inspect_document sends a shallower tree when the document's is too large
and points at get_node for the rest, so get_node was the way out of that
limit -- and it had none of its own. Measured against a document with a
frame of 15,000 children, through figcraft-cli mcp:
inspect_document depth 64, allPages 14.29 MiB (8.00 MiB cap not
applied before #915)
get_node depth 64 10.51 MiB
Both are over the 10 MiB buffer the report is about, and for get_node no
depth makes it smaller: the width is at depth 1, so depth 1 is already
9.12 MiB of text and depth 0 has no children at all.
get_node now takes the same maxBytes, cuts the subtree by depth as the
document's tree is cut, and then takes a window over the children when
even depth 1 is too large. The marker says how many children were shown
of how many, and the offset to ask for next; `offset` is a new argument
for reading the rest. After this:
get_node depth 64 8.00 MiB, children 11427 of
15000 from 0, next 11427
The hint in inspect_document's marker no longer promises that get_node
has no size limit, since that was the hole: it now says a wide node
answers with a window and an offset.
The test pages through a windowed node and asserts every child is
reported exactly once, in order, with every page inside the limit.
Unbounded, the first page is 23,003 bytes against a 7,667-byte limit.
A .fig import's notes were deduplicated and left uncounted, so each kind
was named once with no sign of how much of the file it applied to: a
file whose every text layer lost a setting read exactly like one where a
single layer did. That is what made the loss in #467 invisible -- the
reporter saw one line while three kinds of data went, one of them on all
83 text layers.
The mapper now collects its notes through the same `Warnings` the REST
importer already uses, which keeps the first-seen order and the cap and
appends `(×n)` where a note was met more than once. Nothing else
changes: the note text stays the catalog key it was, and the import
report already reads a `(×n)` suffix off before translating the base
(`ui-egui`'s `note_text`), so a counted note is still translated.
`figcraft-cli info` prints these notes, so it now names each kind with
its count.
With the count dropped again, the test reports the two notes plain:
["connector layers import as static frame/vector snapshots ...",
"connector label was skipped: depth limit prevents nested ..."]
Part of #467.
Found while testing crates/color/src/lib.rs:125
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/doc/src/text.rs:64
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Two things the limit got wrong, both reported by CodeRabbit on #915.
maxBytes was compared with the escaped text block alone, but what goes
on the wire is that block inside a content object, inside the JSON-RPC
result, with the request id and the newline the transport writes: a
block measured at exactly maxBytes left as a larger message than the
caller said it could read. A kilobyte of the caller's figure is now kept
for that wrapping. About ninety bytes of it is fixed and the rest is the
request id, so the figure holds for any id up to roughly 900 bytes --
every id in practice, a number or a UUID. Sizing the real envelope here
would mean handing the transport's request id to every tool, and the
assumption is written down instead, in the schema and at the constant.
A depth: 0 request over the limit returned early, so the one response
that cannot be made any smaller was the one sent without the truncated
marker -- while a bisection that walked down to depth 0 sent the same
tree with it. Depth 0 now takes the same path as the bisection: the tree
goes out, over the limit, with the marker saying so.
Without the first fix the test reports a limit at the exact size of the
tree as no truncation; without the second the depth-0 marker is absent
(left: (None, None), right: (Some(0), Some(0))).
Found while testing crates/engine/src/online.rs:118
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Adds .coderabbit.yaml so CodeRabbit no longer reviews or comments on PRs opened by dependabot[bot].
Co-authored-by: Claude Opus 5.5 <[email protected]>
- Add Polish command labels for arrange.reorder, file.moveTab, pref.startOnHome,
pref.uiScale, text.loadFont, text.refit, and text.rescanFonts.
- Ensures 100% command coverage with zero fallback to English in the Actions palette.
Found while testing crates/collab/src/code.rs:~150 (Local::from_link, u64::from_str_radix on the token)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
- Mark Polish catalog as reviewed and approved by native speaker (@lukasz-sklad)
- Replace "Lekki" with "Jasny" for Light mode
- Translate disambiguated context keys (Fill|sizing, Join|stroke, Light|glass, Open|comments)
- Catch up with 15 UI strings added in recent main commits
- Update docs/translating.md to list Polish as reviewed
A .figcraft, .fig or file JSON dropped on the canvas ran file.import,
which added its pages to the file in front. It now opens as a file of
its own, as on the home screen, in the browser too. Images and SVG
still import at the drop point; File > Import still adds pages.
- pending_open::open_dropped queues a dropped file as file.open; the
home screen and the canvas share it.
- Dragging only documents over the canvas says "Drop to open" (not in
the browser, which names no file until the drop).
- The tab-switch import test drops an SVG: it pinned where an import
lands, using a .figcraft only as a file to import.
MCP stdio is newline-delimited JSON, so one tool result is one message
and has to fit the client's read buffer whole; there is no streaming of
part of one. The TypeScript SDK's stdio transport defaults to 10 MiB and
hosts rarely raise it, so a document whose tree went past that killed
the server: the client stopped reading, the pipe broke, and every later
call in the session failed.
The tree is now reported at the deepest depth that fits, found by
bisection, with a `truncated` object giving the depth asked for, the
depth used, both sizes, the limit and a pointer to get_node for reading
one subtree deeply. `maxBytes` lets a caller that knows its own buffer
say so; the default is 8 MiB, leaving the envelope 2 MiB of the usual
10.
The size is measured as the transport sends it -- pretty-printed and
then escaped inside the text block, which adds about a tenth (342,136
bytes of tree became 379,954 on the wire in the fixture). A kilobyte is
reserved for the marker itself, which is added after the measurement.
A limit below even the shallowest tree still answers, at depth 0, with
the marker saying so: there is nothing smaller to send, and an error
would leave the caller with nothing at all.
The tool description now says that each level of depth multiplies the
response and points at get_node, which the report notes would have been
enough on its own to avoid the crash.
Fixes#912.
Found while testing crates/color/src/lib.rs:165 (also :173)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A component set took its name from the first variant's name up to the first '/' or '=',
so variants named "State=Default, Size=Small" made a set called "State". The set now
takes the path before the last '/' ("Button/State=Hover" -> "Button",
"Icons/Arrow/Dir=Up" -> "Icons/Arrow"), keeps a plain name as is ("Button"), and
otherwise gets the next numbered default ("Component 1"), as a new component does.
- right.rs: drop the unused ACCENT import (main removed it); keep the grid
constants CAPTION_PT/TEXT_PT/TITLE_PT our branch added.
- typography.rs: keep our TEXT_PT size, adopt main's accent_text token rename.
- widgets.rs: keep our TEXT_PT size for the button label; the disabled colour is
now main's text_tertiary from the match, so the old dim() closure is dropped.
- ui-parity.md: keep our row and main's two (#890, #894) changelog rows.
The picker draws on the shared popover, so it closes the way the popover does: the X,
Escape, and a press outside it.
- A press on the canvas closes it; presses in the panel beside it keep it, since a
swatch, a style chip or a field there switches the colour it edits.
- The control that opened it is held out of that closing and its rect is refreshed every
frame, so pressing the same swatch again reaches its own toggle — after the panel
scrolls too (the bug #730 had, in the picker's hands).
- Escape closes it before the binding deselects the layer, and not while one of its own
dropdowns is open.
Tests: a press on the canvas closes it, the same swatch closes it again (panel_remove_tests).
Every section lays its rows out on the same grid: the panel is 240 pt, its content
16 pt in from the left and 8 pt from the right, two 88 pt columns at that width with
8 pt gutters and a 24 pt column for the buttons at a row's end (the columns are flex,
so they share a wider panel); rows are 24 pt in a 32 pt track, captions 10 pt under
them, labels and values 12 pt, the selection's type above them 14 pt. A row's two
buttons (eye, minus) are one group on the row's right edge, 24 pt each and 4 pt apart.
Two things put the right margins out before: block() inset every row by PAD on both
sides (16/16 instead of 16/8), and the panel's frame carried a 1 pt stroke (egui counts
a frame's stroke as its margin), which took a point off each side of the 240 pt panel.
The edge is drawn inside the panel now.
The paint, effect and layout grid rows reserve the whole button group before their
field (8 + 24 + 4 + 24 = 60), so the group lands on the panel's 8 pt margin; the export
preview frame takes the same 16 / 8. Every row that had its own width now asks the
grid: the stroke's align and weight fields split what the row leaves (74 + 8 + 74 with
two buttons, 88 + 8 + 88 with one), the type panel's size field and the full-width
dropdowns take the row's own width, and the Inspect tab's rows end at the 8 pt margin.
The fields' label inset is 8 (was 12) so "Intensity" fits its column whole beside its
value; the glass field's width in field_tests follows the wider column.
Tests: cargo test -p figcraft-ui-egui (549 passed).
Clicking a colour swatch opens the picker; clicking the same one again closes
it (another swatch switches the picker to its paint). The page background, the
text range, the paint rows, the effect rows and the selection colours all
toggle.
The pickers' review fixes on top: the colour picker keeps its popover only
while it is open (a body that closed it left a stale dragged place for the next
picker to inherit); the style picker resets its retained popover when a new
picker opens and on its early-close paths; and Escape belongs to the nested
colour picker first, so closing it doesn't take the style picker down with it.
Opening the style picker, the layout settings or the stroke settings closes the
colour picker instead of leaving it lingering beside them.
`Popover` is the one card detached panels draw on: a header 40 points high
with the title 16 in from its left edge and a 24-point X 8 in from its right,
a hairline under it, the theme's popup shadow, and one slot. Its measurements
are our own — the colour picker's header, which it now uses — not another
product's.
The colour picker and the style picker move onto it in this commit; each keeps
only its own body. The style picker keeps its dark surface through
`Surface::Menu`, and its tools row (view, detach, create) now starts the slot,
under the shared title.
Closing is the widget's now: the X, Escape, and a press outside it — except in
the rects the caller holds out (the button that opened it, a popover it opened
in turn), so that button's own click can toggle it shut instead of the press
closing it and the release opening it again (the bug #730 has). The colour
picker keeps its old habit of staying open while the panel beside it is used.
Tests: press-and-release for the X, for a press on the opener (open) and for a
press elsewhere (closed), plus Escape, the opening rule, the lift and the drag.
The chrome detached panels share, in one place: a 240 pt card with a 40 pt header —
the title 16 in from its left edge, a 24 pt X 8 in from its right — a hairline under
the header and the theme's popup shadow, around one slot:
let mut p = Popover::new("my_popover", tr("Title"));
let closed = p.show(ctx, Open::Anchor(pos), Hold::default(), |ui| { /* the slot */ });
The measurements are the ones the colour picker's header already used, which this widget
now draws; the card is as wide as the picker was. The shadow is the theme's
(`visuals.popup_shadow`), so the dark theme gets its own.
Closing is the widget's: the X, Escape, and a press outside it — except in the rects the
caller holds out (`Hold { rects, escape }`: the button that opened it, a popover it opened
in turn), so that button's own click can toggle it shut instead of the press closing it
and the release opening it again (the bug the fill-rule panel has). Press-and-release
tests pin all of it.
It opens against an anchor (its right edge on the anchor, top on it, lifted when that
would run off the screen's bottom) or at a point; the header drags it.
Six headless tests: the header's geometry, the opening rule, the lift, the drag, the X,
Escape, and the press-and-release of a press on the opener (open) and elsewhere (closed).
The warning still said fixed and content-sized tracks were written as equal
tracks, which the export no longer does. It now reports only the tracks whose
size the format can't hold, and only for the tracks an exported grid has.
Co-authored-by: xiaou61 <[email protected]>
The CLI help and docs/mcp.md described the roots as confining whatever the run
touches, which reads as if they also applied to `mcp --attach`. They don't: an
attached server runs the commands inside the app, and the app confines them
from its own flags. VectorCraft documents the same split, and the missing
sentence is what made the difference confusing, as #835 reports.
Co-authored-by: xiaou61 <[email protected]>
The import has read `gridColumnsSizing` / `gridRowsSizing` since #630, but the
export wrote only equal tracks and warned about it, so a grid with a fixed
column beside two weighted ones came back equal after a save and reload.
Write the sizing map next to the order map, keyed by the same track ids: fixed
px, fractional fill and content-sized tracks keep their sizes. A track the
format has no name for (negative or out-of-range size, or a document track the
frame doesn't list) keeps its id and reads back equal, with the existing
warning.
Co-authored-by: xiaou61 <[email protected]>
Duplicate and Alt-drag copied a main component as a second main. Within
its file the copy is now an instance of it (components::copy_in_file,
shared by edit.duplicate and the Alt-drag copy), as Copy then Paste
already was. A variant duplicated in its set stays a new variant.
Layer data written with node.set {props: {pluginData}} was saved but no
command read it back. node.get (the command, the control method and MCP
get_node) and document.inspect now report pluginData as {key: value}
when a layer or page has any, so a tool keeping state in a file can
read it on its next run. Widget keys are included: the plugin rules
only stop plugins writing them, and widget.list already reports that
state.
Alt+arrows were left out of the nudge keys, so with Alt held to measure
the arrows did nothing and the distances never changed. The arrows now
nudge with Alt held too (Shift: the big step); the hover measurement is
recomputed every frame, so the labels follow each nudge. While measuring
to another layer the selection's handles and size label hide; both
layers keep their outlines.
The dim text token (disabled buttons, placeholders, hints) was 2.1:1 on
the light theme's panels; disabled buttons also faded to 40% opacity,
and selected text was drawn in the dark green, 1.4:1 on the dark
theme's selection highlight. The dim text token is now at least 3:1
against panels, fields and selected rows in both themes while staying
dimmer than secondary text, a disabled button is grey with that text,
and selected text takes the theme's readable green. A test checks the
ratios with the colour contrast helper.
The browser has no system fonts, and since releases moved to the shared
workflow the web build was made without craft-fonts: no font list and no
text drawn. The Georgian change also re-pinned craft-fonts to its main,
which has no fonts/latin-manifest.txt, so the next release would fail
under CRAFT_FONTS_REQUIRED. Pin the Latin commit again until one commit
has both (craft-fonts#42); desktop Georgian uses an installed font.
Presentation tabs now take at most half the free strip while file tabs share it.
They shrink to the minimum tab width first, then the rest go into an overflow
menu, so they never reach the new-tab button, theme toggle or window buttons.
The close X has its own hit rect next to the tab body and is registered every
frame, so a click on it can no longer select the tab instead.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review follow-up. The closed-owner test now picks a device and switches
Responsive in the presenter and checks that the file in front keeps its
prototype settings, so removing either `live` guard in present::show fails
it. The present harness gets frame_output for the AccessKit tree. The new
tab functions get doc comments.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Main plays a presentation in a single "Present" tab. Now each page of a
file gets its own tab, named "file - page", and the name follows
renames. Play on a page that has a tab shows that tab as it was; from
another start frame it starts again. A tab keeps its own copy of the
prototype: it plays on after its file closes, and a saved file opened
again gets its tabs back. A tab whose file is closed or not in front
never reads or changes the file in front (device, responsive,
comments). Escape or the tab's X closes that tab and goes back to its
file, or Home.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Configure server write loop in hub.rs with max_server_message so outgoing ServerMsg payloads above 64 MiB (e.g. document snapshots in Welcome messages) are delivered without socket termination.
- Set core max_message_bytes in local live session HostConfig to max_server_message.
- Map tungstenite Capacity errors in guest read loop to Outcome::Failed with accurate size limits rather than Outcome::Refused which falsely reported 'The host refused'.
- Add loopback tests verifying Welcome delivery above client message limits and accurate diagnostics on client capacity overrun.
CodeRabbit: typing "Auto" in the count of a stretched grid set 0, which the setter clamped to 1. The count field only takes the word where 0 is Auto (a fixed-size grid); a stretched grid's count is a plain number.
Review fixes for #881 (FredApps, CodeRabbit).
- number_field: the word for Auto (the placeholder's own text in the UI's
language, or "auto") is taken only by a field that asks for it (`Empty::Auto`),
not by every field whose placeholder reads "Auto". An auto layout's gap
("Auto" is Space between) and a prototype overlay's offset ("Auto" is no
offset) keep their own meaning of 0; the fixed "自动" match is gone.
- layout: Auto width has no hidden 60 px band — the bands step by the gutter
alone, so the pitch is what the panel shows. A zero-width band snaps to one
line, not two identical edges.
- right: the width field shows "Auto" (not 0) when there is no width, and a
stretched grid shows it dim and not editable (its bands divide the frame),
with the margin and gutter still editable.
- tests: the field suite checks the word on a grid's width (0) and on an auto
layout's gap (nothing); the layout suite pins the gutter pitch, the Auto
count and a stretched grid's bands.
- docs: ROADMAP and ui-parity state the pitch and what changes for saved grids.
Columns and rows can now set their line width to Auto: a width of 0 (or
typing the word "Auto" / "自动" in the field) draws a one-pixel hairline
down the band's middle, like the square grid, instead of a filled band.
- grid_lines: a non-stretched grid with size <= 0 emits a hairline at each
band's middle rather than a zero/negative-width rect.
- canvas: a degenerate (zero-width or zero-height) grid rect is painted as a
1px line; a filled band is still painted as a rectangle.
- number_field: a field whose placeholder is "Auto" also accepts the word
"Auto"/"自动", mapping it to 0.
- right: the grid size field shows "Auto" and the align menu no longer
forces a 60px band on a zero width.
- props: document the Auto semantics of LayoutGrid::size.
- ROADMAP.md / docs/ui-parity.md: note the change.
- layout tests: regression test for the Auto hairline.
Found while testing crates/engine/src/live.rs:85
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Found while testing crates/collab-net/src/invite.rs:242 (percent_decode)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Without craft-fonts the web build has no Georgian face and can't read installed fonts, so the
language picker and System must skip Georgian there, as they already skip Japanese, Korean and
Chinese without their faces.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
Found while testing crates/color/src/lib.rs:63 (clamp01, used by Color::new via parse_css at :155-166)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Launching opens a new empty file, which then has to be closed to get
back to Home. Preferences ▸ "Start on the Home screen"
(`pref.startOnHome {on?}`, off by default) starts the desktop app on Home
instead. Files named on the command line and recovered files still open.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
An explicit edit of a bound width, height, opacity, gap, padding, corner
radius, stroke weight, arc or text field reported ok, but normalize wrote
the variable's value straight back. Now the edit sticks and that field's
binding is dropped, as paint edits already do.
- `styles::bound_before` / `detach_edited`: compare each layer-level
bound field's stored value before and after the edit (`bound_field`),
so keys written unchanged keep their bindings and fields moved along
with another (proportions, min/max) lose theirs. A binding the edit
itself made (`boundVariables`) stays. List and component property
bindings are left to their own rules.
- Hooked into node.set, transform.resize, object.opacityKeys, text.style
and `Session::preview` (canvas drags of the selected layers).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Adds crates/ui-egui/locales/ka.toml (ქართული): every key of en.toml (1,730 UI
strings), every command label (567) and every menu wording (31), marked as a
draft. Georgian has no letter case and keeps a noun singular after a numeral,
so plural entries give the same text for one and other. System picks it for
ka, ka-GE and ka_GE.UTF-8; pref.language's help text lists ka.
Inter and egui's fonts have no Georgian, so the UI draws it in craft-fonts'
Noto Sans Georgian: it joins the built-in faces in theme::install_fonts and
the web build keeps it (crates/text/build.rs). craft-fonts-ref moves to
a71c20e3 (storytold/craft-fonts#27, the commit that adds the face) in the
release and FreeBSD workflows, and the release builds the web version with
craft-fonts (web-craft-fonts: true).
Tests: the Georgian locale tags, draft status and count-neutral plurals, and
that the built-in Georgian face draws every Georgian letter in the catalog
(skipped without craft-fonts). Docs: localization-parity, translating, mcp,
README and ROADMAP.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
Found while testing crates/collab-net/src/invite.rs:valid_host (around line 205)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
`caret` put the index right after a line break (a newline, or a line
separator) at the end of the line above, so after Enter the caret showed
on the previous line until something was typed, and Up/Down counted
lines from there. The index before the break stays at the end of its
line; the one after it is now on the next line, as after a wrap.
Selection highlights already started such lines at their own start
(#665); their comment no longer describes the old caret.
After a click in the Pages list, layers selected since (Select all, say)
took the first Delete, which left nothing selected, so a second Delete
deleted the page. Delete that doesn't take the page now ends the Pages
list's hold on the keyboard.
The Pages list kept the keyboard after a switch to the Assets tab or to
another open file, so Delete with nothing selected deleted a page there.
It now deletes a page only while the list that was clicked is on screen
(File tab, list expanded, panel and interface shown) and its file is the
one open.
Figma stores Shift+Enter as U+2028. `.fig` imports keep it, but layout
only broke lines at a newline, so imported text ran soft-broken lines
together ("Option 1SL: …"), lost the empty line of a separator followed
by a newline, and put glyph bullets on one line.
A line separator now ends its line without ending the paragraph: no
paragraph spacing and no new list marker after it, the next line keeps
the item's indent, and it draws no glyph. A trailing one leaves an empty
last line, and a line it cuts short isn't stretched when justified.
Carets, clicks and selections asked `para_end && !last_line` to mean
"the line ends with a break char". Record that on the line instead, so a
break that doesn't end the paragraph can use the same rules. No
behaviour change: today only a newline ends a line early.
Clicking a page row (or the add-page button) gives the Pages list the
keyboard: Delete or Backspace with no layers selected now deletes the
current page instead of failing with "nothing is selected". No
confirmation; a "Page deleted" toast says so and Undo brings it back.
The context menu's Delete page shows the same toast. Selecting layers,
clicking a layer row or pressing on the canvas gives Delete back to the
layers.
An instance copied its main's variable modes when it was made and kept
the copy as modes of its own. Switching it to another variant kept the
old variant's modes, and a mode set on the main later never reached it.
Instances now take the modes their main sets through the link
(`Document::inherited_modes`), closer than any frame around them. Modes
set on the instance itself still win and stay through variant switches.
Bindings are re-applied for the instances of a main that chose another
mode and for an instance that switched mains. Detaching keeps the modes
the instance took from its main, and the Variable modes panel shows them
as the instance's Auto value.
Instances made before this change keep their copy as modes of their
own; setting one back to Auto lets it follow its main.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
node.set refuses unknown properties, but the effects inside it were read
by serde, which drops keys it doesn't know. A drop shadow given
offsetX/offsetY (the names variable bindings use for its offset) was
stored with the default offset [0, 4] and the call said ok. effect.add
and effect.set dropped them the same way.
The three now refuse what the effect doesn't read and say what it
takes: "effects[0]: unknown properties: offsetX, offsetY (a dropShadow
takes blend, color, offset, radius, showBehind, spread, type, visible)".
A key counts as read when setting it to null makes the effect fail to
parse or come out different, so serde's aliases count as read without a
list of them here.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
component.makeSlot adds its slot property through add_prop, which named
itself in every error: making a second slot called "Content" said
"invalid parameters for `component.addProperty`: a property named
`Content` already exists", a command the caller never ran. add_prop now
takes the command errors name, the fix#807 makes for wrap.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
An assignment and the layers that read it can name different copies of
one definition (the set's own id on one side, a variant's copy on the
other). The values were kept under the id the assignment names and read
under the id the layer names, so such a value never reached its layer.
Both sides now go through prop_root, the definition the copies were
made from (CodeRabbit on #847). In Untitled UI, 143 assignments name a
copy that no layer reads while another copy of it is read.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A layer with individual stroke sides (a header with only a bottom divider,
strokeSides [0, 0, 1, 0]) came out of the CSS, HTML, React and Leptos
code as `border: 1px solid ...` on all four sides, because the border
was written from the stroke weight alone. With individual sides it is now
a border-top/right/bottom/left for each side that has a weight, and a
side bound to a variable names it (border-top: var(--border-top, 3px)).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A component property assignment carries its value in `value`, and in
current files also in the variable form `varValue`. Many assignments
leave `value` empty and say it only in `varValue` (booleans as
boolValue, instance swaps as symbolIdValue, text as textDataValue), and
those were ignored: a hidden layer showed and a swapped icon stayed the
default. In Untitled UI that is 1,643 booleans and 185 swaps, so inputs
showed their hint text, help icon and leading icon on top of the
neighbouring layers. Both forms are read now.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
When a path can't be resolved (a link that leads nowhere, `..` below a
missing folder), the reason is given only for paths inside a root.
That check compared the path as typed with the roots resolved, so a
root reached through a link (`/tmp/work` on macOS is
`/private/tmp/work`) called its own dangling links "outside the folders
that may be written: /private/tmp/work", and a link out of a root still
had its reason told. The existing part of the path is now resolved
before the comparison, as the roots were.
This is also why `links_that_lead_out_of_a_root_are_outside` and
`outside_paths_are_refused_alike_whatever_exists_there` failed on
macOS, where the temp folder is under `/var` -> `/private/var`.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Each variant of a component set keeps a copy of the set's property
definitions (an id and a parentPropDefId, no name or type), and the
variant's layers and its instances' assignments use the copy's id. Those
copies were skipped, so imported variants lost every boolean, text and
instance swap binding, and instances lost their property values. In
Untitled UI (8,074 copies) an input's "Hint text" no longer followed its
property. A copy now names the definition it was made from.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A mistyped or unsupported flag (e.g. --automation-write-roots) used to be
dropped silently, so the app started unconfined. It now exits 2 with
'unknown option', like a bad automation root. A bare -- ends the options.
Adds a desktop control-channel test that file.save / file.saveAs outside the
write root are refused and create nothing, while saves inside succeed, and
documents that the root flags arrived in 0.7.0 (0.6.0 ignores them).
Routing the CLI's writes through `file_access` changed what happens when the
output path is a symbolic link. `file_access::write` stages the bytes and
renames them into place, and a rename replaces the link with a regular file,
where `std::fs::write` followed it and wrote to its target. A render to a link
reported success while the target kept its old contents.
Unconfined writes go back to `std::fs::write`. Confined ones keep the staged
write, which is right there: `Roots::check` resolves the path first and has
already refused a link leading out of the root, so the write lands on the
resolved target inside it.
Found while testing crates/engine/src/cmd/styles.rs:119-124
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
`figcraft-cli mcp` takes `--automation-read-root` / `--automation-write-root`
to confine the files an agent's commands touch, but the file subcommands took
neither: `figcraft-cli run script.json --out /anywhere/x.figcraft` read and
wrote as the user who started it. The README invites agent use, and a script
or agent driving `run`, `export` or `convert` had no way to say which folders
it may reach (#833).
`info`, `export`, `convert`, `render` and `run` take the same two flags, with
the same semantics: repeatable, the folders must exist, reads inside a read or
write root, writes only inside a write root, everything else refused -- so
giving no write root denies writing altogether. Without the flags they behave
as before.
The flags would have been decorative on their own: `figcraft_engine` enforces
roots through `file_access`, which the engine's own loads already go through,
but the CLI wrote its output with `std::fs` and read its script the same way.
Those three calls now go through `file_access` too, so the confinement reaches
the files the CLI itself touches.
ExportFormat::VectorDrawable ("XML", .xml) and ExportSetting.vd_compat
(bestEffort / exact / ignore, stored only when not the default) wire the
figcraft-vectordrawable writer into file.export, file.exportLayers and
export.add/update: a vdCompat param (junk refused), svgIdAttribute writes
android:name, and the results carry the writer's warnings (per file for
export layers). The Export section lists XML with an "Android Vector
Drawable" tooltip and its id attribute and compatibility options, File >
Export... gets an XML preset, and the toasts say when content was changed
or left out. CLI: --format xml / .xml outputs, --vd-compat, warnings as
note: lines. MCP: format xml returns the XML as text plus its warnings.
A .fig export writes XML settings as SVG with a note.
Docs: file-format-parity (row and revision), ui-parity and roadmap
revisions, architecture (the crate), cli and mcp.
A new L4 crate that writes layers as an Android Vector Drawable (XML):
shapes, booleans and text as baked path data, solid and gradient fills
(linear, radial and sweep, exact for any handle frame via a group and a
2x2 SVD), native strokes where Android draws them the same and filled
stroke outlines otherwise, clip frames and masks as <clip-path> groups.
What a drawable can't hold is governed by Compat (bestEffort / exact /
ignore): layer opacity over overlapping shapes, even-odd fills and clips
(rewritten to non-zero via pathops), diamond gradients, alpha masks;
effects, image and pattern fills and blend modes are left out with a
warning. Output is pretty-printed, numbers are finite and short, names
(android:name) are optional, escaped and unique.
frame_and_text_marks_are_inset fails if either mark reaches the box edge
again; slice_is_one_connected_knife fails if the handle stops short of the
blade's base or the icon goes back to separate marks.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
File ▸ Open, Import and Place image did nothing in the browser (the pickers
return a path, which a browser never has) and dropped files were ignored
(eframe hands their contents over asynchronously).
- Services::pick_open_contents: a platform can pick a file's name and
contents; the web build uses an <input type="file">. Open and Import pass
them to file.open / file.import as data + name.
- A dropped file without a path waits behind the opening card until its
contents are read (bytes_async on wasm), or says it couldn't be read.
- Canvas drops still import, as on the desktop.
Closes#759.
Co-authored-by: Cursor <[email protected]>
Found while testing crates/engine/src/codegen.rs:865 (swift_layer) and :1038 (compose_layer)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The closed Spacing row paired each icon with the other axis: the
side-bars field wrote top/bottom and the top/bottom-bars field wrote
left/right. The H field now binds paddingHorizontal (left+right) and the
V field paddingVertical (top+bottom); the PaddingH/PaddingV paintings
were swapped to match so every icon is pixel-identical.
(shut_pair_fields_edit_their_own_axis).
From review: the three-field models keep the values last typed while the
colour is still theirs, so a chroma lowered to fit sRGB or a hue typed on
a grey isn't lost on the next frame. One table per model (Polar) with its
own field ids. lch()/oklch() read deg, rad, grad, turn and none through
one helper, with a junk sweep. lch.rs reuses the sRGB transfer functions
from profile.rs.
The Spacing row's toggle opens per-side Top/Right/Bottom/Left fields in a
2x2 grid in place of the H/V fields, each bound to its own padding key with
one undo step; shut H/V fields show both values ("16,54") when the sides
disagree instead of "Mixed". Original code-drawn side icons, English and
Spanish strings (padding_individual_tests).
Every mark in the bottom toolbar is now painted in one 18-point box. The
mode switcher drew its marks in a 14-point box, so Draw, Prototype and
Inspect read smaller than the tools. The flyout carets move from a
10-point box, whose stroke rounded to one physical pixel thinner than the
row's, to a 14-point box that rounds to the same width at 1x, 1.5x and 2x.
Frame and Text are inset by a unit: their long pixel-hinted strokes read
heavier than the anti-aliased outlines beside them. Slice is a knife (a
slim blade and a solid handle), one silhouette instead of a box crossed
by a diagonal.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- text.rescanFonts offers a loaded family the files it hasn't read (styles
installed since) and refits every family that gained faces; `added` still
lists only the families that were unknown
- text.refit refuses an `ids` that isn't a list instead of refitting the
whole document
- text.loadFont with a file already loaded returns its families without
keeping its bytes or adding its faces again
- doc comments on the functions this change touches
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The same JSON methods as the desktop control server (engine.execute, ui.inspect, ui.screenshot, …),
as promises on window.figcraft, like FilmCraft's window.filmcraft: an agent driving the editor in a
browser tab or an <iframe> uses the same commands as the CLI and the MCP server. openFile opens
bytes the way a dropped file opens; captureDownloads keeps saves in the page for readFile.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A text box keeps the size it was measured at, so text laid out while its
family was missing kept the fallback font's box after the font arrived (a
hugging chip clipped its label).
- text.refit {ids?} measures auto-sizing text again (whole document by
default) → {refitted}
- text.rescanFonts and text.loadFont refit the open document's text in the
families they add, as one undo step
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The system fonts are indexed once per process, so a font installed while
FigCraft (or `figcraft-cli mcp`) runs stayed missing until a restart.
- text.rescanFonts walks the font folders again and offers the families
installed since (loaded families keep their faces) → {added}
- text.loadFont {path | data} adds a font file without installing it → {families}
- both clear the layout cache so text lays out again in the new face
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- PluginEntry no longer carries save_file: postcard is positional and existing
messages keep their fields (docs/collab-protocol.md), so mixed app/server
versions keep reading the directory. Manage plugins still shows "can save files".
- save_file checks the data length before reading it and refuses with status 1
and the reason, as documented, instead of failing the run with a limit error.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The picker's model menu offers LCH and OKLCH beside Hex, RGB, HSL, HSB
and CSS, on desktop and web. Lightness, chroma and hue fields; an edit
outside sRGB lowers chroma and keeps lightness and hue (CSS gamut
mapping). The CSS field and Color::parse_css read lch() and oklch().
edit.copy and edit.cut mirror a selection made only of text layers to the system clipboard as plain text, one layer per line, instead of SVG. Other selections stay SVG; pasting back into FigCraft still pastes the copied layers.
Closes#808
Co-authored-by: Cursor <[email protected]>
layout.remove and node.set {layout: null|"none"} now drop the bindings of auto layout fields (gap, crossGap, padding...), as frame-to-section already did for some of them (it now drops all of them). Re-applying a binding no longer creates an auto layout on a frame without one: the binding waits for the next auto layout.
Closes#810
Co-authored-by: Cursor <[email protected]>
Polish integer plural rules use Plural::One strictly for n == 1, while
higher counts ending in 1 (e.g. 21, 101) use Plural::Many.
Also adds unit tests for Polish integer plural coverage.
Co-Authored-By: Antigravity <[email protected]>
Escape closed the whole Preferences dialog even while a menu was open over
it. That held while Preferences had no menu of its own; the Interface size
menu is the first, so one Escape threw away the dialog with the menu.
Dialog keys now leave Escape to an open menu, and the next Escape closes the
dialog as before.
Adds a keyboard test of the Interface size menu: Tab reaches it, Enter opens
it, the arrows move through the sizes, Enter picks one and closes the menu,
and Escape closes only the menu.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Draws the interface the way a real window does (a fixed number of physical
pixels, its size in points being that divided by the scale drawn now) over
10 window sizes, 6 system scales and every interface size. Guards the rule a
fixed size promises: above 100 % the window still holds the smallest layout,
the panels leave room for a canvas, and the size settles without flicker.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Preferences > Interface size and the `pref.uiScale` command make the whole
interface larger or smaller. Auto keeps the system's display scale; a fixed
size from 75 % to 300 % is absolute (100 % is one screen pixel per point on
every system). The size is held back while the window is too small for it.
Strings are added in en, es and fr; the control protocol, MCP, UI parity and
roadmap docs are updated.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
#783 completed the French catalog to every current string, superseding
the translations added in #681. This keeps only the ~50 UI strings and
8 command labels whose wording a native French speaker corrected during
that review (keyboard focus, "Off" as "Désactivé", blend modes, stroke
join, spotlight, the no-break space before : ; ? !…), applied on top of
#783. The draft status is unchanged.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
* docs: drop references to competing commercial software
* docs: clarify parity checklist coverage in status callout
* #786: say interface design once in the tagline
---------
Co-authored-by: bflatastic <[email protected]>
Add full Polish localization for FigCraft UI, commands, and menus.
Includes Polish CLDR integer plural rules (one, few, many, other)
and eliminates all untranslated strings and borrowings.
Co-Authored-By: Antigravity <[email protected]>
Both reach wrap() (a page in a mask, a section outlined into a group)
and named `object.group` before.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
fill.set, stroke.set and node.set fill/fills/stroke/strokes reported ok
while normalize put the bound colour straight back, and node.set
{fills: []} brought the bound paint back. They now drop a paint's
colour, opacity or stop binding where the edit changed that value, as
selection.setPaint already did (its comparison moves to
lists::rebind_edited_paint, which also lets stop bindings follow a
removed stop). boundVariables applies last in props::apply, so a
binding given with the paints in one call stays.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Grouping, framing, auto layout, masks, outline stroke and the boolean
commands share wrap(), which named `object.group` in every refusal
(`layout.add` on a page: "invalid parameters for `object.group`").
wrap() now takes the calling command's id.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A negative width or height was clamped to zero by the scale helper, producing a layer smaller than asked for; vector.create now says so. Regression test covers width -10 with height 8.
Review follow-up. A file written by this version carries editorOutlines, so a stroke someone chose is kept even when it matches the old decoration; only files without the marker get the one-time legacy cleanup. vector.create now errors when a requested side cannot scale the geometry or when both axes would collapse, instead of creating an invisible layer.
A path's intrinsic box replaced the width and height asked for, so an icon imported at 24 stayed on its source grid. Scale the geometry to the requested size; with one side the other keeps its proportion. A path or point set that draws nothing is an error now, not an invisible 0x0 layer.
Node::new gave every component set a purple dashed stroke, so export and screenshot drew the editor frame into the image. Sets now start with no paint; the canvas draws the dashed frame itself, and files written by earlier versions lose exactly that stored decoration when opened. A stroke someone chose is kept.
Collect the folders whose lock is held, wait 100 ms once and retry only those,
instead of up to 100 ms per held folder. Recovery::start takes its folder without
waiting again (a held name already falls through to another). Windows doesn't
pass the lock to child processes, so the wait and its test are Unix-only; the
test holds the lock 10 ms against the 100 ms wait.
Fills, strokes and effects already merged item by item, so both sides
changing different items of one of those lists was no conflict. Every other
list conflicted as a whole: both sides editing different export settings, or
different guides, or different prototype interactions, or different
accessibility annotations or Dev Mode links, each had to be resolved by hand
for the whole list, though the items stand on their own.
Export settings, guides, prototype interactions, accessibility annotations and
the Dev Mode code links and resources now merge item by item, with the same
rule: the same item changed both ways, an item removed on one side, or
different items added on both sides still conflict.
Properties that are keyed maps rather than item lists (bound, varModes,
overrides, propRefs, pluginData) are deliberately left out: their items line up
by key, not by position. Text runs are left whole too: they live in the node's
text data, so merging them item by item means splitting a property the change
format keeps whole.
Replaces the test that documented the gap (other_lists_still_conflict_as_a_whole)
with one that checks each new list merges, and adds one for the same item
changed both ways still conflicting.
A child process forked before its exec inherits the run's open descriptors,
and a flock lasts while any descriptor for the file is open. A crashed run
whose folder lock was still held by such a child read as a run that was still
going, so recover_all skipped it and its unsaved documents were not opened at
the next start. The desktop app's tests failed intermittently for the same
reason whenever another test spawned a process in parallel (the reduced-motion
read spawns gsettings on Linux and the BSDs): recovery_tests failed in 23 of
60 runs of the whole binary and never on its own.
Taking a folder's lock now retries try_lock briefly, so a lock that frees
itself is no longer read as a live run while a run that really is running
still keeps its folder.
Regression test: a_lock_freed_a_moment_later_is_not_a_run_that_is_still_going
holds the crashed run's lock on a second descriptor, the way a forked child
does, and frees it shortly after; the copies come back. The recovery tests
that collide with a spawning test went from 8 failures in 10 runs to 0 in 10.
- **Quit (macOS):** `on_exit` now calls `services::shutdown` and then, on macOS only,
`std::process::exit(0)`. This happens before eframe tears the windows down, which is where the
crash in #781 is thrown (same approach as the sibling apps, tracked in craft-libs#3).
- **`services::shutdown(app, dir, recovery)`** replaces `save_state`. It runs on every platform and
everything in it is finished before it returns:
1. the preferences (`state.json`) are written atomically, as before;
2. this run's recovery copies are settled by the new `Recovery::quit(session, asked)`:
- if the unsaved-changes question was asked (`ui.quit`: last window closed through the
question, or `app.quit`), or nothing is unsaved, it works as before: `finish()` removes
them;
- otherwise the unsaved documents are copied (waiting for any copy still being written)
and the copies stay for the next start, as after a crash. This is the macOS Quit menu
(Cmd+Q): winit's menu calls `terminate:`, which reaches `on_exit` without asking. Until
now that path removed the recovery copies of files that had unsaved changes;
3. live sessions end explicitly: a session this app hosts tells everyone and stops; a joined
session is left. Before, this only happened in the app's thread-local destructor at
process exit, if at all.
- If the app can't be reached in `on_exit` (it's busy), the recovery copies are left on disk for
the next start rather than removed.
Quit paths checked:
- **Cmd+Q / the menu bar's Quit (macOS):** `terminate:` → winit `applicationWillTerminate` →
eframe `exiting` → `on_exit`. Covered.
- **Last window closed** (close button, title bar Close, Dock/Finder quit Apple event → `Close`):
`on_close_request` → `begin` sets `ui.quit` → root `Close` → `on_exit`. Covered.
- **`app.quit` (control channel / MCP):** sets `ui.quit` and closes the root window → `on_exit`.
Covered. Its reply goes to the connection thread before the window closes. The process now ends
sooner after that, so a client that reads very slowly could in theory see EOF instead of the
reply (not seen).
- **The frame guard (panic):** isn't a quit. It writes recovery copies and the app keeps running,
so it doesn't need this. A panic outside the guard aborts as before.
- **eframe error mid-run:** eframe also calls `on_exit` before `run_native` returns `Err`. On macOS
that error now only reaches the log (eframe logs it), not the "window couldn't open" box. Startup
failures still show the box: `on_exit` only runs once the app exists.
Part of #781.
The 91 UI strings, 12 command labels and 2 menu wordings the strings added since the French catalog was drafted left in English are translated in the catalog's conventions (NBSP before : ! and inside guillemets, one/other plurals, 0 singular): starter templates and the Templates tab, the home screen's recents table with its sorting and dates in the French way, dropping files to open or import, the Getting started tips with sharing by code, the Playground's bound-variable chip, smart-selection spacing, outline mode's options, the converting and panel-width commands. Appended to each section; still requires native review.
When an interaction noodle or interaction editor was selected in the
Prototype tab, pressing Delete or Backspace (or executing edit.delete)
targeted the selected layer/component instead of deleting the prototype
connection.
- Intercept Backspace and Delete in shortcuts::key_event and run_binding
when an interaction is selected in the Prototype tab to remove the
interaction using proto.interaction.remove rather than deleting the layer.
- Ensure clicking on empty canvas or outside noodles/handles deselects the
active interaction noodle/editor.
- Close the interaction editor when navigating away from the Prototype tab.
- Add regression tests in proto_editor_tests verifying interaction deletion
with Delete/Backspace keeps the component intact, and Escape closes the
interaction editor before deselecting the component.
From review:
- ru and pt-BR "Intensity" were too long for the glass field and ended
in "…"; they are now "Яркость" and "Brilho", which fit.
- A value too long for its field ends in "…" instead of being clipped,
so "100000" never shows as "10".
- A label cut short shows whole on hover.
- The every-language test now checks the Intensity label fits whole in
each catalog (it fails with the old ru translation); the panel test
checks the label isn't cut and that the field is the 98.5 points the
language test measures against.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
While drawing, the Pencil previewed its stroke in the theme's text
colour, 2 points wide: white in dark theme, so over a white shape
nothing showed until the mouse was let go. The preview now uses the
colour and weight the Pencil draws with (one constant, PENCIL_STROKE,
black at 2 pt), scaled by the zoom, at least 1 pt. It doesn't show pen
pressure or a brush, which only the finished vector has.
Refs #256 ("The 'draw' tool is not visible when you are drawing. You
can see it after you release the mouse.").
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The glass effect's light intensity field had no label, and "Refraction
80%" and "Dispersion 50%" ran into the edge of their fields. The field
is labelled "Intensity" in every catalog. Text labels of number fields
take their measured width instead of 7 points per character, and a
label too long for its field ends in "…" before the value, which is
clipped to its own area. This changes every text-labelled number field:
Latin labels sit a little tighter, CJK labels get their real width.
Refs #256 ("There is one parameter that is not signed in the glass
effect editing options. The inscriptions also go beyond their
borders.").
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
reverseLoop reversed the list and skipped shared segments, which only turns a
loop of three or more round: a one-segment loop has no order at all, and a
two-segment ring reads the same whichever way its list is turned, so both
answered reversed: true without moving.
Three or more segments: reverse the list and leave every stored end alone (a
shared one belongs to the neighbour too). One or two: store one segment the
other way round, skipping any segment another short loop lists; when none is
left to swap, refuse with a clear error and leave the document as it was.
Tests: two 2-segment areas over one line that share it, the shared segment
listed first in one area and last in the other; each reversal flips that area's
winding sign and leaves the shared segment and the neighbour alone. A shared
one-segment loop is refused. The 4-segment squares stay as the long-loop case.
Drop the two catalog.rs menu lines: they aren't in the reference app's menus,
and the Vector menu runs commands without parameters, so both would fail there.
Two vector commands behind the fill-rule panel, split out so they can land
on their own (the panel is held for a maintainer design decision):
- vector.fillRule {ids? | id?, region?, rule: nonZero|evenOdd} sets the
winding rule a filled area fills by (the whole layer when region is left out).
- vector.reverseLoop {ids? | id?, region, loop} turns one loop round, which is
how a hole is made and unmade.
reverseLoop no longer swaps the stored ends of a segment a loop shares with a
neighbour: reversing the loop's list already flips this loop's traversal of the
shared segment, so flipping its segment would turn the neighbour's area round
too (bug called out in review). A new test builds two areas that share a segment
and checks the shared segment and the neighbour's fill are left alone.
Adds the commands to the Vector menu (catalog/parity), their locale labels, and
vector_fill_rule_tests (rule per area or whole layer, junk refused, shared segment).
variable.export takes format "dart", and Local variables > Export...
offers "Dart constants for Flutter...". The file has an abstract final
class per collection and mode (the default mode named after the
collection, the others after the collection and the mode) with a
static const Color, double, String or bool per variable.
Names become camelCase Dart identifiers, kept apart when two read the
same, with reserved words escaped; strings are escaped for Dart;
aliases are resolved to their value in that mode, so each class stands
alone. A generated file passes the Flutter analyzer and its constants
work in const widgets.
Plugins can hand the host a file name, MIME type and bytes through the new
`save_file` import (plugin API 4, `permissions.saveFile`). The host asks the
user where to save (browser download on the web), so the plugin never sees a
path. Includes the SDK wrapper and the `export-names` example plugin.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Assisted-by: Claude (Anthropic)
- Code generation gains a Flutter target (ours; the incumbent generates CSS, iOS and Android only), which makes Dart the seventh language: `language: flutter` (with `dart` as an alias, so `CodeLang::from_id` maps both, `id()` is `flutter` and the label `Flutter`).
- `inspect.code` writes a Dart widget tree for the layer: a `Container` for a box, with its size, padding and a `BoxDecoration` (colour, radius, border, shadows); `Row` / `Column` for auto layout, with `spacing`, `crossAxisAlignment` and `textBaseline`, and `stretch` wherever a child fills the cross axis; `Expanded` for a child that fills the main axis; `Stack` with `Positioned` for free children; `Text` with a `TextStyle` (family, size, weight, style, height, letter spacing, decoration and alignment); `Image.asset` for images; `Opacity` for layer opacity; `SizedBox` for a capped subtree. Colours are `Color(0xAARRGGBB)` and `$` is escaped in string literals, as in Compose.
- `inspect.componentCode` writes a `StatelessWidget` for a component or set: properties become constructor parameters with their defaults, a variant property becomes a Dart enum (lowerCamel members, never a keyword), a boolean a `bool`, a text a `String`, a swap a `Widget` and a slot a `List<Widget>?`. Instances are calls that pass only what they change, and an instance's snippet puts the `import 'package:flutter/material.dart';` line before the call, since Dart's imports come first.
- Code connect takes Flutter too: `codeConnect.set {language: flutter}` stores an import path, a name and a template, and the snippet carries a Dart `import '…';` line. The default template passes named parameters (`Card(size: {size})`), as SwiftUI does, and import paths are checked as Dart imports (letters, digits, `_ . / : -`).
- The Inspect panel's language picker iterates `CodeLang::ALL`, so Flutter appears there and in the Connect to code note; the 13 UI catalogs and the component playground's language list follow.
- Docs: README, ROADMAP, docs/mcp.md, docs/gaps.md, docs/components-parity.md, docs/file-format-parity.md, docs/target-app-parity.md and docs/dev-mode-parity.md.
- Tests: Flutter cases in the codegen, component codegen and inspect behaviour suites.
Fonts activated through Adobe Fonts rendered in the fallback font without
a warning: Creative Cloud keeps them in its own folder as hidden files
named by number, which the font scan never read. The macOS scan now
includes ~/Library/Application Support/Adobe/CoreSync/plugins/livetype,
and files without an extension are indexed when they start with a font
signature. Fonts are read in place.
Closes#298
ReadOne replies with a variant body (signature v); deserializing it
straight to u32 (signature u) is a signature mismatch that zvariant
rejects, so the initial colour scheme never arrived and system mode
kept its light fallback. Take the Value and read the uint32 out of it.
A filled area of a vector network can now carry its own paint stack. A region without paints of its own keeps drawing with the layer's fills, and a region with an explicitly empty list draws nothing, so the two cases stay distinguishable. The canvas, SVG and PDF output, boolean and hit-test geometry, vector editing and region operations all honour the per-region paints and keep them when regions are split, merged or reordered.
The .fig importer resolves the individual paint stacks of a layer's filled areas through its normal paint mapping and hands them to the document builder as private mapper data. Public REST vectorNetwork JSON never consumes that data and its region paints are dropped, so only validated .fig input can set them. Invalid entries are skipped with an import note. Live style and variable links on individual regions and writing region paints in the experimental .fig exporter are not part of this change.
winit 0.30 reports no system theme there (EventLoop::system_theme is
None), so the "system" preference resolved to light every time. When
the windowing stack gives no theme, the interface now follows the
freedesktop colour scheme instead: read once and watched live through
the settings portal (org.freedesktop.appearance / color-scheme) in the
desktop services (Services::system_dark), so GNOME's and KDE Plasma 6's
appearance setting switches the app without a restart. Where the OS
theme is reported (macOS, Windows, web) it still wins, and without any
answer the light fallback stands.
A glass layer's coverage, distance field and normals depend only on its
outline and the zoom. A shape asked for again in a later render (the
next frame of a drag, a slider drag, a canvas patch, an export of the
same view) is now worked out whole, in its own pixel frame (translation
reduced to its fraction, in 1/65536 px with carry), and kept: renders
reuse it instead of redoing the distance transform and normals for
their crop.
A shape asked for only once (each step of a zoom, a pane dragged by
fractions of a pixel, a resize) is still worked out for just the part
shown, so those don't get slower: working out the whole pane when only
a corner is in view cost 1.8x before this rule.
The cache is bounded in bytes (128 MB, 32 MB on the web), keeps the
least recently used last, skips shapes over 4M px (1M on the web), and
parallel bands share one computation (OnceLock slot). Renders count
once per render_view / render_view_single; bands don't.
Output: digests of 30 renders (5 looks, square and rounded, scales 1,
1.37, 2) are identical to main; a cut through a rounded pane now equals
the whole render exactly.
Release timings (glass_render_timings, 2400 × 1600 px, min of 5):
canvas drag under the glass 49.5 → 26.5 ms, glass slider drag
104.8 → 61.6 ms, same view again 94.8 → 54.0 ms; zoom 105.0 → 93.7 ms
(no reuse: each zoom is new).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
L6 depends on egui/eframe/rfd/winit. **The UI crate is swappable.**
- **The UI is thin**: panels read engine state and act through `app.run(id, params)` /
`app.set_props(ids, props)`. Colours come from `theme::Tokens`.
- **Time goes through `chrono`** (workspace dependency). Never hand-roll UTC/calendar conversion (days-to-date algorithms, leap-year math); a crate owns only its output format string. The one exception is range extension: past the years chrono can represent, shift the date by whole 400-year Gregorian cycles (146,097 days, after which the calendar repeats), let chrono compute the month and day, and add the cycles back to the year (splitting seconds into hours, minutes and seconds is plain arithmetic).
@@ -130,3 +131,8 @@ commit. Grade by behaviour against the incumbent, not by whether a menu item exi
GitHub profiles). Never hand-edit `contributors.json`.
- Maintainers refresh the stats with `python3 ../../craftrules/scripts/contributors.py .` (it also
re-verifies who wrote each `people.toml` entry; `--check` only verifies).
## CI workflows
- Workflows in `.github/workflows/` call the shared workflows in [`storytold/craft-common-actions`](https://github.com/storytold/craft-common-actions) at `@main`, with their inputs and secrets. No other jobs: if no shared workflow does what's needed, it doesn't go here.
- Don't add or change CI in a feature or fix PR. If the existing workflows can't build or test a change, link an existing craft-common-actions issue or PR, or open a new issue, or a PR only for a small fix that is safe for every app and does not increase CI time.
- Some workflow files here still run their own jobs. Don't add to them. Moving a job into craft-common-actions is its own task, never part of another PR, and the job stays here until a shared workflow runs it.
@@ -25,8 +25,24 @@ Release builds embed fonts from the optional craft-fonts build input
| Family | Licence | Upstream |
|---|---|---|
| Inter | SIL Open Font License 1.1 | https://github.com/rsms/inter |
| Roboto, Montserrat, Lato, Open Sans, Poppins, Nunito Sans, Merriweather, Playfair Display, Source Sans 3, Source Serif 4, Carlito, Caladea | SIL Open Font License 1.1 | Google Fonts / upstream repositories (see craft-fonts) |
| Archivo | SIL Open Font License 1.1 | [Google Fonts / Archivo](https://github.com/google/fonts/tree/6c70c829f09ea345d3590406693220ea35c6553f/ofl/archivo) |
| Roboto, Montserrat, Lato, Open Sans, Poppins, Nunito Sans, Merriweather, Playfair Display, Carlito, Caladea | SIL Open Font License 1.1 | Google Fonts / upstream repositories (see craft-fonts) |
| Liberation Sans, Serif, Mono | SIL Open Font License 1.1 | https://github.com/liberationfonts |
| BIZ UDPGothic, BIZ UDMincho, Shippori Mincho, Noto Sans Arabic | SIL Open Font License 1.1 | Google Fonts (see craft-fonts) |
| BIZ UDPGothic, BIZ UDMincho, Shippori Mincho, Noto Sans Arabic, Noto Sans Georgian | SIL Open Font License 1.1 | Google Fonts (see craft-fonts) |
The central catalog also contains Adobe-authored Source Sans 3 and Source Serif 4.
FigCraft filters both families from every build in accordance with its asset policy.
egui's default fonts (compiled into the `egui`/`epaint` crates) keep their own licences (OFL / Apache-2.0 / Ubuntu Font Licence); see those crates.
## egui dependency font icons
| Asset | Author | Source | Licence | Use |
|---|---|---|---|---|
| `epaint_default_fonts::EGUI_ICONS`, `epaint_default_fonts::EMOJI_ICON` (dependency font inputs) | John Slegers | [egui 6d518fc7](https://github.com/storytold/egui/tree/6d518fc799681571963842f95b55a6b68431b15a/crates/epaint_default_fonts/fonts), [emoji-icon-font](https://github.com/jslegers/emoji-icon-font) | MIT (Copyright (c) 2014 John Slegers; full upstream text in `NOTICE`) | egui-icons.ttf is the five-glyph subset, included by default_fonts independently of monochrome_emoji_fonts; emoji-icon-font.ttf is the full monochrome fallback. Dependency assets unmodified here; no font file copied into this repository. |
| `crates/ui-egui/locales/ar.toml` | @hussain-hz with OpenAI Codex (AI-assisted) | Original Arabic wording from FigCraft's English catalog and engine command labels; no proprietary localisation consulted | MIT OR Apache-2.0 | Draft; native-speaker review and maintainer approval pending |
- No `unwrap()`, `expect()`, `panic!`, `unreachable!`, `todo!`, `unimplemented!` in shipped code
(workspace clippy lints deny them; tests may use them). Return `Result`/`Option`; use `.get()` for
data-derived indexes; slice strings only at char boundaries; check floats are finite; cap sizes.
- New workspace crates must include `[lints] workspace = true` in `Cargo.toml` to inherit `[workspace.lints]`. Production crate roots must declare the required crate-level lint attributes before modules or other crate items; crate-level `//!` documentation may precede them.
-`figcraft_engine::guard` turns a panicking command into `EngineError::Internal` and the document
is rolled back; the desktop app also guards each UI frame and writes recovery copies. These are
safety nets, not substitutes for `Result`.
@@ -81,6 +82,7 @@ code.
L6 depends on egui/eframe/rfd/winit. **The UI crate is swappable.**
- **The UI is thin**: panels read engine state and act through `app.run(id, params)` /
`app.set_props(ids, props)`. Colours come from `theme::Tokens`.
- **Time goes through `chrono`** (workspace dependency). Never hand-roll UTC/calendar conversion (days-to-date algorithms, leap-year math); a crate owns only its output format string. The one exception is range extension: past the years chrono can represent, shift the date by whole 400-year Gregorian cycles (146,097 days, after which the calendar repeats), let chrono compute the month and day, and add the cycles back to the year (splitting seconds into hours, minutes and seconds is plain arithmetic).
@@ -130,3 +132,8 @@ commit. Grade by behaviour against the incumbent, not by whether a menu item exi
GitHub profiles). Never hand-edit `contributors.json`.
- Maintainers refresh the stats with `python3 ../../craftrules/scripts/contributors.py .` (it also
re-verifies who wrote each `people.toml` entry; `--check` only verifies).
## CI workflows
- Workflows in `.github/workflows/` call the shared workflows in [`storytold/craft-common-actions`](https://github.com/storytold/craft-common-actions) at `@main`, with their inputs and secrets. No other jobs: if no shared workflow does what's needed, it doesn't go here.
- Don't add or change CI in a feature or fix PR. If the existing workflows can't build or test a change, link an existing craft-common-actions issue or PR, or open a new issue, or a PR only for a small fix that is safe for every app and does not increase CI time.
- Some workflow files here still run their own jobs. Don't add to them. Moving a job into craft-common-actions is its own task, never part of another PR, and the job stays here until a shared workflow runs it.
@@ -51,3 +51,20 @@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
egui icon font subset (dependency asset)
----------------------------------------
epaint_default_fonts 0.37 embeds egui-icons.ttf, a five-glyph subset of
emoji-icon-font by John Slegers. The full monochrome emoji-icon-font fallback
uses the same MIT license. Upstream license text follows unchanged.
MIT License
Copyright (c) 2014 John Slegers
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
<img src="assets/app-icon/hicolor/256x256/apps/ai.storyteller.figcraft.png" width="128" height="128" alt="FigCraft icon: an engraved chameleon on a mint-green tile">
</p>
<h1 align="center">FigCraft</h1>
<p align="center">
<b>Interface design; an open-source, clean-room reimplementation of Figma, rebuilt in pure Rust.</b><br>
<b>Interface design and prototyping: opensource, collaborative and built in pure Rust.</b><br>
Frames, auto layout, components and variants, vector networks, styles, variables and prototypes,<br>
in a native app that opens instantly and works offline. Open source, and yours.
</p>
@@ -20,6 +24,9 @@
<img alt="macOS · Windows · Linux · FreeBSD · Web" src="https://img.shields.io/badge/macOS%20%C2%B7%20Windows%20%C2%B7%20Linux%20%C2%B7%20FreeBSD%20%C2%B7%20Web-native-0EA5C6?style=flat-square">
<img alt="License: MIT OR Apache-2.0" src="https://img.shields.io/badge/license-MIT%20%2F%20Apache--2.0-3a3a3a?style=flat-square">
<img alt="Status: early alpha" src="https://img.shields.io/badge/status-early%20alpha-d69e2e?style=flat-square">
<br>
<a href="https://sonarcloud.io/project/overview?id=storytold_figcraft"><img alt="Scanned by SonarCloud" src="https://img.shields.io/badge/Scanned_by-SonarCloud-4b9fd5?style=flat-square&logo=sonarqubecloud&logoColor=white"></a>
<a href="https://coderabbit.ai"><img alt="Reviewed by CodeRabbit" src="https://img.shields.io/coderabbit/prs/github/storytold/figcraft?style=flat-square&label=Reviewed%20by%20CodeRabbit&logo=coderabbit"></a>
</p>
<p align="center">
@@ -62,7 +69,7 @@
<tr>
<td width="25%" valign="top">
<h3>🎛️ Familiar by design</h3>
The toolbar, layers panel, Design panel, menus and shortcuts sit where your hands expect them. If you know Figma, you already know FigCraft.
The toolbar, layers panel, Design panel, menus and shortcuts sit right where your hands expect them.
</td>
<td width="25%" valign="top">
<h3>⚡ Native and fast</h3>
@@ -90,13 +97,13 @@
<p align="center"><sub><b>Components and variants</b>: component sets, variant properties and the Assets panel · <code>examples/ridgeline-design-system.figcraft</code></sub></p>
</td>
<td width="50%" valign="top">
<img src="docs/images/shot-3-prototype.png" alt="Four onboarding phone screens in the Prototype tab, joined by blue connection curves from the Get started, Continue, Allow location and Not now buttons to the next screens; the Get started button is selected and the Prototype panel shows its On click, Navigate to 2 Interests interaction" width="100%">
<img src="docs/images/shot-3-prototype.png" alt="Four onboarding phone screens in the Prototype tab, joined by green connection curves from the Get started, Continue, Allow location and Not now buttons to the next screens; the Get started button is selected and the Prototype panel shows its On click, Navigate to 2 Interests interaction" width="100%">
<p align="center"><sub><b>Prototyping</b>: connections, transitions, back actions and a flow · <code>examples/ridgeline-onboarding.figcraft</code></sub></p>
</td>
</tr>
<tr>
<td width="50%" valign="top">
<img src="docs/images/shot-4-devmode.png" alt="Dev Mode on the Ridgeline home screen: the frame carries a Ready for dev badge, a Development annotation callout points at the featured trail card, a trail row instance is selected and the Inspect panel shows generated React code with a TrailRow props interface and the instance's title and details values" width="100%">
<img src="docs/images/shot-4-devmode.png" alt="Dev Mode on the Ridgeline home screen: the frame carries a Ready for dev badge, a Development annotation callout points at the featured trail card, a trail row instance is selected and the Inspect panel shows a Playground preview of it and generated React code with a TrailRow props interface and the instance's title and details values" width="100%">
<p align="center"><sub><b>Dev Mode</b>: ready-for-dev status, annotations and React code where component properties become props · <code>examples/ridgeline-app.figcraft</code></sub></p>
</td>
<td width="50%" valign="top">
@@ -114,6 +121,16 @@
<p align="center"><sub><b>Live sessions</b>: three apps on one <code>figcraft-server</code> file, with cursors, selections, cursor chat and spotlight</sub></p>
</td>
</tr>
<tr>
<td width="50%" valign="top">
<img src="docs/images/shot-8-board.png" alt="A board file named Sprint retro on a dot grid: a Sprint 14 retro made from the retrospective template, with Went well, To improve and Next steps sections of sticky notes, vote counts of 3, 2, 1 and 1 on four stickies after a vote has ended, and below it a pull request flow chart from the flow chart template whose shapes are joined by straight and elbow connectors with Yes and No labels; a timer pill in the top right shows 4:30 left with pause, add a minute, stop and Results buttons" width="100%">
<p align="center"><sub><b>Boards</b>: a whiteboard file kind with sticky notes, templates, connectors, a shared timer and voting, in the same live sessions</sub></p>
</td>
<td width="50%" valign="top">
<img src="docs/images/shot-9-devmode-measure.png" alt="Dev Mode on the Ridgeline home screen with the featured trail card selected: red redlines kept on the canvas show 118.3 from the card to the search field, 62.1 from the card to the first trail row and 256.3 along the card's left side, and the Inspect panel's Measurements section lists the three with their values under an Add measurement menu" width="100%">
<p align="center"><sub><b>Dev Mode measurements</b>: redlines placed between layers or along a side, saved in the file and following the layers · <code>examples/ridgeline-app.figcraft</code></sub></p>
</td>
</tr>
</table>
## Made in FigCraft
@@ -191,7 +208,7 @@ exposes to agents, and exported at 2x by FigCraft's own renderer. The source fil
</td>
<td width="50%" valign="top">
<h3>Prototype and hand off</h3>
Connections, triggers, actions and Smart Animate-style transitions, variables with conditionals (if / else if / else built in the interaction editor, expressions like <code>count + 1</code>), flows, a presentation view with device frames, hotspot hints and an Inspect panel, and a live preview pane. Alt to measure distances and padding on the canvas, Inspect mode with measurements, annotations in colour-coded categories, a ready-for-dev list with a focus view and Compare changes (before / after with property-level diffs since a frame was marked ready), and code (CSS, HTML, React, Leptos, SwiftUI, Compose) where components take their properties as parameters, comment threads pinned to frames (replies, @mentions, reactions, read state) and named version history. Export any layer at any scale to PNG, JPG, WebP, SVG or PDF.
Connections, triggers, actions and Smart Animate-style transitions, variables with conditionals (if / else if / else built in the interaction editor, expressions like <code>count + 1</code>), flows, a presentation view with device frames, hotspot hints and an Inspect panel, and a live preview pane. Alt to measure distances and padding on the canvas, Inspect mode with measurements, annotations in colour-coded categories, a ready-for-dev list with a focus view and Compare changes (before / after with property-level diffs since a frame was marked ready), and code (CSS, HTML, React, Vue, Leptos, SwiftUI, Compose, ArkUI, Flutter) where components take their properties as parameters, comment threads pinned to frames (replies, @mentions, reactions, read state) and named version history. Export any layer at any scale to PNG, JPG, WebP, SVG or PDF.
</td>
</tr>
</table>
@@ -227,8 +244,8 @@ The desktop app also offers a loopback-only control channel (`figcraft --control
- **Its own renderer:** tiny-skia on every core for the canvas and exports, with layer blending, effects and masks; booleans from linesweeper, geometry from kurbo, text from skrifa and HarfRust.
- **Runs in the browser:** the whole engine and UI compile to WebAssembly.
- **Accessible:** screen readers get names, roles, values and states for the tools, layer rows, fields, dropdowns, tabs and menus (AccessKit on desktop), F6 moves the keyboard between the toolbar, layers, canvas and properties (Tab within them, Escape back to the canvas), the system's reduce-motion setting is honoured, and panel text meets WCAG AA contrast in both themes.
- **Languages:** English, Russian and Turkish (reviewed by native speakers), and Spanish, French, German, Brazilian Portuguese, Japanese, Korean, Simplified Chinese, Czech and Indonesian drafts, picked in Preferences or taken from the system or browser and switched live; adding a language is one catalog file ([`docs/translating.md`](docs/translating.md)).
- **Plugins:** sandboxed WebAssembly plugins (a metered, pure-Rust interpreter with a memory cap and no file or network access, in the browser too) that act through the same commands, each run one undo step, within the permissions their manifest asks for. A Rust SDK and three example plugins are in [`plugins/`](plugins); see [`docs/plugins.md`](docs/plugins.md).
- **Languages:** English, Russian and Turkish (reviewed by native speakers), and Spanish, French, German, Brazilian Portuguese, Japanese, Korean, Simplified Chinese, Czech, Indonesian, Georgian and Vietnamese drafts, picked in Preferences or taken from the system or browser and switched live; adding a language is one catalog file ([`docs/translating.md`](docs/translating.md)).
- **Plugins:** sandboxed WebAssembly plugins (a metered, pure-Rust interpreter with a memory cap and no file or network access, in the browser too) that act through the same commands, each run one undo step, within the permissions their manifest asks for. A Rust SDK and four example plugins are in [`plugins/`](plugins); see [`docs/plugins.md`](docs/plugins.md).
- **Live sessions (preview):** share the open file from the app (Share ▸ Start sharing: online through a server, or on your local network), give the others the short join code it shows (or the invite link), and edit it together: everyone's changes sync live, with cursors, selections, avatars and follow, and undo takes back only your own changes. Others join from the desktop app (invite links open it) or from the browser build. See [`docs/collab-protocol.md`](docs/collab-protocol.md).
- **A server of your own (preview):** `figcraft-server` keeps a team's files, publishes them from the app (File ▸ Publish to server…, or Share ▸ Online with your server in Preferences) and hands out share links and join codes that let people edit or only view; files survive restarts and apps reconnect where they were. No accounts yet: links are the access. Releases ship it for Linux, and the desktop app joins over `wss://`; [hosting it on a rented Linux server](docs/server.md#deploy-on-a-linux-server) takes a systemd unit and Caddy or nginx (or Docker). See [`docs/server.md`](docs/server.md).
- **Never crashes:** no `unwrap`, `expect` or `panic` in shipped code, enforced by clippy; commands run behind a guard so a bug fails one action, not your session.
@@ -262,7 +279,7 @@ New contributors and AI agents: start with [`AGENTS.md`](AGENTS.md), then [`docs
Installers for macOS (universal, signed and notarized), Windows (x64, x86 and arm64), Linux (AppImage, `.deb`, `.rpm`, Flatpak and tarball), FreeBSD and the web are attached to each [GitHub release](https://github.com/storytold/figcraft/releases). Maintainers: [`docs/releasing.md`](docs/releasing.md) explains how releases are built, signed and published.
> [!IMPORTANT]
> **Status:** FigCraft is in **alpha**: about 49% of the way to replacing Figma for real work, by our own estimate. Nearly all of Figma's main menu is wired ([`docs/parity-checklist.md`](docs/parity-checklist.md)) and the core editor works end to end, but FigCraft opens `.fig` files only best effort, has only a first preview of multiplayer (live sessions hosted from the app or a self-hosted server, with per-person undo and share links to edit or to view; no accounts, teams or cloud yet), runs its own WebAssembly plugins rather than Figma's, and hasn't been checked against real Figma documents. The honest, area-by-area picture is in the [roadmap](ROADMAP.md) and [`docs/target-app-parity.md`](docs/target-app-parity.md). Expect rough edges and please file issues, or tell us what broke on [Discord](https://discord.gg/artcraft).
> **Status:** FigCraft is in **alpha**: about 49% of the way to replacing proprietary tools for real work, by our own estimate. Nearly all tracked interface design actions in the menu checklist are wired ([`docs/parity-checklist.md`](docs/parity-checklist.md)) and the core editor works end to end, but FigCraft opens `.fig` files only best effort, has only a first preview of multiplayer (live sessions hosted from the app or a self-hosted server, with per-person undo and share links to edit or to view; no accounts, teams or cloud yet), runs its own WebAssembly plugins, and hasn't been checked against real design documents. The honest, area-by-area picture is in the [roadmap](ROADMAP.md) and [`docs/target-app-parity.md`](docs/target-app-parity.md). Expect rough edges and please file issues, or tell us what broke on [Discord](https://discord.gg/artcraft).
## Privacy and online sharing
@@ -348,8 +365,6 @@ ArtCraft Team and are not covered by this license. They may be used only unmodif
part of this repository and FigCraft, under [`docs/brand/LICENSE-brand.txt`](docs/brand/LICENSE-brand.txt).
Forks and modified versions must remove them.
<sub>Figma is a trademark of Figma, Inc. FigCraft is an independent, open-source project and is not affiliated with, sponsored by or endorsed by Figma, Inc.; the name is used only to describe the workflows FigCraft is compatible with.</sub>
**Stage: alpha** · next: beta, ~26 points and ~162–292 h away (one Opus 5.5 agent; ~33–67 h wall-clock with 5 agents)
> **Last reviewed:** 2026-10-10 · **Last updated:** 2026-10-17 · **Change:** minor (community batch #753: Selection colors, smart-selection spacing, a Present tab; language table re-measured from the catalogs; community fixes from #524 up (55 PRs: `.fig` import fidelity, paste from other apps, the Actions palette, editor fixes, server limits) and a boolean-op crash guard; number variables in the text content picker (#629); frames and sections convert into each other (#719); bounded canvas pan overscan; community fixes from #478 up and a control-channel fix (#589); bounded canvas pan overscan; Typography spacing icons support scrubbing (#373); font scans skip named pipes; system theme on first run; recent files as cards with thumbnails on the home screen; community fixes: lost images reported on loading (#207), SVG import budgets (#190), Leptos code (#223), full screen (#227), server snapshots readable again (#188), owner-only preferences (#200), ruzstd 0.9.1 (#182), layer cycle repair (#205), batch rollback, Direct3D 12 on Windows, plugin metering (#192); vector edit mode regions: hover hatch, selection and drag; SVG/PDF exports refuse over-depth documents (#208); bounded MCP/control framing, connections and queues (#201); plugin log limits (#191); live-session revert guard (#198); attached MCP no longer replays after a lost reply (#194); repeat command recursion fix; the Layout panel's constrain-proportions link scales the other side, #220; core editor leftovers: component playground, auto gap handles, variables inside styles, inner shadows and background blur on groups, command-made text readable on dark pages, a one-step frame/group convert, delays on mouse triggers and manual overlay offsets, wrapped-line alignment; `.fig` export: instance sizes, variables with modes, shared styles, variable-width strokes; starter templates, a Templates tab and Getting started tips on the home screen; your online files; everyday editing partials: number fields, outline options, Hide when publishing, cursors, panel widths, dark pages; sharing with a join code; text box sizes measured against documents re-saved by other software; live sessions per document; .fig export checked by opening written files in other software; a public figcraft-server: open hosting, owner tokens, per-address limits, idle expiry, moderation, install.sh; experimental .fig export; keyboard shortcut parity suite and bindings; Simplified Chinese draft catalog; Latin American Spanish overlay; prototyping tail: gamepad triggers, widgets in prototypes, mode lookups; plugin directory on figcraft-server; Japanese and Korean draft catalogs and CJK UI fonts; plugin widgets and the plugin panel tail; French, German and Brazilian Portuguese draft catalogs; cursor chat and spotlight in live sessions; layers and assets panels behaviour suite and fixes; vector multi-edit and edit tail; dev resources links; extended collections tail; branch version history and finer list merges; headline percentages recomputed from the area table: feature breadth ~84%, core ~94%, mainstream ~50%; text lists, hanging punctuation and decoration styles; comments and version history behaviour suite and fixes; canvas navigation, guides and grids behaviour suite and fixes; export and inspect behaviour suite and fixes; .fig import: editable vector networks and more layer kinds; shapes and vector behaviour suite and fixes; paint and effects behaviour suite and fixes; frames, groups, sections and masks behaviour suite and fixes; styles and variables behaviour suite and fixes; prototyping behaviour suite and fixes; components behaviour suite and fixes; embedded image colour profiles; layers panel keyboard: F2 rename, Shift+F10 context menu, keyboard box selection, 10k-layer QA; per-side stroke weight variables; text behaviour suite: CJK line breaking, metrics, vertical trim; selection and transform behaviour suite and fixes; `.fig` import, best effort; Windows QA pass with eight fixes; auto layout behaviour suite and fixes; Display P3: gamut warning and soft-proof (true P3 blocked upstream); several windows; localization layer with a draft Spanish catalog; hosting figcraft-server: wss in the desktop app, server releases, deployment guide; plugin panels and settings; canvas fixes: boolean hit testing on the result, group drags, frame outlines and slices; code connect; stretch and scatter brushes, pen pressure; text on a path and pattern fills; team libraries on the server and swap library; branches: create, update from main, review and merge; animated GIF fills and TIFF images; Open recent, outlines for the selection and mask outlines: menu breadth 100%; extended collections, variables import/export, gradient stop and variant bindings; feature breadth ~80%; ready ~49%) · **Target:** Figma Design (desktop app 126.10.7)
> **Last reviewed:** 2026-10-10 · **Last updated:** 2026-10-30 · **Change:** minor (pasted layers whose image the clipboard lacks are named and show a placeholder that reads as missing (#1646); paint-style colours on layers pasted from another design app keep their style instead of turning black (#1638); progress docs cleaned of what merges left twice: rows merged into one, repeated status items removed, and `cargo xtask docs` (in `ci`) fails on a repeated table row or status item; a name typed for a variable or collection finds the one with that exact id before one named like it, and stored mode ids, font family variables and the panels' labels read exact ids (#1613); plugin widgets have hover styles: a layer's `hover` fill, stroke, opacity or text colour shows while the pointer is over it (gap 6); `.fig` elbowed and curved connectors without a saved path are routed like board connectors instead of drawn straight (gap 1); every file a command reads whole by path is a regular file within a cap that fits it: documents 16 GiB, placed images 256 MiB, CLI scripts 1 GiB; `figcraft-cli connect publish` publishes a code repository's code connect mappings to a design file (gap 16); while a drag only moves a large selection, its outlines, box, slots, spacing and Design panel values are worked out once and shifted, not rebuilt on every move (#1571); Shift+Cmd/Ctrl+Enter on a selected board shape adds the next shape already joined to it, and goes on typing (#1597); an imported or pasted file finds an installed family under the other names its fonts give it, and with other case and spacing, instead of reporting it missing (#1292); a development example generates figcraft-cli's shell completions and man pages from its clap definitions (part of #1396); `.fig` ZIP image reads share a total byte budget and count skipped images in the import warnings (#1076); German terminology corrections (M10): gradient stops, OpenType stylistic sets and the Home-screen preference; a component set's property rows fit the 240 px right panel: variant values wrap two to a line there and every + and − stays in view; the server key is checked by comparing SHA-256 hashes of the presented and the kept key, as the deployment guide says (#1130); opening a file contacts no team library server it names until you check it (#1097); a hostile server can no longer grow a client's memory with made-up people, and a session remembers a bounded number of deleted layer ids (#1121); `doc.list` gives AI agents and plugins the open files' names, not their paths (#1117); with a large selection, frames where nothing changed reuse the selection's outlines, box and slots and the Design panel's instance and position values instead of working them out again (#1571); team libraries and plugins larger than the 16 MiB download burst can be downloaded again, more slowly (#1089); a batch keeps one undo snapshot, not one per step (#1099); files nested 256 levels deep import on the browser's 1 MB stack (#1101); SVG and PDF exports count pattern sources and the renders they embed from the layer's depth, so one cap bounds a whole export (#1082); an explicit variables file path (`variable.import`, the `variables` of `file.open` and `file.import`) is refused past 256 MiB or when it isn't a regular file, as the sibling `variables.json` already was (part of #1114); joining and leaving a file nobody is editing reads and writes nothing on the server: it stays loaded a minute after everyone leaves (#1074); opening a `.figcraft` file unpacks at most 6 GiB in all (2 GiB in the browser), every area together (#1079); hovering the canvas hit-tests once per pointer move, not on every frame, and moving an auto layout frame no longer lays out its contents again on every move (#1571); `.fig` instance expansion no longer walks variable nodes for every instance or copies every outer override row into each nested instance (#1077); `comment.people` lists names that were only @mentioned, and `comment.next` and `comment.prompt` carry the mentions too (#1415); grid auto-placement looks for free cells a 64-bit word at a time and skips what blocks it, with the same placements (#1083); with the pen out, Enter or a double-click ends the path and edits the vector with the Move tool instead of drawing on (#1430); a recovery copy made while a version is previewed holds the file, not the version shown; a FigJam `.jam` file opens as a board (#1342); the Actions palette offers only commands that run from it, and no second row reading like a menu item: Place image opens the file dialog (#1574); share links can comment: a third access level beside edit and view, whose holder sees the file live and adds, replies to, reacts to and resolves comments but can't change the design; Preferences ▸ Scroll wheel zoom: a mouse wheel zooms the canvas about the pointer and Cmd/Ctrl+wheel scrolls (#1390); `comment.list {mentions}` lists only the threads that @mention a name (#1415); in the browser an internal error thLine truncated
FigCraft is a clean-room, open-source, pure-Rust reimplementation of the Figma design workflow,
native on macOS, Windows, Linux and FreeBSD and in the browser, fully drivable by agents through
@@ -19,13 +19,13 @@ start with [`docs/gaps.md`](docs/gaps.md).
| Essentials user | **~59%** | estimated: core features 77.8% × 0.92 launch/stability × 0.94 discoverability × 0.88 opening files people send ([method](docs/target-app-parity.md#essentials-user)) |
| Remaining to beta | **~155–280 h** one agent | estimated |
| Remaining to full parity | **~262–519 h** one agent (~83–145 h wall-clock with 5 agents) | estimated |
| Remaining to full parity | **~260–515 h** one agent (~83–145 h wall-clock with 5 agents) | estimated |
### Readiness by audience
| Audience | Ready % | Opus 5.5 agent wall-clock hours to ~95% | Work that dominates |
|---|---|---|---|
| Full target (ready for real work) | ~53% | ~268–522 h | multiplayer (42–84), localization (29–51), plugins (8–26), the interaction pass (40–60), GPU canvas (30–60), AI (30–60), the `.fig` import's tail (5–10) |
| Full target (ready for real work) | ~53% | ~266–518 h | multiplayer (42–84), localization (29–51), plugins (6–22), the interaction pass (40–60), GPU canvas (30–60), AI (30–60), the `.fig` import's tail (5–10) |
| Mainstream practitioner | ~50% | ~120–229 h | core editor depth incl. the `.fig` import's tail (49–98), the interaction pass (40–60), real-file corpus and platform QA (25–50), checking the rest of the experimental `.fig` export elsewhere: effects, images, prototypes, real files (5–15) |
| Essentials user | ~59% | ~55–100 h | the `.fig` import's tail so the files people send open well (5–10), core tool polish (25–45), interaction pass on the core tools (15–25), stability QA (10–20) |
@@ -50,7 +50,7 @@ against real Figma documents. Ready-for-real-work is ~53%, inside alpha's ~40–
~75% **and** reliable `.fig` import: the `.fig` import's tail (5–10 h), a real-file corpus with measured
fidelity (15–30 h), the interaction pass (25–40 h of the 40–60), core depth to ~80% (40–75 h),
basic multiplayer on the landed core (30–55 h), stability and platform QA (20–40 h), a partial GPU path (15–30 h);
native speakers' reviews of the six language drafts need people.
native speakers' reviews of the nine language drafts need people.
## By dimension
@@ -58,14 +58,14 @@ native speakers' reviews of the six language drafts need people.
| File formats | 60% (`.fig` read best effort with editable vectors; experimental write, round trips through our own importer and written files opened in other software) | 5–10 | [file-format-parity](docs/file-format-parity.md) |
| Collaboration | 13% (core in; live sessions with document sync, cursors and text carets, cursor chat and spotlight, the browser joins, invite links open the app; a self-hosted server with edit and view links and team libraries; preview) | 46–92 | [collaboration-parity](docs/collaboration-parity.md) |
| File formats | 60% (`.fig` read best effort with editable vectors and a total ZIP image byte budget; experimental write, round trips through our own importer and written files opened in other software) | 5–10 | [file-format-parity](docs/file-format-parity.md) |
| Collaboration | 13% (core in; live sessions with document sync, cursors and text carets, cursor chat and spotlight, the browser joins and publishes to a server, invite links open the app; a self-hosted server with edit and view links and team libraries; preview) | 46–92 | [collaboration-parity](docs/collaboration-parity.md) |
| Ecosystem / plugins | 38% (own WASM plugins with panels, settings and widgets; a self-hosted plugin directory) | 8–26 | [gaps](docs/gaps.md) |
| Ecosystem / plugins | 40% (own WASM plugins with panels, settings and widgets that take clicks, text and a property menu; a self-hosted plugin directory) | 6–22 | [gaps](docs/gaps.md) |
| AI features | 0% | 30–60 | [gaps](docs/gaps.md) |
## Features
@@ -86,13 +86,14 @@ native speakers' reviews of the six language drafts need people.
Percentages measured 2026-10-17 over 2,272 entries (strings added since a draft show English). Also shipped: Russian (`ru`, 99%, reviewed by VanTan), Swedish (`sv`, 99.6%, command labels since 2026-10-17; native review welcome), Turkish (`tr`, 99%) and Czech (`cs`, 99%, draft). Figma ships English, Japanese, Korean, French, German, Spanish and Portuguese
(Brazil); FigCraft has the localization layer, English and Spanish (Spain and Latin America),
French, German, Brazilian Portuguese, Japanese, Korean, Simplified Chinese and Indonesian drafts (Preferences ▸
Language; Traditional Chinese has no catalog; CJK UI
text uses an installed CJK font, loaded only when needed; adding a language:
[`docs/translating.md`](docs/translating.md)). Document text in any script (bidi, Arabic shaping,
@@ -126,14 +127,53 @@ Ranked; detail and milestones in [`docs/roadmap.md`](docs/roadmap.md).
1.`.fig` import's tail: fixes from real files, connector labels, table cells, filled areas with their own fills (the best-effort import, editable vector networks, the other layer kinds and variable prototype actions landed 2026-10-11): 5–10 h; then checking the rest of the experimental `.fig` export (landed 2026-10-12; written files open and draw in other software) on effects, images, prototypes and real files: 5–15 h.
2. Real-file corpus and measured fidelity: 15–30 h.
3. Tool-by-tool interaction pass against Figma: 40–60 h.
4. Languages: native reviews of the es, fr, de, pt-BR, ja and ko drafts (people), Latin American Spanish (1–2 h), then Simplified Chinese (3–5 h).
4. Languages: native reviews of the es (with es-419), fr, de, pt-BR, ja, ko, zh-Hans, id and cs drafts (people), the entries added since each catalog was written, then Hindi and Vietnamese; Arabic native review and panel mirroring ([localization-parity](docs/localization-parity.md)).
5. Multiplayer (M9): accounts and teams on top of the live sessions and the standalone server (transport, hosting, invite links that open the app, document sync, presence with remote text carets, cursor chat and spotlight, per-user undo, reconnect with offline edits, collision-free ids for styles, variables and comments, the browser client, `figcraft-server` with persistence and share links to edit or view, team libraries on the server landed): 42–84 h.
## Progress log
- **2026-10-29** · M10: correct nine German labels in `de.toml`: gradient stops use Verlaufsstopp, OpenType ss01–ss05 use Stilsatz, and the Home-screen preference names the Startbildschirm consistently in the UI and Actions palette. This is a partial terminology pass; German remains a draft awaiting a full native-speaker review and maintainer approval. Numbers unchanged.
- **2026-10-28** · M3, #922: an instance scaled with the Scale tool (or `transform.scale`) keeps its layers' scaled size and points when its main is rebuilt. The scaled layers' new size and place were read as what constraints give (an ancestor resized in the same edit), so no override was kept, and adding a layer to the main put a 14 px icon's 24 px triangle back. The Scale tool now records them as `size` and `position` overrides on layers inside instances; strokes already kept theirs. The report's own sequence (clearing effects on the component holding the icon) no longer reproduces on main. Regressions for a scaled instance, one scaled with the tool (with undo) and a scaled icon nested in a component and its instances. Numbers unchanged.
- **2026-10-28** · M6, #1059:`.fig` import copies each image carried inside the message once per blob, shared by every hash that names it, instead of once per hash: a file pointing many hashes at one large blob no longer multiplies it in memory (100 hashes on a 250 MB blob were 25 GB). Regression test. Numbers unchanged.
- **2026-10-28** · M1, #1476: page ruler guides survive save and reopen. A page kept its guides in the page kind while the layer's own `guides` field (frame guides, #70) took that key when a file was read, so a reopened page's guides vanished, and saving it wrote `guides` twice, which then refused to open. Every layer now keeps its guides in the one field, and a page with two `guides` lists opens with both. Regressions in the doc and format crates. Numbers unchanged.
- **2026-10-27** · M6, #1341: FigJam sticky notes and shapes with text import their generated layers (#1397): the background (its saved vector network, or the outer fill or stroke outline when the generated geometry is damaged, before a rectangle), text with its typography and paints, saved bounds and sticky author lines, as editable frame children; missing or invalid generated text falls back to the outer text within the layer budget. Synthetic regressions and a local check of the report sample; native board metadata and direct `.jam` opening remain outside this fix. Numbers unchanged.
- **2026-10-27** · Batch continue-on-error (M0.9, #1386): `command.batch` and MCP `batch` accept `onError: "continue"`, roll back failed steps individually and return one outcome per step. References to failed steps fail with both step numbers; successful edits stay one undo step, and the default abort mode is unchanged. Engine and MCP regression tests. Numbers unchanged.
- **2026-10-26** · Non-ASCII layer names transliterate in code export (M6, #1235): `class_name` runs the name through `deunicode` before its ASCII filter, so 로그인 화면 → `rogeuin-hwamyeon`, кнопка → `knopka`, über → `uber` in the HTML/CSS/React/Leptos class names, component roots and `data-*` variant attributes (SwiftUI/Compose type names unchanged); a name left with nothing (symbols only) takes the node's kind (`.text`, `.frame`) instead of `layer-`, and the 40-character cut lands on a word boundary. New dependency: deunicode (pure Rust tables, zero deps, BSD-3-Clause, the `slug` crate's base). `html_codegen_tests`. Numbers unchanged.
- **2026-10-25** · M6, #1143: offline Penpot 2.x open/import/export through the editor, CLI and MCP using the existing document model. Unsupported layout settings retain saved positions and bounded source metadata with warnings; font binaries and toggle-overlay are reported, and provider fonts are never downloaded. Shared ZIP/JSON/token budgets guard untrusted input. Figma parity numbers unchanged.
- **2026-10-25** · egui 0.37 capture lifecycle: screenshots with hidden or unknown visibility queue, request a hidden paint, and expire after 60 seconds if no capture arrives (headless-tested contract). Logic-only edits, saves and inspection continue while hidden. Native visible GUI and forced hidden-window paint, browser and Wayland behavior remain unqualified; parity percentages are unchanged.
- **2026-10-24** · Open, Save and the folder dialogs start in the folder of the last file or folder picked in the run (#1290): the Linux portal does not remember it between dialogs as the macOS and Windows panels do. A folder deleted since falls back to the system's choice. No numbers change.
- **2026-10-24** · Boards (M12): board pages get a dot grid under their layers (drawn by the canvas render after the page colour, thinned out when zoomed out); a selected connector shows its ends as handles, and dragging one onto another object re-attaches it there, onto empty canvas leaves it free (`board.reattach`, one undo step, also over MCP); connectors can be curved, and elbows go around the objects they join when the short way would cross them. Checked in the app, through MCP and in the web build ([boards-parity](docs/boards-parity.md)).
- **2026-10-24** · Installed weights of an embedded font family (M7, #1246): release builds embed some families with only a few styles (Poppins Regular, Italic and Bold), and the first font scan dropped every installed face of such a family, so an installed Poppins SemiBold or Medium was never offered (a rescan did offer them). The first scan now keeps the embedded faces, which still win for the styles they cover, and offers the installed styles they lack beside them. Regression test in the font database tests. Numbers unchanged.
- **2026-10-24** · Dev Mode tail (M6, gap 16): Measurements kept on the canvas (gap 16): `devMode.addMeasurement {from?, to?, side?}` places one between two layers, along one side of a layer or (neither) to the layer around it, stored on the layer it starts from (saved in `.figcraft`, synced, undoable; old files open without them); Dev Mode draws them as labelled redlines that follow the layers, the Inspect tab lists them (those of the selected layer, or every one on the page), places and removes them; `devMode.removeMeasurement` and `devMode.measurements` remove and list them; a copy keeps those to layers copied with it. The layers panel is read-only in Dev Mode: rows select and expand, but don't rename, reorder, hide or lock. Dev Mode 98% / 78%.
- **2026-10-24** · Board facilitation (M12): a timer everyone on a board sees (presets or any minutes, pause, resume, a minute more, stop; synced as its end time, so nothing ticks over the network) and voting sessions (votes per person, on everything, on sticky notes or on the selection; a click votes, your badge takes it back; your votes show while it runs, everyone's with a ranked list once it ends; clear results), in a pill at the top right of the board. `board.timer`, `board.timerState`, `board.vote`, `board.voteResults`; ballots sync per voter. Checked in the app, with two app instances in one live session through a local `figcraft-server`, and through MCP; the web build compiles ([boards-parity](docs/boards-parity.md)).
- **2026-10-24** · Board templates and tables (M12): a templates picker in the board toolbar places a retrospective, a brainstorm, a weekly planner, a flow chart or a kanban board (`board.template`, all FigCraft's own) as ordinary board objects in one undo step; `board.table` makes a table of auto layout frames (up to 50 × 20) whose rows grow with their text, edited by row and column from the bar over the selection or `board.tableEdit`, typed into with a double-click. Boards ~35% → ~40% breadth, 30–60 h → 26–54 h; the weighted numbers are unchanged.
- **2026-10-24** · Plugin widgets take text input and a property menu (M11, gap 6, plugin API 4; older plugins run
unchanged): a double-click on a widget's editable field edits it like any text, and when editing ends the value goes
into the widget's state and its plugin redraws it, one undo step (`widget.setText` from scripts and MCP); a widget
can carry a few actions, toggles and colour choices, shown in a small bar over it while it's selected
(`widget.menu`). The Counter example has a label field and a menu (reset, steps of ten, a colour). Plugins & widgets
65%.
- **2026-10-24** · Boards (M12): File ▸ New board makes a board, a file kind of its own saved in `.figcraft` and carried to people joining a live session. Sticky notes (8 colours, text that shrinks as the note grows, the author's name, Shift+Cmd/Ctrl+Enter for the next one), shapes with text (7 outlines), connectors that re-route when what they join moves (computed in the engine's normalize pass, so local edits, undo and others' edits agree), marker, highlighter and eraser, stamps with the placer's name on hover; a board toolbar, a contextual bar over the selection and the Design panel hidden until asked for; prototyping and Dev Mode refuse on boards. Design files and boards have their own marks (a pen nib, a note) on tabs, the home screen and the Open files list. Checked in the app, with two app instances in one live session, and through MCP; the web build compiles ([boards-parity](docs/boards-parity.md)).
- **2026-10-22** · CJK line breaking completes JIS X 4051's kinsoku (M4): the line-start set now also has the halfwidth small kana ｧ–ｯ and small tsu, the halfwidth (semi-)voiced sound marks ﾞﾟ, the spacing (semi-)voiced sound marks ゛゜, the kana iteration marks 〱–〵 and the halfwidth closing parenthesis ｠; the line-end set has the halfwidth opening parenthesis ｟. Text layers, auto layout and the canvas otherwise wrap Japanese as before: ideographs break anywhere and the behaviour suite pins every category plus that a manual newline may still start a line with a prohibited character. Numbers unchanged.
- **2026-10-15** · Auto-save (M1, #305): Preferences ▸ Save files automatically (`pref.autoSave`, off by default, desktop only) saves each open file that has a path 3 s after its last change, never during a gesture, a live text preview or while a version is shown; the file is encoded off the UI thread, a failed save says why and is tried again a minute later. Numbers unchanged.
- **2026-10-15** · A text box by dragging (M1, #363): a drag with the Text tool makes a fixed-size text box the size of the drag; a click still makes auto-width text. Numbers unchanged.
- **2026-10-15** · Batch step references (M0.9, #400): a `command.batch` step can use the id an earlier step made (`{"$step": n}`, a reserved key), so a nested tree takes one call and one undo step. Numbers unchanged.
- **2026-10-15** · Removing someone from a live session (M9, #338): the host or the file's owner removes a person from the avatar's menu or with `collab.removePerson`; they keep their copy and can't rejoin while the session runs. Numbers unchanged.
- **2026-10-15** · Languages (M10, #257, #263, #469, #477, #288): Russian (one/few/many plurals) and Turkish catalogs reviewed by native speakers, Czech and Indonesian drafts, the Japanese draft completed; Preferences ▸ Language lists 12 languages and Latin American Spanish. Numbers unchanged.
- **2026-10-15** · Zoom previews (M8, #312): when renders from scratch are slow, a new zoom shows the whole view coarsely first, then refines it. Numbers unchanged.
- **2026-10-15** · Web clipboard (M1, #375): copy and paste in the browser go through the system clipboard. Numbers unchanged.
- **2026-10-15** · Alt-drag copies snap to their original and Duplicate (Cmd+D) repeats the last copy's offset (M1, #283). Numbers unchanged.
- **2026-10-22** · `fill` and `stroke` props accept `"none"` and `null` (M1, #1036): `node.create`, `node.set` and the other prop appliers map them to an empty fills/strokes list, as `fills: []` already did, so an outline-only shape is one call (agents write `"none"` for no paint; the empty-list form was undocumented). The parameter docs (`KEYS`, shown by `list_commands` and the CLI) and `docs/mcp.md` say so, and the error message names the accepted forms. `fill_none_tests` covers create, set, undo, text and the error. Numbers unchanged.
filenames open normally; the global `--no-system-fonts` flag also respects the terminator.
Binary regression coverage opens five such native files and checks their JSON info.
- **2026-10-17** · Control channel while the window is minimised (M8, #720): a minimised or covered window answers every request that acts on the app's state (edits, `ui.inspect`, `ui.pointer`, `ui.key`, `ui.type`, `ui.set`, `app.quit`…) without drawing, and what needs a drawn frame (`ui.screenshot`, `ui.input`, `window.new`) is an error at once; before, those waited up to 60 s and held up every later request on the connection (an attached agent's `inspect_ui` or window screenshot stalled until the window came back). Measured on Windows 11, minimised: requests ~100 ms each (eframe wakes a minimised window at most ten times a second; ~2 ms on screen), a window screenshot fails in ~100 ms instead of 60 s, idle CPU 0. Numbers unchanged.
- **2026-10-17** · Number variables on text content (M3, #629): the Typography section's variable picker also lists number variables for a text layer's content, as the engine already accepted (the text shows the number, e.g. 42.5, and follows the variable). Other text fields keep their own types. Numbers unchanged.
- **2026-10-17** · Frames and sections convert into each other (M1, #719): `object.convertToSection` turns a frame on the page or in a section into a section in place (same box, name, fill and layers, which stay where they are; what only a frame has goes: auto layout, clipping, layout grids, guides, scrolling), and `object.convertToFrame` now also turns a section into a frame that keeps its fill and doesn't clip, so the look stays. Nested frames, instances, main components and sections holding sections are refused with the reason; one undo step each. Both are in the canvas context menu (Convert to section / Convert to frame) and in Spanish. Numbers unchanged.
- **2026-10-16** · Adobe Fonts found on macOS (M7, #298): fonts activated through Adobe Fonts rendered in the fallback font without a warning, because Creative Cloud keeps them in its own folder as hidden files named by number, which the font scan never read. The macOS scan now includes `~/Library/Application Support/Adobe/CoreSync/plugins/livetype`, and files without an extension are indexed when they start with a font signature. Fonts are read in place. Regression test in the font database tests; docs/cli.md lists the folder. Numbers unchanged.
- **2026-10-15** · Dev Mode tail (M6, gap 16): generated layer trees start with the import lines of the connected components they call, once each per language; Compare changes takes any saved version as the before; a behaviour suite made the Dev Mode canvas inspect-only (no moving, resizing, drawing or editing; menus and shortcuts refuse design edits with a note), added Changed to the ready list and refused statuses on layers inside frames; read-only plugins show their panels from the Inspect tab. Dev Mode 97% / 76%.
- **2026-10-14** · Corner handles win over the corner-radius handles (M1, #292): a press on the inner half of a corner handle's square picked the radius handle 10 px in, so the drag neither resized nor (dragging outward) rounded anything. Inside the drawn square the corner now resizes both sides (in ratio with Shift or the link); further in the radius handle still answers. Regression test in the transform behaviour suite. Numbers unchanged.
- **2026-10-14** · Home screen recent files as cards (M7): each card shows the picture its file carries on its first page's colour, the file's icon, name and "Edited ... ago"; sort by name, creation or last change either way; a table view; a click selects, a double-click or Enter opens, arrow keys move; Open, Show in folder and Remove from recents on right-click. Numbers unchanged.
- **2026-10-13** · Vector edit mode regions (M4): with the Move tool, the filled area under the pointer is hatched with "/" lines in the selection colour, clipped by the area's own fill rule so holes stay clear; the hatch goes near points (3.5 px) and segments (2 px), and its spacing follows octave folding (8 px at every power-of-two zoom), fixed to the shape. A click selects the area with its points (they move with it but keep the unselected look), Shift adds or removes one, a press on a selected area drags the whole selection as one undo step, and a click on an empty spot clears the areas and stays in edit mode. `vector.edit.select {regions}` and `vector.edit.state` (`regions`, `regionVertices`, `regionHover`); `vector_regions_tests` and `regions_tests`. Numbers unchanged.
@@ -316,15 +356,179 @@ Ranked; detail and milestones in [`docs/roadmap.md`](docs/roadmap.md).
| Date | Change | Summary |
|---|---|---|
| 2026-10-30 | minor | M6, #1646: a layer pasted from another design app whose image the clipboard didn't carry came in flat grey, like a grey the design chose, and the paste said only that some kinds of content were left out. Now the placeholder an import stores for an image it doesn't have is light grey crossed by darker diagonal stripes (REST JSON, `.fig` and paste alike), the fill keeps the image's own name, the `.fig` and paste note is a fixed, translated one per image ("image fills whose image isn't in the file show a missing-image placeholder and keep its name"), `edit.pasteDesignLayers` returns `missingImages` (the names of the pasted layers concerned), the paste's message names them, and a paste with notes opens the import report like an import. Tests: `fig/missing_image_tests.rs`, `paste_missing_image_tests.rs`, `import_report_tests.rs`. |
| 2026-10-30 | minor | M6, #1638: layers pasted from another design app lost the colours of their shared paint styles: a component's glyphs, and the same glyphs in its instances, came in black. The clipboard links each style by the guid it has in the library it came from, which the clipboard doesn't hold, as well as by its library asset, whose copy it does hold; the `.fig` mapping took the guid and found nothing. Now a guid names a style, variable collection or other resource only when the file holds a node by it, and otherwise the asset it also names finds the file's copy, so `.fig` import and paste agree. A link to a style the file really doesn't hold keeps the colours stored with the layer, as before, now with a note. Tests: `paste_style_tests.rs` (design data built in code, read as paste reads it). |
| 2026-10-30 | minor | Docs (found while landing #1649): merges that kept both sides of a conflict had left rows and status items twice. One row each now for `.fig` in `docs/file-format-parity.md` (four copies: #1234, #1236 and the #1076 ZIP budget kept, with the newer four-values-per-byte budget wording and #1649's connector routing), Real-time co-editing (three: #1069, #1065, #1063) and Live cursors (two: #1380, #1416) in `docs/collaboration-parity.md`, Windows x64 / arm64 in `docs/hardware-parity.md` (#1320, #965) and M0.9 in `docs/roadmap.md` (#1396, #996). The status lines of this file, `docs/roadmap.md`, `docs/gaps.md`, `docs/target-app-parity.md`, `docs/ui-parity.md`, `docs/hardware-parity.md`, `docs/localization-parity.md` and `docs/file-format-parity.md` drop items listed more than once (302 in all, the first kept), and an unclosed parenthesis in target-app-parity's status line is closed. No wording changed otherwise. New `cargo xtask docs`, run by `ci`: a table row that repeats an earlier one (header rows apart) or a status item listed twice fails it. Numbers unchanged. Test: `docs_check::tests`. |
| 2026-10-30 | minor | M3, #1613 (after #1618): the rest of the stored-id work. `Variables::get` and `Variables::collection` take the variable or collection with that exact id before one named like it (as `find_mode` does), so typed ids in `variable.bind`, `variable.update`, aliases and the prototype action and expression input can't be taken by a renamed one. Stored mode ids go through the new `Collection::mode_by_id` (expression evaluation, display and type checks, the prototype's mode switch), font family variables read their stored ids exactly, and the colour picker, prototype editor and panels' labels (variables, prototype, selection colours, typography, style picker, assets, variable fields) read exact ids. Tests: `stored_ids_tests.rs` (exported aliases, deletion, Dev Inspect, expressions, mode switch, library import, interaction edits) and the doc crate's lookup test; #1618's expression test now expects the exact id to win. |
| 2026-10-30 | minor | M6, gap 6 (first slice): plugin widgets have hover styles, as the reference tool's widgets do. Any layer of a widget description, the widget itself too, may have `hover: {fill?, stroke?, opacity?, color? (text)}`, checked like the rest; on the canvas, while the pointer is over such a layer, it shows that style as a live preview (the text-preview mechanism: no undo step, never saved or sent to a live session), and moving away, leaving the canvas or any command (a click on the part acts on the document itself) puts it back; another live preview showing isn't replaced. `widget.hover {ids}` does the same for scripts and agents. No new plugin API: older FigCraft draws the layer as it is. The Counter example's buttons darken under the pointer. Timers and panels from widgets are still missing. Numbers unchanged. Tests: `widget_hover_tests`, `widget_tests::hover_styles_are_checked_and_normalised`. |
| 2026-10-30 | minor | M6, gap 1: `.fig` connectors saved without their path that are elbowed or curved were drawn as straight lines. They now run as FigCraft's board connectors do: in right angles (around the layers they join when the short way would cross them) or along one curve, leaving each layer from the side the file's `magnet` names, else the edge its point lies on, else the side facing the other end. The router (`route_points`) moved from the engine's boards into `figcraft-doc`, unchanged, so the importer and the board connectors share it. Straight connectors are unchanged. Checked by rendering the synthetic fixture with figcraft-cli; not on real files. Numbers unchanged. Tests: `fig::tests::tail::connectors_without_a_saved_path_follow_their_line_style`, the elbowed connector in `other_layer_kinds_import_as_the_closest_layers`, the engine's board connector tests. |
| 2026-10-30 | minor | M6, gap 6: plugin panels (and plugin command forms) take a `slider` control: a number between a required `min` and `max`, drawn as a slider with the value field beside it, named for screen readers as a slider with its range. Dragging shows the value live and calls the plugin back once, when let go (one undo step); typing in the field or double-clicking the slider (back to the panel's value) do the same. The slider is the one the Type settings' variable axes used, moved into the shared widgets, which also gives those axis sliders a screen-reader name. Numbers unchanged. Tests: `v2_tests::panels_are_validated_and_keep_entered_values`, `plugin_panel_tests::a_slider_calls_the_plugin_back_once_when_let_go`. |
| 2026-10-30 | minor | M6, gap 16: `figcraft-cli connect publish <file.figcraft> [--mappings FILE] [--dry-run]` publishes a code repository's code connect mappings (`figcraft.connect.json`, as Export code connections writes it) to a design file: each link connects the components or sets of that name to their code for its language; links that don't fit are warnings and are skipped; the file is saved in place through the engine's `file.save` (whole or not at all, version history kept); a dry run saves nothing; nothing linked is exit code 1. Numbers unchanged. Tests: `files::tests::connect_publish_links_components_and_saves_the_file`, `args` tests. |
| 2026-10-30 | minor | M8, #1571 (part 3): the first frame after each move of a drag rebuilt the selection's outlines and box, its slots, the selection spacing and the Design panel's values from the new document. Now the engine notes when a drag has only moved the selected layers, each by the same offset from where it started (`Interaction::moved_by`: not for a layer in an auto layout flow, nor an Alt copy), and the canvas and panel work those out once from the document the drag started from and shift what depends on position (the outlines on screen, the box); the Design panel shows that document's values, all but the position fields, which follow the move. The Export section's values are kept too. Nothing looks different. Dragging 648 of about 1,900 synthetic layers (release build, two runs each): the UI pass 2.10 to 1.44 ms a frame, CPU 135% to 117% of a core; the Design panel 0.9 to 0.5 ms, the overlays 0.5 to 0.3 ms. Not yet: the engine's preview of each move (copying and normalising the document) and the canvas texture's whole upload. Tests: `moved_by_tests.rs`, `canvas/selection_cache_tests.rs`. |
| 2026-10-30 | minor | M12, #1597 (the third point): Shift+Cmd/Ctrl+Enter (`board.nextSticky`) now works on a selected or edited shape too: it adds the next shape of the same kind, size and colour on its right (or the `side` given), joined to it by a connector and with typing going on in it, as the "+" beside it does (`board.addConnected`, #1631). On a sticky it adds a sticky beside it without a connector, as before. Numbers unchanged. Tests: `board_connected_tests::the_next_sticky_shortcut_joins_the_next_shape_and_leaves_stickies_as_they_were`, `board_ui_tests::the_next_shape_from_the_keyboard_comes_connected_and_goes_on_typing`. |
| 2026-10-30 | minor | M12, #1597: boards create the next shape already connected. A selected sticky or shape shows a "+" beside each of its four sides (clear of the contextual bar, the size label and the toolbar); hovering one shows where the next goes and the connector to it; clicking makes a sticky, or a shape of the same kind, size and colour, 80 units away on that side, joined to it by a connector the way the Connector tool makes one, and starts typing in it. One undo step takes both away. `board.addConnected {id?, side?, edit?}` does it from commands, MCP and plugins. Shift+Cmd/Ctrl+Enter still adds a sticky without a connector. Tests: `board_connected_tests.rs` (engine), `board_connect_tests.rs` (the handles, headless). |
| 2026-10-30 | minor | M4, #1292: a `.fig` import said a family was not installed and drew it in a fallback font although the font picker listed it. Fonts name a family two ways: a typographic name ("Economia", style "Medium") and a legacy name per weight ("Economia Medium", style "Regular"); fonts without typographic names are installed under the legacy ones only, while a file may use either. The font database now resolves a family + style either way (`FontDb::resolve`, used by every lookup: import, paste, REST import, rendering, the missing-fonts check and the font report): a family named with weight words finds the family without them, a style the family lacks finds the family named with it, and the Style list offers those weights. Family names also match across case, spaces, hyphens and underscores, and a weight number style ("600") goes by that weight. `.fig` and REST imports wait for the installed fonts to be indexed before their text is laid out. Widths aren't taken for weights (a missing "Roboto Condensed" stays missing). Tests: `fontdb_names_tests.rs` (synthetic names), `fig_font_names_tests.rs` (a `.fig` round trip). |
| 2026-10-30 | minor | M0.9, part of #1396: a development example generates completions for Bash, Elvish, Fish, PowerShell and Zsh and a man page for the CLI and each subcommand, nested ones included, from figcraft-cli's clap definitions, without running commands. Generating these files doesn't install them; package installation and the other RFC-0012 exceptions remain outstanding. Regression tests in `apps/figcraft-cli/tests/it/generated_docs.rs` cover artifacts, command names and options, matching Bash dispatch states, repeat generation and filesystem errors. Numbers unchanged |
| 2026-10-30 | minor | M8, #1076: bound total `.fig` ZIP image reads to 4 GiB (1 GiB on 32-bit targets) of decompressed ZIP-member bytes containing encoded images, keeping the 256 MiB member cap. Members that do not fit are skipped before allocation and counted in the import warnings; failed reads still consume their bytes, and repeated references keep their cached resolution. Embedded message images, decoded pixel memory and native `.figcraft` reads are unchanged. Regression tests in `fig::tests::image_budget` (format). Numbers unchanged. |
| 2026-10-30 | minor | M8, #1114 (the rest): `file.open` and `file.import` (and recovery copies), `image.place {path}`, the image fill's file picker and `figcraft-cli run`'s script read the path they were given whole, so a device (`/dev/zero`), a pipe or an absurd file could fill memory or wait forever. Every whole-file read of a path now goes through `file_access::read_capped`, which looks at the path before opening it and refuses a folder, a device or pipe, or a file over its cap, with a message saying which, and reads at most the cap even if the file grows; the uncapped `file_access::read` is gone. Caps: documents 16 GiB (a `.figcraft` keeps its images as they are: #1313's file is 7.4 GB on disk), placed images 256 MiB (what a document takes from outside, `ImageData::MAX_BYTES`), scripts 1 GiB; fonts (256 MiB), plugin files and variables files already had theirs. Numbers unchanged. Tests: `path_read_caps_tests`, `file_access` tests, the CLI's `a_script_that_isnt_a_regular_file_is_refused`. |
| 2026-10-30 | minor | M8 (found while retaking the README screenshots): at the default (and narrowest) 240 px right panel, a component set's variant property rows ran off the panel: values past the second ("Outline", "Disabled") were cut off, and the + after them and the − of each property sat past the edge; a shared boolean property's − was pushed out too. Values now wrap as many to a line as fit (two at 240 px, three from about 320 px as before), the property name field gives way first so the default and − stay in view, and a boolean's checkbox no longer takes the whole row. Numbers unchanged. Tests: `component_set_narrow_tests`. |
| 2026-10-30 | minor | M9, #1130: the server key is compared by its hash. `Store::manage` hashes the presented and the kept key with SHA-256 and compares the two digests; the key on disk, `figcraft-server key` and rotation are unchanged. Regression tests in `store_tests` (`operator_key_*`). Numbers unchanged |
| 2026-10-30 | minor | M8, #1097: while the Assets panel showed a file, the app asked the server of every team library the file named for updates, with the link's token and a usage report, so a file someone sent could make the app contact any server. Now only team libraries you added (`library.add`) or checked on this computer are checked by themselves (kept with the preferences as hashes of their links); another one waits in **Assets ▸ Libraries** as "Not checked yet" until you click **Check for updates** (`library.check`, which plugins can't call). Placing or applying something from it is your own action and asks its server as before. Numbers unchanged. Tests: `library_beacon_tests` (engine and UI). |
| 2026-10-30 | minor | M8, #1121: client-side maps were unbounded. A hostile host or server could mint people without end: the presence waiting for the UI, the people a connection had seen and the session's list of who is here each grew with every new id; and a session server's set of deleted layer ids (kept so undo may create them again) only grew. Now a client keeps track of at most `MAX_PEERS` (4,096) others: past it, someone new is left out with a logged warning (in a welcome too), someone already here can still be renamed, and someone leaving makes room; a server's `max_people` above 4,096 is lowered to it with a startup warning (the server still starts). A session remembers at most 262,144 deleted ids, forgetting the lowest first (on restore too). Threat model in `collab-protocol.md`. Tests: `client_cap_tests.rs` (collab-net), `live_peer_cap_tests.rs` (engine), `deleted_cap_tests.rs` (collab), the server's `max_people` clamp. |
| 2026-10-30 | minor | M8, #1117:`doc.list` returned the path on disk of every open file to any caller, plugins and attached AI agents included. It now gives an agent's calls (MCP, or a control call naming an `agent`) and plugins each file's `id`, name and state without `path`; the user's own scripts (the CLI, control calls without an agent) still get paths, and the app's UI reads its tabs directly. Numbers unchanged. Tests: `doc_list_paths_tests`. |
| 2026-10-30 | minor | M8, #1571 (part): a large selection redid the same work on every frame. A pointer move draws two to five egui frames, and each one worked out every selected layer's outline, the selection box and size, the slots shown, and the Design panel's instance checks and position values (X, Y, rotation, constraints, alignment) again, though only the first frame after a move or an edit can differ. Now they are kept with the document snapshot, the selection and, for the overlay, the view's mapping to the screen and the overlay green, and made again when any of those changes (`selection_cache`); the hover and drop-target highlights are drawn every frame as before. Nothing looks different. Dragging 648 layers of a synthetic file of about 1,900 (release build, control-driven, two frames per move): the canvas overlays 1.99 to 0.86 ms a frame, the Design panel 2.54 to 1.40 ms. Not yet: the frame of each move still works everything out (the selection spacing, the export section, the outlines), the canvas texture is uploaded whole after each render. Tests: `canvas/selection_cache_tests.rs`. |
| 2026-10-30 | minor | M8, #1089: libraries and plugins larger than one address's download burst (16 MiB) could never be downloaded: the per-address token bucket never held more than its burst, so a private server's 40 or 64 MiB team library was refused for good. Now a download larger than the burst is served once the address's budget is full, leaving it in debt that refills at 1 MiB a second before its next download (`Bucket::take_spread`, also for everyone's library budget); within the burst nothing changes. Over any stretch of time an address still gets at most the largest object plus the rate times that time. `server.md` (rates, threat model). Tests: `download_burst_tests.rs` (the bucket's loan and payback, ten minutes of 40 MiB requests kept to the rate, a 40 MiB library served and then the address waits). |
| 2026-10-30 | minor | M8, #1099:`command.batch` (and every grouped run: plugins, scripts, MCP `batch`) kept a full undo snapshot per step until the batch ended, so a long batch held one document copy per step. Inside a group, edits now add to the one undo step the group pushed, so a batch holds one snapshot however long it is; it is still one undo step that restores exactly the state before it, and a failure still rolls everything back. An `edit.undo` step inside a batch now undoes the batch's edits so far together. Measured on a 2,000-layer file with 160 KB of document colours (debug build): a 1,000-step batch went from +165 MiB peak and 80 to 100 ms to +10 MiB and about 50 ms (4,000 steps: +35 MiB, the step list and command journal). Numbers unchanged. Tests: `batch_undo_snapshot_tests`. |
| 2026-10-30 | minor | M8, #1101: the importers' 256-level depth cap could overflow the browser's 1 MB stack before it was reached. Measured per level: the REST import's layer walk (also the second stage of every `.fig` import) took about 4 KB, a whole megabyte at 256, and overflowed a 1 MB thread; it now runs on a work list in the same order (groups fitted once their children are in), a few KB in all. The `.fig` mapping, which builds each layer from its children's results, stays recursive at about 1.8 KB a level, 455 KB at the cap. Desktop and browser keep the 256-level cap, so nothing that loads today is cut off. Numbers unchanged. Tests: `deep_import_tests` (256 and 400 levels on a 1 MB thread). |
| 2026-10-30 | minor | M8, #1082 (the rest): the SVG and PDF exporters drew a pattern fill's source from depth 0 again, and the renders they embed (a layer with effects drawn alone, a pattern cell with effects) started at 0 too, so an export could nest its own levels, the renderer's and a pattern's on top of each other. A pattern source now exports one level below the layer it fills, and what the renderer draws for a layer goes on from that layer's depth (`figcraft_render::at_layer_depth`), so one cap bounds the whole export; the render of what lies beneath a layer (background blur, glass) is of the page's own layers and still counts from the page. Canvas masks, mask content and boolean and outline walks already carried the depth or have their own small caps. Numbers unchanged. Tests: `render_depth_tests`, `figcraft-svg` and `figcraft-pdf``pattern_depth_tests`. |
| 2026-10-30 | minor | M8, #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; its source now goes on from the depth of the layer it fills, and the one cap bounds the whole render. Measured, a level takes about 6.5 KB of stack in a render (6.8 KB in a debug build), so the deepest allowed render needs about 1.7 MB: the canvas render threads, the save thread (it renders the thumbnail) and a band pool of their own now get 8 MiB stacks, like the desktop app's main thread (rayon's and other threads' default is 2 MiB). Numbers unchanged. Tests: `render_depth_tests` (the cap reached through a pattern; three 250-level chains of patterns in patterns rendered on a render thread's stack). |
| 2026-10-30 | minor | M8, #1114 (explicit variables files only): sibling variables files were read with a 256 MiB cap and a regular-file check, but an explicit path given to `variable.import`, or as the `variables` of `file.open` and `file.import`, was read whole with no limit. That path now goes through the same capped read: an oversized file, a folder or a device fails with "not a file of at most 268435456 bytes" and the document, tabs and recent files stay as they were. The check reads the file's size before reading it, so a file that grows or is replaced in between is not bounded; primary document reads and `image.place` are left for the rest of #1114. Tests: `import_export_tests::an_explicit_variables_path_is_capped_like_the_sibling_file`, `variables_io_tests::variables_import_caps_the_file_it_reads`. Numbers unchanged |
| 2026-10-30 | minor | M8, #1074: on `figcraft-server` every join of an idle file read its snapshot and log back and every join and leave rewrote its `meta.json`, all on the hub thread, so join/leave loops amplified into disk I/O. A file everyone left (its log synced) now stays loaded for a minute (at most 16 such files, 256 MiB together), and a join in that time takes it as it was with nothing read; its open time, which expiry counts in days, is written at most hourly and when the server stops. Edits are saved as before (logged, compacted when everyone leaves). Measured: 20 join/leave cycles of an unchanged file did 0 reads and 0 writes (40 reads before, plus a `meta.json` write per join and leave). Numbers unchanged. Tests: `idle_io_tests`. |
| 2026-10-30 | minor | M8, #1079: one budget for everything a `.figcraft` file unpacks when it opens. Each area had its own cap (document 4 GiB, images 4 GiB, library 1.5 GiB, versions 1 GiB, branches 1 GiB), so one file could make the app unpack about 10 GB. Now they also share one total, 6 GiB (2 GiB on 32-bit targets such as the browser), counted member by member in reading order: document, images, published library, version history, branches; the per-area caps stay. The document (a 900 000-layer one unpacks to 1.5 GB) and the images up to their own cap fit within it, so a file whose images pass 4 GiB (#1313) keeps as many as before, and images left out by the total are treated like those past the images' cap: they stay in the file and saving over it is refused rather than dropping them. History and branches past the total are skipped, and a document past it refused, each with the note "the file's contents together pass … MB". Numbers unchanged. Tests: `unpacked_total_tests`. |
| 2026-10-30 | minor | M8, #1571 (part): laggy drags and high CPU while the pointer only moves. Measured on a synthetic file of about 1,900 layers (24 auto layout frames of 48 rows, 600 shapes, 24 images; release build, 300 to 600 pointer moves at 60 per second): every pointer move draws about three egui frames (the move, egui's follow-up, a finished render), and the canvas hit-tested the hover on each of them; a drag of an auto layout frame laid its contents out again on every move, though a move can't change them. Now a hover where nothing changed since the last (same place, file, page, revision, selection, tool and modes) isn't sent to the tools again, and a frame that only moved keeps its own layout (its parent's still runs). Hover: the UI pass 0.91 to 0.49 ms a frame, CPU 33% to 22% of a core; dragging 648 layers: the UI pass 3.8 to 3.2 ms, the engine's normalise 3.8 to 2.1 ms a move, CPU about the same (117% to 116%). `ui.inspect` reports `frameTimes` ({frames, meanMs, maxMs}). Not yet: the selection overlay and the Design panel recompute every frame for a large selection (about 2 ms each at 648 layers), the canvas texture is uploaded whole after each render, and egui draws its follow-up frames. Tests: `canvas_hover_tests.rs`, `auto_layout_move_tests.rs`. |
| 2026-10-30 | minor | M8, #1077: two quadratic costs in the `.fig` import's instance expansion. Variable and collection nodes sat in the layer tree, so every instance of a main walked the ones under it again, and as they never become layers the layer cap never stopped them; they are now left out of the tree (variables are read from the node list as before), so everything an expansion visits counts against the cap. A nested instance copied all of its outer instance's override rows, and its property values and slot contents; it now takes only the rows addressed below it (they sort together, so the rest aren't visited) and shares the outer values, keeping only its own. Normal files map as before. Crafted files: 20 000 variables under a main with 2 000 instances 13.8 s → 21 ms; 50 000 override rows over 2 000 nested instances 40.5 s → 98 ms (native release, the same output). Numbers unchanged. Tests: `expansion_cost_tests`. |
| 2026-10-30 | minor | M9, #1415 (the rest): `comment.people` lists every name a message @mentions besides those who wrote or reacted (once each in any case, a known name spelt as the file spells it, never an email address), so a name that has only been tagged shows up. `comment.next` takes the `mentions` filter and, like `comment.open`, returns the thread as `comment.list` lists it, with each message's `mentions`; `comment.prompt`'s bounded `comments` carry each thread's mentions (at most 32, names cut to 80 characters). One mention helper serves them all. Numbers unchanged. Tests: `comment_mentions_tests`. |
| 2026-10-30 | minor | M8, #1083: grid auto-placement tested every cell of a child's span at every position it tried (children × rows × columns × span area), so a crafted grid of wide children kept every layout pass busy. The grid's taken cells are now a bitmap: a span is checked a 64-bit word at a time, the search jumps past the cell that blocks it, a row's search ORs the span's rows once, and a column-anchored child looks on from where the last one of the same column and spans found room (cells only ever get taken). Placements are the same, checked against the old search on 400 random grids. A crafted 1000 × 1000 grid with 1000 full-width children: 192 ms → 0.3 ms; 3000 column-anchored wide children: 4.2 ms → 0.7 ms; ordinary grids unchanged (20 000 1 × 1 children 5.2 ms either way). Numbers unchanged. Tests: `grid_place_tests`. |
| 2026-10-30 | minor | M1, #1430: from the pen into vector edit mode. A double-click with the pen drew on: its second press added a point on top of the last one (a zero-length segment), a double-click meant to edit a layer drew a stray two-point vector on it, and Enter ran what Escape does. Now a press on the pen's last anchor adds nothing; a double-click while drawing ends the path at that point, and one that only started a path drops it and edits the layer under the pointer; Enter ends the path. Both go on editing the vector with the Move tool (points and handles); the next Enter leaves edit mode. In edit mode the pen's double-click and Enter take the Move tool too, a double-click taking back the point or new path its first press began. Escape still keeps the pen in edit mode for the next path (#760). Numbers unchanged. Tests: `pen_vector_edit_tests`. |
| 2026-10-30 | minor | M1: desktop recovery copies made while a version is previewed held the version shown instead of the file, 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, `figcraft_engine::branches::recovery_copy`: the document itself (`DocState::document_itself`, the same document the unsaved-changes check compares, not the version or live text preview shown in its place), its version history and its branches. The desktop copies and the browser's (#1580, which already took the document itself) both use it. Tests: `recovery_tests.rs` in the desktop and web apps (a copy made during a version preview recovers the file and its history). |
| 2026-10-30 | minor | M12, #1342: a FigJam `.jam` file opens as a board, our whiteboard file kind, instead of a design file: the `.fig` reader marks a canvas with FigJam's `fig-jam.` header as `FileKind::Board`, so the file opens with the board toolbar, the dot grid and the board panels, and saves as a board. A `.fig` stays a design file, and `file.import` of a `.jam` adds its pages to the open file without changing its kind. Its stickies and shapes with text are still frames with their generated layers and its connectors vectors, not native board objects. Tests: `board_import_tests.rs` (the kind, `.jam` against `.fig`), `jam_open_tests.rs` (open as a board, import into a design file). |
| 2026-10-30 | minor | M6, #1342: FigJam `.jam` files open. A `.jam` is the same archive as a `.fig` (its canvas header reads `fig-jam.`), and the `.fig` reader already took it when a copy was renamed; now File ▸ Open, its file filter, drag and drop (home screen and canvas), `file.import` and the command line take `.jam` too, and the home screen's recent files show their thumbnails. What comes in is what the `.fig` reader makes of FigJam layers: stickies and shapes with text as frames with their generated layers, tables, stamps and the like as frames with an import note, connectors as vectors, and a layer type it doesn't know left out with a note naming it. The file opens as a design file, not a board, and its objects carry no board metadata yet. Tests: a synthetic `fig-jam.` canvas, bare and zipped (`board_import_tests.rs`), a `.jam` made from our own `.fig` export opened and imported (`jam_open_tests.rs`), the extension checks (`home_tests.rs`). |
| 2026-10-30 | minor | M8, #1574: the Actions palette (Ctrl+K) offered engine commands that need parameters it can't give, such as `image.place` ("Place image", needing the image's `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`), and a row no menu item gives whose label reads like a menu item's (any case, without the ellipsis). Place image runs File ▸ Place image…, which opens the file dialog. Tests: `palette_tests.rs` (no row needs parameters unless a menu item runs it, no duplicate label across several searches), `unknown_params_tests.rs`. |
| 2026-10-30 | minor | M9, #1525: share links can comment. A third access level, `comment`, sits between `edit` and `view`: someone who joins with a comment link (or its join code) sees the file live, as with a view link, and adds comment threads, replies, reactions and resolves them; any other edit is refused, by their app and by the server, which 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, and the Share popover offers Can edit, Can comment and Can view; the toolbar shows a "Can comment" badge with Move, Hand and Comment. `collab.share {access: "comment"}`, `collab.manageFile {action: createLink, access: "comment"}` and `figcraft-server links create FILE --access comment` make and pick it. Older apps are unaffected: one that joins with a comment link is welcomed as a viewer, links listed to it read as view links, and a level an app doesn't know reads as view; an older server makes no comment link and sharing goes on without one. Comment authorship is the name in Preferences, as before (no accounts), so a commenter can edit or remove other people's threads. Tests: `comment_access_tests.rs` (collab), `comment_link_tests.rs` (engine, server end to end), the Share popover and toolbar in `live_share_tests.rs` and `live_server_tests.rs`. |
| 2026-10-30 | minor | M1, #1390: Preferences ▸ Scroll wheel zoom (`pref.scrollWheelZoom`, off by default, kept with the preferences). On, a mouse wheel zooms the canvas about the pointer at the rate Cmd/Ctrl+wheel zooms otherwise, and Cmd/Ctrl+wheel scrolls; Shift+wheel still scrolls sideways, and trackpad scrolls, pinches and touch screens are unchanged. A wheel is told from a trackpad by what the platform reports: macOS names the device; elsewhere a wheel scrolls by lines and a trackpad by pixels, so a mouse whose wheel reports pixels (Chromium-based browsers) keeps scrolling, and a trackpad that reports lines (some X11 setups) zooms. Numbers unchanged. Tests: `wheel_zoom_tests`. |
| 2026-10-30 | minor | M9, part of #1415:`comment.list` takes `mentions` (a name, any case, a leading `@` allowed) and lists only the threads where a message, the first or a reply, @mentions it, combined with the other filters (`{"mentions": "Claude", "filter": "unread"}`). It matches the names each message's `mentions` reports, by the same parser the thread view highlights with (`figcraft_doc::mentions`), so `@Ada Lovelace` is Ada Lovelace, not Ada, and an email address mentions nobody; the name counts as one of the file's people, so an unknown name can be several words. Numbers unchanged. Tests: `comment_mentions_tests`. |
| 2026-10-30 | minor | M8, #1123: a safety net for internal errors in the browser. wasm32 can't catch a panic (stable Rust has no unwinding std for it), so there the command guard and the frame guard do nothing and a panic stopped the page, losing every unsaved change. The page now follows its files with unsaved changes after every frame (shared copies, taken again only when a file changed); its panic hook writes them to the page's local storage as `.figcraft` copies, as many as fit (browsers give a page about 5 MB), and says on the page that it stopped and what reloading brings back; the next page opens them again as unsaved files (`file.recover` also takes the copy as base64 `data`). Numbers unchanged. Tests: `recovery_tests` (figcraft-web). |
| 2026-10-30 | minor | M8, #1104: boolean operations had no input cap, so a crafted file could keep the sweep busy for unbounded time and memory (any number of segments, crossing as often as they like, and a fold re-sweeping its growing result for every operand). Each sweep now takes at most 250 000 segments and 2 million possible crossings (pairs of segments whose bounds overlap, counted exactly in O(n log n)), and a fold at most 4 million segments swept over all its steps; past them the operation is refused with `TooComplex` ("the shapes are too complex to combine") before it runs. The example files' boolean groups sweep at most 55 segments and 155 possible crossings per step. A boolean group past the caps draws without its result, with a warning, once, like layers past the render's memory budget; the refusal is remembered, so redrawing it costs nothing; Flatten refuses it instead of dropping it. Numbers unchanged. Tests: `budget_tests` (pathops), `boolean_budget_tests` (render), `boolean_cap_tests` (engine). |
| 2026-10-30 | minor | M0.9, #1396:`figcraft-cli` parses its arguments with clap, following RFC-0012 (craft-contrib), so help comes from the same definitions the parser uses and bad arguments fail before any work starts. Changed syntax: an option given twice is a usage error (it used to take the last value; the automation roots still repeat), a value starting with `-` needs `=` (`--filter=--json`), `--o` for `-o` and the `version` subcommand are gone (`--version` stays), `--format` is checked before the file is opened, help and usage errors are clap's text (`Usage: figcraft-cli …`, `error: …` with a hint instead of the whole usage). Every subcommand, flag, value form and exit code is otherwise unchanged. Tests: `args::tests` (each documented form, every usage error, every subcommand's help) and the CLI integration tests. Numbers unchanged |
| 2026-10-30 | minor | M7, #1506: the hover and selection outline of a rectangle, frame, component, instance or section with rounded corners follows its geometry: each corner's own radius and the corner smoothing, as the fill draws them, whatever the stroke alignment (inside, centre or outside); square corners keep the box. It used to be the square box, so it matched only the square corners of a layer with mixed radii. Tests: `canvas::outline_tests`. Numbers unchanged |
| 2026-10-30 | minor | M3, #1536 and #1537: component properties are listed in the order they were made in, and renaming one keeps its place (they were sorted by name, so a rename moved the property); that holds for every kind, on every variant of a set and on instances, in `component.properties`, the Design panel and `.fig` export. Files from before keep their properties by name until one is added or renamed. Make slot (the Slot menu of a frame inside a component) asks for the slot's name first: a field under the Slot row with the frame's name selected, Enter or Make slot to make it, Escape to cancel; a name taken keeps the field open. Renaming stays in the component's Properties list. Tests: `engine/src/prop_order_tests.rs`, `ui-egui/src/make_slot_naming_tests.rs`. Left: dragging the shared properties to reorder them, and renaming a slot with only its frame selected. Numbers unchanged |
| 2026-10-30 | minor | M9, #917: the browser build publishes to a `figcraft-server`. Share ▸ Online (the server in Preferences) and Share ▸ Publish to server… upload the open file over the page's own WebSocket, make the links and join codes others join with, and keep the tab in sync with the new file, as the desktop app does; asking a server about its hosting, stopping a share and sharing again go out the same way. The Share popover offers only online sharing there (a page can't listen, so the local network stays the desktop app's); without a server in Preferences it says so and offers joining. A failed publish says why and leaves the file as it was. The browser keeps only your name across reloads, so the owner token of a file it put on an open server lasts as long as the page (`crates/collab-net/src/exchange.rs`, run by the browser client's driver; the requests and what their answers mean are shared with the desktop transport in `requests.rs`). Tests: `exchange_tests` (scripted server: with and without the key, sharing or not, refusals, an old server without join codes, time-outs, leaving while publishing, the background requests), `live_join_tests`; checked in a browser against a local server. Percentages and hours unchanged |
| 2026-10-30 | minor | M4, #1051: fonts installed on this computer in the browser build. Where the browser lists them (Chromium's Local Font Access, `window.queryLocalFonts`), the Design panel's font menu offers "Use fonts installed on this computer…"; the browser asks for permission, and the fonts join the font list by name, with their styles. A font's bytes are read only when text uses it, one style at a time (its preview in the menu waits until then), and auto-sizing text fits it once it is in; one that can't be read is dropped and its text keeps the fallback. A family the build embeds keeps its faces; the computer's styles it lacks join it. Other browsers show no row (`figcraft_text::add_local_fonts`, `apps/figcraft-web/src/local_fonts.rs`). Tests: `fontdb::local_tests`, `local_fonts_tests`; checked in a browser. Percentages and hours unchanged |
| 2026-10-30 | minor | M8, #1550: SVG import reads CSS colour functions its SVG parser doesn't (it painted them black, or left strokes out, without a word): `color(srgb …)` and `color(display-p3 …)`, with alpha, converted to the document's profile, and `color-mix(in srgb|oklab|oklch, …)`, evaluated at import (percentages, alpha and nested mixes as CSS Color 5 has them), in colour attributes, `style` attributes and `<style>` sheets. Other colour spaces are taken as sRGB values and other mixes mixed in sRGB, each with an import note; a call that can't be read becomes a mid grey with a note. Bounded: calls over 2,048 characters, mixes nested more than 8 deep and more than 100,000 calls are not read. Tests: `css_color_tests` (svg). Numbers unchanged |
| 2026-10-30 | minor | M1, #1320: settings in portable mode. The settings folder now comes from craft-settings, the suite's rules: `FIGCRAFT_CONFIG_DIR` names a folder (set but empty: none, nothing read or written); a `portable` file next to the executable (next to the `.app` on macOS) puts the preferences, recovery copies and installed plugins in `config/` beside it; otherwise they stay in `FigCraft` in the platform folder, where every earlier version kept them, so nothing moves for an installed app. The Windows portable zip now ships the `portable` file. A portable folder without preferences yet starts with a copy of the platform folder's `state.json` (owner-only, the original left for an installed app), so a portable copy that kept its settings in `%APPDATA%` before keeps them after the update. Checked on Windows 11 with a portable copy beside an installed copy's settings: the first start took them (its name and start screen), a name changed there was kept in `config\` only, and without the marker the app read the installed settings again. Tests: `settings_dir_tests`. Numbers unchanged |
| 2026-10-30 | minor | M1, #965, #1438: "Parent device is lost" at start. Both reports lost the device while wgpu compiled its indirect-call validation shaders, which it builds with every device: on Windows Direct3D 12 took another program's `dxcompiler.dll` from the search path and failed without its `dxil.dll`; Intel Macs' Metal compiler aborts on them. Direct3D 12 now compiles with FXC (part of Windows; `WGPU_DX12_COMPILER` still chooses), and the validation shaders are left out (FigCraft draws nothing indirectly; `WGPU_VALIDATION_INDIRECT_CALL` turns them back on). When the graphics still fail before the window's first frame, the app starts itself again with the next attempt, from a fixed list: on Windows Direct3D 12/OpenGL, Vulkan, OpenGL, then the first again on the power-saving GPU with WebGL2-level device limits; on Linux and FreeBSD the defaults, OpenGL, then lower requirements; on macOS the defaults, then lower requirements. The last failure's message names what was tried; the log names the backend and GPU that drew. `WGPU_BACKEND` is still the one attempt. A restarted app waits up to 5 s for its `--control` port. Checked on Windows 11 with a local build that made the first one, two, three and then all four attempts fail: each fallback opened the window (Vulkan, OpenGL, Direct3D 12 on the integrated GPU) and four failures ended with the message. Not run on the reporters' machines or on macOS. Tests: `graphics_tests`, `startup_tests`. Numbers unchanged |
| 2026-10-29 | minor | German terminology corrections (M10): gradient stops, OpenType stylistic sets and the Home-screen preference; partial review, catalog remains a draft. Nine existing translations corrected; keys, placeholders, coverage, grade and estimates unchanged |
| 2026-10-29 | minor | M4, #1219: Thai, Lao, Khmer and Burmese text, written without spaces between words, wraps between words instead of where a line overflows (`พัฒนาซอ|ฟต์แวร์` → `พัฒนา|ซอฟต์แวร์`). `figcraft_text::breaks::word_breaks` finds the word starts inside each run of these scripts with ICU4X's line segmenter (`icu_segmenter` 2.3): its dictionaries on desktop, its LSTM models in the browser (about 0.39 MB of wasm instead of 1.86 MB, which the 24 MiB gate can't take; they split a few compounds the dictionaries keep whole). Where a run meets spaces, Latin or CJK the UAX #14 rules decide as before, and a word wider than its box still takes #1400's emergency break. Text without these scripts does no extra work. Grade unchanged. Evidence: `breaks::tests::thai_lines_break_between_words`, `lao_khmer_and_burmese_lines_break_between_words`, `word_breaks_survive_junk`, `behaviour_tests::thai_lines_break_between_words` (fails on main). |
| 2026-10-30 | minor | M8, #1416: an AI agent attached over MCP shows where it works, as a person in a live session does: its avatar in its own colour in the people row, named as its comments are signed ("Claude (agent)"), and an outline with its name around the layers its last edit changed or it last selected. It goes 8 seconds after its last command and shows only in the file it works in. Display only (no undo step, not saved, never synced to a live session); the engine records it in `Session::execute_as` (`figcraft_engine::agent`). Tests: `agent_presence_tests` (engine), `live::agent` (ui-egui) |
| 2026-10-30 | minor | M5, #405: a manual overlay can be placed on the canvas. While the interaction editor has an Open overlay action with Manual position open, the overlay frame shows dimmed and outlined where it opens, from the top-left of the layer that opens it (right below it without an offset), and dragging it moves it: the X and Y fields follow while dragging, and the drag is one undo step. Tests: `overlay_ghost` (ui-egui) |
| 2026-10-29 | minor | M9, #1380: your name is one setting, shown where you join. Preferences calls it "Your name" (it signs your comments and is what others see in live sessions). The Join dialog shows it, editable, beside the code, as the Share popover does, and the field is empty while it is the default. A new name is kept at once, and the browser build keeps it in the page's local storage, so a reload joins as you (the other preferences still start afresh there). Changing it while in live sessions renames you in each of them: a `name` signal, which the server cleans and answers with your `PeerJoined` to everyone, and newcomers' welcomes list the new name. A rename made while joining, or before a reconnect, goes out once the host reads signals. An older server ignores it, and clearing the name keeps the session's name until you join again. Tests: `collab/src/rename_tests.rs`, `collab-net/src/rename_tests.rs` (two apps over loopback), `ui-egui/src/display_name_tests.rs`, and `figcraft-web`'s `kept` round-trip. Numbers unchanged |
| 2026-10-29 | minor | A marquee from anywhere in vector edit mode (#1508): a press on another layer used to leave edit mode at once, so a drag there moved that layer; it now draws a marquee that selects the points inside it. A click there still leaves edit mode and selects what it lands on. Numbers unchanged |
| 2026-10-29 | minor | M9, #1381: joining a live session shows its progress. From the click on Join (the Share popover, the home screen or the Join dialog, which an invite link opens) the Join dialog says "Joining live session…" with a spinner and Cancel until the shared file is in its own tab and ready, then the file shows; while it is on its way the join fields don't start a second one. A join that fails or times out says why in the dialog, with Try again (a new join with the same code or link) and Cancel; Escape and the close button cancel too. `collab.leave` takes `session` to end a session by its id, whichever tab is in front. Tests: `live_joining_tests`. Numbers unchanged |
| 2026-10-29 | minor | M9, #1075:`figcraft-server` gives requests that change its files (join codes, links, renames, removals, blocks, publishing with the key) a budget of 60 a minute per address, with the server key and owner tokens alike (100 renames in a row from one address: 100 served before, 60 now; another address is unaffected), and finds join codes through an index of their hashes instead of opening every link's code. Same results, same protocol. Tests: `admin_budget_tests`. Numbers unchanged |
| 2026-10-29 | minor | M8, #1080: SVG import draws a pattern's tile once per pattern and scale and reuses it for every other layer it fills (1,000 rectangles filled with one pattern: 1,000 tile drawings before, 1 now, same image fill), and the layers inside a tile count in the import's 100,000-layer budget instead of each tile starting a fresh one, so patterns nested in patterns stop at the budget. Tests: `tile_tests` (svg). Numbers unchanged |
| 2026-10-29 | patch | M4, #1478: a letter and its combining marks are shaped with one font that has them all. Text typed decomposed (NFD, `E` + U+0302 + U+0309 for Ể) in a font with the letters but not the marks drew missing-glyph boxes beside and above the letter, while the precomposed form fell back to a font that has it; each grapheme cluster now takes the first face (the style's own, then the fallbacks) that covers every char in it (joiners and variation selectors need no glyph), else the letter's own face as before. Fonts that cover Vietnamese were already right: mark positioning (GPOS `mark`/`mkmk`) stacks the marks over the letter on the canvas, in PNG, SVG and PDF. Tests in `crates/text/src/vietnamese_tests.rs`. Numbers unchanged |
| 2026-10-29 | minor | M9, #1415:`comment.list` (CLI, MCP `run_command`, the `figcraft://comments` resource) gives every message, the first and each reply, its `mentions`: `{name, known, you}` for each name it @mentions, by the rule the thread view highlights them with (`figcraft_doc::mentions`, which the highlighting now shares). A name resolves against the file's people and you, as the file spells it. An unknown name is reported with `known: false`. `[email protected]` mentions nobody. Filtering threads by a mentioned name and listing mentioned names in `comment.people` stay open. Numbers unchanged |
| 2026-10-29 | minor | M8, #1061: a mask group's pixmaps (the mask, its own layer, the content and the pixels it goes onto) cover only where the mask or its content can paint (effects included) within the view, rounded out to whole pixels, instead of the whole view: a 40 × 40 mask on a 3000 × 2000 view held 96 MB, now 61.5 kB, drawing the same pixels. Isolated layers (effects, opacity, masks, clipping frames) also share one budget of 2 GiB held at once per render thread (`figcraft_render::layer_budget`); a layer past it is left out with a warning, so 64 nested full-view layers no longer ask for tens of GB. Tests: `mask_bounds_tests`. Numbers unchanged |
| 2026-10-29 | minor | M2: components can't contain their own instances (#1084). Any edit that would make a component hold an instance of itself, directly or through other components (an instance-swap property's current value included), is refused with the loop named: creating, placing, swapping, switching variants, moving, pasting and instance-swap values. Each variant counts as a component of its own, so a variant holding a sibling variant's instance is fine unless that one leads back to it (`components::check_component_cycles`, beside the instance structure check); a dropped instance lands on the page instead. A file holding such a loop (opened or imported) has one instance per loop detached, with a note; a loop merged in a live session is reported, and a rebuild never expands an instance inside a copy of itself, so its copies stop growing (12 rebuilds of a two-component loop: 25 → 113 layers before, 25 throughout now). Tests: `component_cycle_tests`. Numbers unchanged |
| 2026-10-29 | minor | Which copy is which (M8, #995): `edit.duplicate` with `map: true` also returns `map`, each original layer's id, descendants included (nested frames, groups, instances and their sublayers), paired with its copy's id, so a batch can edit inside a copy through a step reference (`"field": "map/6"`) instead of inspecting it. Without `map` the result is unchanged. Regression tests in `crates/engine/src/duplicate_map_tests.rs`. Numbers unchanged |
| 2026-10-29 | minor | Bounded replay log (#1069, from a security audit): the batches a session keeps for apps that reconnect were bounded only by count (10,000 batches, 1,000,000 changes) on a desktop host and a private `figcraft-server`, so a long busy session could hold any number of bytes. `Limits::max_log_bytes` now defaults to 64 MiB, and `figcraft-server` takes it from a new `max_log_mb` setting (default 64; open to the public, also at most a quarter of `max_file_mb`, as before). An app further behind than the trimmed log catches up with the whole file, as it already did past the count limits. Nine 8 MiB renames with default limits kept all 75.5 MB before; now 58.7 MB, and a client that missed them gets a snapshot. The protocol is unchanged. |
| 2026-10-29 | minor | M8, #998: MCP tools that act on a document (`run_command`, `batch`, `create_node`, `set_properties`, `inspect_document`, `get_node`, `select`, `export`, `screenshot`, `save_file`, `undo`, `redo`) take an optional `document`, an open file's `id` (`doc.list` and the `document.inspect` outline now carry it) or path, and so do the control methods `engine.execute`, `document.inspect`, `node.get` and `ui.render`: the call acts on that file and the tab in front stays in front, so an attached agent keeps editing its own file when the user switches tabs. Calls without it act on the file in front as before; other control methods refuse a `document` behind the tab in front. Numbers unchanged |
| 2026-10-29 | minor | Sibling placement in live sessions (#1065, from a security audit): every replica (the server's and each app's) placed a created, moved or reordered layer by counting each sibling that sorts before it, so a batch of many layers under one parent took quadratic time, on the server's single hub thread too. It now binary-searches the siblings' `(key, id)` order on each side of the layer's old place. The order is unchanged, ties and fractional positions included: a test replays 5,000 creations and 2,500 moves against the counting rule. 20,000 creations and 10,000 moves under two parents took 20 s in a debug build; they now take 0.3 s. |
| 2026-10-29 | minor | Comments from an AI agent (M8, #1414): an agent attached over MCP commented, reacted and read threads as the user (the name from Preferences) unless it passed `author` on every call. Every command an MCP client runs now acts as that client's name from `initialize` (`clientInfo.title`, else `.name`) with " (agent)" after it, or "AI agent" without one, carried to the app as `engine.execute {agent}`; its messages, reactions and read state are its own, and its messages are marked (`agent: true` in the file and in `comment.list`, an Agent label by the name in the thread). The user's name and the comments made in the app are unchanged, and in a live session the name and marker travel with the thread to everyone. Tests in `comment_agent_tests`, the MCP `agent_comments` and attached suites, the UI comments suite and the collab diff round trip. Numbers unchanged |
| 2026-10-29 | minor | M3, #1338: the instance-swap property's menu (on an instance, its main component's default and the component playground) and the instance's swap menu open with a search field at the top that has the keyboard; typing filters the components by name (any case, headings without matches drop out), an empty result reads "No matching components", Up and Down step through the matches and Enter picks one (the first without a step), Escape clears the search and then closes the menu. Only the list scrolls; the menu grows back when the search is cleared. Tests: `component_swap_search_tests`. Numbers unchanged |
| 2026-10-29 | minor | A refused batch's fixes in linear time (#1063, from a security audit): the server restores each deleted layer once, walking each subtree once and stopping at layers already restored, and looks up and sends each style, variable, comment or document property fix once. The fixes sent back are unchanged. Deleting every layer of a 20,000-deep chain in a refused batch took 27 s on the hub thread in a debug build (each deleted layer's whole subtree walked again, each entry fix compared with all the earlier ones); it now takes 0.1 s (`a_refused_deep_delete_restores_each_node_once_in_linear_time`). |
| 2026-10-29 | minor | M12, #1387: while a connector is dragged out with the connector tool, the canvas shows the connector itself as it will be made (its straight, elbow or curved route, line, weight and arrowhead, built by the same code as `board.connector`), not a thin line; the object under the pointer is outlined with its four attach points (the middles of its sides), the one the end goes to ringed; within 10 px of one the end keeps that side (`fromSide`/`toSide`, and `side` when re-attaching). Hovering with the tool shows the attach points too. Escape drops the preview; releasing makes the connector in one undo step. Tests: `board_connector_tests`. Numbers unchanged |
| 2026-10-29 | minor | Large images on the canvas (#1483): an image shown smaller than its pixels draws on the canvas from a copy scaled down to 4096 or 8192 px on its longer side, the smallest that holds what the screen shows (`figcraft_render::images::shown`), cached beside the image itself in the same 512 MB budget. Two 16 000 × 10 000 images took 640 MB each and evicted each other on every frame that drew both; as copies they take 80 MB together zoomed out and 336 MB at 8192. Moving one over the other went from about 2.1 s a frame to 11 ms zoomed to fit and 15 ms at 30% in the app (debug build). Closer than that, and for tiles, the canvas draws from the image itself, so 100% and closer stay sharp; a view of both at 100% still takes about 2 s a frame, and an image with shadows still decodes in full for them. Exports, thumbnails and every other render are unchanged. Tests: `large_image_tests`. Numbers unchanged |
| 2026-10-29 | patch | M8: the progressive JPEG check from #1491 (a frame sampling a component more finely than its first, which the JPEG decoder asserts on, and in the browser a panic aborts the page) moved to `figcraft_format::jpeg` and now also guards the pictures files carry (`thumbnail::decode_thumbnail`: the home screen and recent files), so a crafted `.figcraft`, `.fig` or `.sketch` thumbnail or image file shows no picture instead of crashing. The canvas, exports and SVG/PDF keep it through the renderer. Regression tests in `jpeg_tests` (format) and `jpeg_sampling_tests` (render). Numbers unchanged |
| 2026-10-29 | minor | M6, #1234:`.fig` import resolves style links that name a library style by its asset key and version instead of a guid (the file's own copy of the style), as variable links already were; text in instances of library components kept a stale copy of the style's values and no link. The linked text style beats the copies of its values on the layer and on the override rows below the row that links it, and rows above that row (an outer instance's size, say) still win. Override rows that restate a nested level's overrides (`overrideLevel`) apply at that level, below the instance's own rows, instead of in wire order. Synthetic regression tests in `text_style_tests`. Numbers unchanged |
| 2026-10-29 | minor | M6, #1236 (gap 1): imported `.fig` text keeps the line breaks the file saved with it (`derivedTextData`), so a font whose advances differ slightly from the file's (another build of the family, a stand-in) no longer wraps a word early or splits one in a narrow box. The breaks hold only while the box width and everything that moves a break are as saved (characters, fonts, sizes, spacing, case, axes, features, lists, paragraph settings; paints, links and vertical settings don't count), and are given up where a line would come out more than 3% (+1 px) wider than its box; they are kept in `.figcraft`. The line breaker was checked and is unchanged: trailing spaces and the letter spacing after a line's last glyph already stay out of the fit test, box widths aren't rounded and the tolerance is 1e-6 px, so the differences come from the fonts. Synthetic tests in `saved_breaks_tests` and `fig::text_wrap_tests`; not checked on real files. Numbers unchanged |
| 2026-10-29 | minor | #754, #1025: the Missing fonts window offers Download for a family the suite's font store has (Google Fonts, Noto: craft-libs' `craft-font-store`, its pinned indexes compiled in). The first Download asks the user to allow downloading fonts (off until then; the setting is the Craft apps' shared one); the files download on a thread of their own, each checked against the index's size and sha256 before it is installed into the shared font folder, which the font scan now reads; the row shows progress or why it failed, and once installed the file's text fits the real font and the family leaves the list. `font.download {family}` and `font.downloads` are commands, but no command turns downloading on; agents install only while the user has it on. Desktop only. No cancel, no turning it off from FigCraft, no substitute picker. Tests: `fonts::download::tests`, `missing_fonts_tests`. Numbers unchanged |
| 2026-10-29 | minor | Web wasm size (#1472): the release wasm is 23,915,213 bytes (22.8 MiB), down from 25,982,540 (24.8 MiB), 1.2 MiB under the gate. Catalog rows hold their plural forms behind a reference (two words, not eight) and the browser build compiles with wasm SIMD, so the glyph rasterizer's scalar fallback is left out; the 24 MiB gate in `packaging/web/package.sh` is restored. The browser now needs WebAssembly SIMD (Safari 16.4 or later). Numbers unchanged. |
| 2026-10-29 | minor | Shadows of images with transparent pixels (M3, #1289): a drop shadow is cast from the picture's alpha and hidden only behind its pixels, not its whole box (a transparent PNG's shadow was cut away inside its bounding box), unless shown behind transparent areas; its inner shadow shades the inside of its alpha, not the box's edge; the same for Fit images, crops smaller than the box and pattern fills. Opaque pictures that fill their box keep their outline (spread still offsets it exactly); each image is scanned for transparency once. Canvas, PNG, SVG (knocked out by the layer's alpha; an inner shadow embeds a render) and PDF (`image_shadow_tests` in render and svg). Numbers unchanged |
| 2026-10-29 | minor | M1, #1464: Crop image. Object ▸ Crop image, the right-click menu (one image layer) and a crop button on the image fill row run `image.crop`: the layer's top visible image fill (or the one `index` names) switches to Crop with the picture left where Fill or Fit drew it, as one undo step, and the canvas enters crop mode (the image shows dimmed outside the layer, whose box gets corner brackets and edge bars; the layer's edges and corners crop it with the picture held still, the image drags to move and its corners scale it; Enter, Escape or a click outside finishes). A layer without an image fill, or several layers, are refused with a reason. Tests: `image_crop_tests`, `crop_image_tests`. Numbers unchanged (the menu item is FigCraft's own and not counted by `parity`) |
| 2026-10-29 | minor | The Design panel's sections on one grid (M4, #858): 16 pt left and 8 pt right margins, two 88 pt columns with 8 pt gutters and a 24 pt button column, 24 pt rows under 10 pt captions, 4 pt between a row's two buttons; section titles stay 11 pt semibold; the panel's edge is drawn inside it, so its right margin no longer loses 1 pt to a frame stroke. Numbers unchanged |
| 2026-10-29 | minor | The toolbar on one 32 pt grid (M4, #1210): the pill's cells, padding, gaps and divider share one grid in every state, and the Resources button is gone, as in UI3; the Assets panel opens from the left rail, Shift+I, Alt+2 and View ▸ Panels ▸ Assets. Numbers unchanged |
| 2026-10-28 | minor | Arabic draft: 1,857 UI strings, 620 command labels, 31 menu wordings; count-neutral plurals and web font gate. Native review and panel mirroring remain open; grade and estimates unchanged. Evidence: `crates/ui-egui/locales/ar.toml`, i18n catalog tests and [translation notes](docs/translating.md#arabic-draft) |
| 2026-10-28 | minor | M3, #922: instances scaled with the Scale tool keep their layers' scaled size and points when their main is rebuilt (`size`/`position` overrides recorded on scale). Numbers unchanged. |
| 2026-10-28 | minor | M6, #1059:`.fig` import copies embedded image blobs once per blob, not once per hash naming them. Numbers unchanged. |
| 2026-10-28 | minor | M1, #1476: page ruler guides survive save and reopen; files written with a page's `guides` key twice open again with both lists. Numbers unchanged. |
| 2026-10-27 | minor | M4: zero-length round-capped dashes draw dots; dash expansion uses the existing safety budget; extracted from #1143. Numbers unchanged. |
| 2026-10-27 | minor | M6, #1341: FigJam sticky notes and shapes with text import their generated layers (#1397): the background (its saved vector network, or the outer fill or stroke outline when the generated geometry is damaged, before a rectangle), text with its typography and paints, saved bounds and sticky author lines, as editable frame children; missing or invalid generated text falls back to the outer text within the layer budget. Synthetic regressions and a local check of the report sample; native board metadata and direct `.jam` opening remain outside this fix. Numbers unchanged. |
| 2026-10-27 | minor | Placed images keep their own size (M7, #1360): dropping, Place image, paste and `image.place` put an image at its pixel size and fit only a side past 4096 px (they fitted it into 1000 px); a PNG whose `pHYs` records a whole 2x–4x multiple of 72 dpi (macOS screenshots record 144 dpi) is placed at its point size, so an 800 × 600 px capture comes in at 400 × 300; other densities, such as 300 dpi scans, keep their pixel size; the image data stays at full resolution and an explicit width or height still wins. On macOS, paste reads the clipboard's PNG as is (its density and colour profile kept) and falls back to the decoded image. Numbers unchanged |
| 2026-10-26 | minor | Non-ASCII layer names transliterate in code export (M6, #1235): `class_name` transliterates through the new `deunicode` dependency (pure Rust tables, zero deps, BSD-3-Clause), so Korean, Cyrillic, Latin-diacritic and CJK names become readable ASCII classes (`rogeuin-hwamyeon`, `knopka`, `uber`, `an-niu`) in HTML/CSS/React/Leptos output, component roots and `data-*` variant attributes; symbols-only names take the node's kind (`.text`) instead of `layer-`, and the 40-character cut lands on a word boundary. `html_codegen_tests` updated and extended. Numbers unchanged |
| 2026-10-25 | minor | M6, #1143: offline Penpot 2.x open/import/export through the editor, CLI and MCP using the existing document model. Unsupported layout settings retain saved positions and bounded source metadata with warnings; font binaries and toggle-overlay are reported, and provider fonts are never downloaded. Shared ZIP/JSON/token budgets guard untrusted input. Figma parity numbers unchanged. |
| 2026-10-25 | minor | egui 0.37 capture lifecycle: screenshots with hidden or unknown visibility queue, request a hidden paint, and expire after 60 seconds if no capture arrives (headless-tested contract). Logic-only edits, saves and inspection continue while hidden. Native visible GUI and forced hidden-window paint, browser and Wayland behavior remain unqualified; parity percentages are unchanged. |
| 2026-10-24 | minor | Server memory (#1134): when the last person leaves a file nobody edited, figcraft-server keeps the client ids it issued in the log instead of rewriting the snapshot (compaction still follows `compact_after`/`compact_mb` and edits), and a snapshot is encoded straight into its file buffer, without a second copy of the body; the file format is unchanged. Numbers unchanged |
| 2026-10-24 | minor | Boards (M12): a dot grid on board pages under the layers, connector ends dragged onto other objects or empty canvas (`board.reattach`), curved connectors, elbows that go around what they join. Boards row ~35% → ~37%, ~20% → ~21%, 30–60 h → 28–56 h; headline numbers unchanged (boards are outside the weighted sum) |
| 2026-10-24 | minor | Installed weights of an embedded font family (M7, #1246): release builds embed some families with only a few styles (Poppins Regular, Italic and Bold), and the first font scan dropped every installed face of such a family, so an installed Poppins SemiBold or Medium was never offered (a rescan did offer them). The first scan now keeps the embedded faces, which still win for the styles they cover, and offers the installed styles they lack beside them. Regression test in the font database tests. Numbers unchanged |
| 2026-10-24 | minor | Dev Mode tail (M6, gap 16): Measurements kept on the canvas (gap 16): `devMode.addMeasurement {from?, to?, side?}` places one between two layers, along one side of a layer or (neither) to the layer around it, stored on the layer it starts from (saved in `.figcraft`, synced, undoable; old files open without them); Dev Mode draws them as labelled redlines that follow the layers, the Inspect tab lists them (those of the selected layer, or every one on the page), places and removes them; `devMode.removeMeasurement` and `devMode.measurements` remove and list them; a copy keeps those to layers copied with it. The layers panel is read-only in Dev Mode: rows select and expand, but don't rename, reorder, hide or lock. Dev Mode 97% → 98% breadth, 76% → 78% ready, 2–4 h → 1–2 h; headline numbers unchanged at this rounding |
| 2026-10-24 | minor | Board facilitation (M12): a shared timer and voting sessions (`board.timer`, `board.vote` and their queries; timer and vote sync as document state, ballots per voter; neither is an undo step). Boards row ~35%/~20% → ~40%/~24%, 30–60 h → 26–54 h; headline numbers unchanged (boards are outside the weighted sum) |
| 2026-10-24 | minor | Board templates and tables (M12): five built-in templates (`board.template`, `board.templates`, a toolbar picker with pictures) and tables (`board.table`, `board.tableEdit`; rows and columns from the bar over the selection). Boards row ~35%/~20% → ~40%/~24%, 30–60 h → 26–54 h; headline numbers unchanged (boards are outside the weighted sum) |
| 2026-10-24 | minor | Plugin widgets take text input and a property menu (gap 6, plugin API 4; API 1–3 plugins and existing widgets unchanged): a widget's text layer can be an editable field, edited on the canvas with a double-click like any text (or `widget.setText`), its value cleaned and capped at 1,000 characters, written into the widget's state under the field's name and the field's text, and the plugin called back to redraw, all as one undo step that syncs in live sessions; a widget can carry a property menu of up to 8 actions, toggles and colour choices (up to 8 colours each), shown in a small bar over the widget while it is selected and driven with `widget.menu`; descriptions, values and menus read back from files are checked, an API 3 plugin that draws a field or a menu is refused, and older FigCraft builds refuse API 4 manifests. The Counter example gained a label field and a menu (reset, steps of ten, a colour); the SDK reads the new events. Hover and timers are still missing. Plugins & widgets 60%/35% → 65%/38%, 8–26 h → 6–22 h; Ecosystem dimension 38% → 40%; weighted totals unchanged at this rounding; full parity ~262–519 h → ~260–515 h |
| 2026-10-24 | minor | Branches in live sessions (gap 17): create, edit, review, update from main and merge while the file is in a session; the branch stays with its author (ids in their own client space, far above the main file's), merging is one synced edit and one undo step, peers' edits merge against the branch's own base and conflicts resolve as offline, a view link refuses merging, a branch made outside the session that adds layers is refused with a reason; a branch's history gets entries for its creation and updates from main; no wire format change. Left: branches others in the session can see, text runs item by item. Numbers unchanged |
| 2026-10-24 | minor | Boards (M12): a file kind of its own (`file.newBoard`, saved in `.figcraft`, carried to live-session joiners) with sticky notes, shapes with text, connectors that re-route when what they join moves (an engine recompute, the same for local edits, undo and others' edits), marker, highlighter and eraser, stamps with the placer's name, a board toolbar and a contextual bar; prototyping and Dev Mode refuse on boards; design files and boards have their own marks on tabs and file rows. Boards row added outside the weighted numbers; headline numbers unchanged at this rounding |
| 2026-10-21 | minor | Best-effort `.sketch` import (Sketch 48 and newer): open, import, drop, CLI and MCP; symbols become live components and instances, shared styles and Color Variables come across, Sketch 2025 frames, stacks (as auto layout) and tints too, text placed as Sketch places it; checked on three real files against Sketch's own renders ([file-format-parity](docs/file-format-parity.md)). Headline numbers unchanged |
| 2026-10-22 | minor | CJK line breaking: JIS X 4051 line-start/line-end sets completed (halfwidth small kana, halfwidth and spacing voiced sound marks, iteration marks 〱–〵, halfwidth parentheses) with behaviour cases; numbers unchanged |
| 2026-10-18 | minor | Colour picker models LCH and OKLCH beside Hex, RGB, HSL, HSB and CSS (desktop and web): lightness, chroma and hue fields (OKLCH lightness in percent, chroma to two places), an edit keeps lightness and hue and lowers chroma until the colour fits sRGB, as CSS gamut mapping does; the CSS field reads `lch()` and `oklch()` too. Values read the stored channels as sRGB, like the other models (a Display P3 file's colours aren't converted first). Conversions checked against the CSS Color 4 and OKLab reference values; checked in the desktop app over the control channel. Numbers unchanged |
| 2026-10-15 | minor | Auto-save (#305, #296): Preferences ▸ Save files automatically (`pref.autoSave`, off by default; desktop only, the browser has no files to save to) saves each open file that has a path 3 s after its last change, so steady editing doesn't save every few seconds, and never during a gesture, a live text preview or while a version is shown. The file is encoded on a thread of its own and written as of that moment (edits made meanwhile stay unsaved); a failed save says why in a toast and is tried again a minute later, and version history checkpoints keep coming with it on. Numbers unchanged |
| 2026-10-15 | minor | A text box by dragging (#363, #360): a drag with the Text tool makes a fixed-size text box the size of the drag (at least one line tall); a click, or a drag shorter than the drag threshold, still makes auto-width text. Numbers unchanged |
| 2026-10-15 | minor | Batch step references (#400): inside `command.batch` (MCP `batch`), `{"$step": n}` anywhere in a step's params stands for the `id` step n returned (`{"$step": n, "field": …}` for another field; `$step` is a reserved key), so a frame with its children takes one batch and one undo step; a malformed reference fails the batch, which rolls back, and a nested batch keeps references to its own steps. Numbers unchanged |
| 2026-10-15 | minor | Removing someone from a live session (#338): right-click a person's avatar ▸ Remove *name* from session, or `collab.removePerson {client}`, when you host the session or own the file on a server; the session ends for them (they keep their copy), the code or link they hold no longer gets them back in while it runs, and everyone else stays. Numbers unchanged |
| 2026-10-15 | minor | Pages list (#272, #258): a long pages list scrolls in its own area, so the layers below stay in view; page and layer drops count only inside the list's visible rows. Numbers unchanged |
| 2026-10-15 | minor | Variables export in the browser (#368): Export… ▸ Variables JSON / CSS custom properties did nothing in the web build and now arrives as a download; importing there still needs a file picker. Numbers unchanged |
| 2026-10-15 | minor | Font picker keys (#352, #345): Up and Down step through the font families from the current one, keeping the highlight in view and previewing it on the canvas; Enter applies it, Escape leaves the text as it was. Numbers unchanged |
| 2026-10-15 | minor | Leaving Design ends the pen path (#308, #307): switching to Prototype or Inspect while the Pen draws ends the path and goes back to Move (pressing Design again keeps drawing), and a click on the toolbar no longer reaches the canvas under it. Numbers unchanged |
| 2026-10-15 | minor | Vector edit mode keeps the canvas still (#322): the edit mode hints in the Design panel wrap, so entering edit mode and picking a region no longer shift the canvas by about 25 points in the middle of the click (which dragged the region picked). Numbers unchanged |
| 2026-10-15 | minor | Drop shadows under a background blur (#317, #255): a layer with both casts its shadows from what it paints (a translucent fill casts a lighter shadow than an opaque one) and reads the backdrop only after drawing them, so a shadow shown behind transparent areas is seen through the blur. Numbers unchanged |
| 2026-10-15 | minor | When the window can't open (#306, #261): Windows release builds have no console, so a graphics failure at start quit without a word; the error is now logged and shown in a message box in the interface language, with what to try (unset `WGPU_BACKEND`, update the driver, OpenGL) and where to report it, skipped with `--control`. Numbers unchanged |
| 2026-10-15 | minor | Crashes on Windows (#271, #314, #337; #224 and two more): a text field focused before it was drawn aborted the screen-reader layer; double-clicking a page, creating a variable collection and renaming a style folder no longer close the app. Numbers unchanged |
| 2026-10-15 | minor | Deeply nested layers (#330, #210): one depth guard (256) on the three recursive layout paths, so a deep file can't run the layout out of stack. Numbers unchanged |
| 2026-10-15 | minor | Image memory budgets (#326; #203, #189): decoded images and a file's images have budgets; the decode cache evicts its least recently used entries. Numbers unchanged |
| 2026-10-15 | minor | macOS file dialogs (#253, #250) no longer re-enter the event loop; a dialog's result goes to the file and window it was opened for, and dialogs are parented to the main window. Numbers unchanged |
| 2026-10-15 | minor | Recovery copies (#297, #303, #331, #336; #202, #185, #187): a copy counts only after its write succeeds and a failed one is retried; saves and closes wait for the write in flight; copies keep version history and branches; a recovered copy stays until this run has its own. Numbers unchanged |
| 2026-10-15 | minor | Text edits (#280, #278) no longer roll back earlier changes; a drag released off the canvas ends at the pointer's last canvas position. Numbers unchanged |
| 2026-10-15 | minor | Rollback (#282, #301; #196): a failed command rolls back during a drag or other gesture too; a failed batch keeps the redo steps. Numbers unchanged |
| 2026-10-15 | minor | Atomic save and export (#284): staging files are uniquely named and created exclusively, so a file or link already at the old fixed staging name is never written through, and two writers of one file no longer share one. Numbers unchanged |
| 2026-10-15 | minor | Alt-drag copies (#283) snap to their original, and Duplicate (Cmd+D) then repeats the last copy's offset. Numbers unchanged |
| 2026-10-15 | minor | The pen (#277) shows its Bezier handles while a point is dragged out. Numbers unchanged |
| 2026-10-15 | minor | Corner handles (#293, #292): a press on a corner handle resizes instead of picking the corner-radius handle. Numbers unchanged |
| 2026-10-15 | minor | New text (#372, #362) takes a readable colour from the frame it is drawn in, not only the page (text drawn in a white frame on a dark page was white). Numbers unchanged |
| 2026-10-15 | minor | Outline stroke (#386) works on every selected layer. Numbers unchanged |
| 2026-10-15 | minor | Send to back (#444) counts per parent when the selection spans frames. Numbers unchanged |
| 2026-10-15 | minor | Align buttons (#276) are off when the selection has nothing to align to. Numbers unchanged |
| 2026-10-15 | minor | Gradient type icons (#275) show their gradient. Numbers unchanged |
| 2026-10-15 | minor | Corner resize cursors (#388) stay diagonal at any proportions; the text selection highlight (#390) stays inside the box on the last line. Numbers unchanged |
| 2026-10-15 | minor | New pages (#311, #295) take the current page's background; the theme still decides a new file's page. Numbers unchanged |
| 2026-10-15 | minor | A slice's export preview (#265) shows the content under it. Numbers unchanged |
| 2026-10-15 | minor | SVG export (#327, #325) keeps a nested frame's fill and stroke inside its parent's clip. Numbers unchanged |
| 2026-10-15 | minor | Home screen (#318, #324): recent files notice files deleted while it is open; a long path no longer draws over the file name. Numbers unchanged |
| 2026-10-15 | minor | The Small caps sample in Type settings (#355) draws from capitals, no longer a box where the UI font lacks the small-capital letter. Numbers unchanged |
| 2026-10-15 | minor | Web clipboard (#375): copy and paste in the browser go through the system clipboard, so a paste after a copy brings the copied layers instead of whatever text the clipboard held. Numbers unchanged |
| 2026-10-15 | minor | Zoom previews (#312): when renders from scratch are slow, a new zoom first shows the whole view coarsely instead of the page colour, then the refined image; images of another file or page never show. Numbers unchanged |
| 2026-10-15 | minor | Hidden windows (#313, #332; #269): a minimised or covered window answers control and MCP requests (commands, inspection, `ui.render`) at once, each window on its own; screenshots and input still wait for it to be on screen. Numbers unchanged |
| 2026-10-15 | minor | Server (#340, #302): a compaction cut short closes the file instead of confirming lost edits (#183); each team library version is kept in its own file (#206). Numbers unchanged |
| 2026-10-15 | minor | Counters (#343, #344, #353; #309): collab sequence numbers, layer name counters and opacity digits can't overflow. Numbers unchanged |
| 2026-10-15 | minor | Commands refuse bad indexes and ids (#475, #384, #310): `page.select` and `doc.switch`, `effect.set` and `export.update` / `export.remove`, `page.move` with a non-page id. Numbers unchanged |
| 2026-10-15 | minor | Find and replace (#358, #361) changes every occurrence, case-insensitively, also when the replacement contains the search text. Numbers unchanged |
| 2026-10-15 | minor | The gradient shorthand (#254, #252) is read by the engine, so the CLI, control channel and MCP share it. Numbers unchanged |
| 2026-10-15 | minor | .fig import (#279, #281): resized instances lay out their layers by their constraints (#266); explicit variable modes whose collection is named by its asset key, as library collections are, come across (#267). Numbers unchanged |
| 2026-10-15 | minor | Colour emoji (#300, #268) draw from their own colour tables (sbix and CBDT/EBDT bitmap strikes, COLR v0 layers), a colour face tried first; COLR v1 still draws outlines. Numbers unchanged |
| 2026-10-15 | minor | macOS (#290, #289): `.figcraft` files open from Finder, Open With and the Dock. Numbers unchanged |
| 2026-10-15 | minor | Windows (#294, #291): the first preferred language alone picks the UI language. Numbers unchanged |
| 2026-10-15 | minor | Unix font discovery (#474) follows `XDG_DATA_HOME`. Numbers unchanged |
| 2026-10-15 | minor | Linux packaging (#346, #328): the window id matches the desktop entry (FreeBSD keeps its class); the RPM requires glibc 2.35. Numbers unchanged |
| 2026-10-15 | minor | Tests (#472, #471, #339): the rename shortcut test, the hanging punctuation test and the font-scanning tests no longer depend on the platform, font or test order. Numbers unchanged |
| 2026-10-15 | minor | Languages (#257, #263, #469, #477, #288): Russian (with one/few/many plural forms) and Turkish catalogs, reviewed by native speakers; Czech and Indonesian drafts; the Japanese draft completed. 2,177 entries now (1,587 UI strings, 559 command labels, 31 menu wordings); strings added since a catalog was written read English there. Localization 39% and 29–51 h unchanged at this rounding (Indonesian is now a draft, while the five older drafts lack 88 newer entries) |
| 2026-10-22 | minor | Type settings keeps its width (#1029): in languages whose labels run past the 110 px label column (French "Retrait de paragraphe", "Espacement de liste"), the List spacing, Paragraph indent, Max lines and decoration rows were wider than the popover, which then widened every frame while open. Those fields now take the width left after their label; English is unchanged. Regression test in `type_settings_tests`. Numbers unchanged. |
| 2026-10-22 | minor | `fill`/`stroke` props accept `"none"` (either case) and `null` as no paint (M1, #1036): mapped to an empty fills/strokes list, as `fills: []` already did (undocumented), so an outline-only shape is one call; `KEYS` and `docs/mcp.md` document it, the error names the accepted forms, `fill_none_tests` covers create, set, undo, text and the error. Numbers unchanged |
| 2026-10-20 | minor | A document dropped on the canvas opens (#916): a `.figcraft`, `.fig` or file JSON dropped on the canvas opens as a file of its own, as on the home screen, in the browser too; it used to add its pages to the file in front. Images and SVG still import at the drop point, File ▸ Import still adds pages, and dragging only documents over the canvas says "Drop to open" (not in the browser, which names no file until the drop). Numbers unchanged |
| 2026-10-20 | minor | Duplicating a main component makes an instance (#892): Duplicate and Alt-drag copied a main component as a second main; within its file the copy is now an instance of it, as Copy then Paste already was (a variant duplicated in its set stays a variant). Numbers unchanged. |
| 2026-10-20 | minor | Readable disabled and selected text (#890): the dim text colour (disabled buttons, placeholders, hints) is at least 3:1 against panels, fields and selected rows in both themes (it was 2.1:1 in the light theme), a disabled button is grey with that text instead of fading to 40%, and selected text (renaming a page or layer, any text box) takes the theme's readable green on the highlight (1.4:1 in the dark theme before, now 4.2:1 or more); checked with the colour contrast helper (`theme::tests`). Numbers unchanged |
| 2026-10-20 | minor | Layer data reads back (#879): `node.get` (the command, the control method and MCP `get_node`) and `document.inspect` report a layer's or page's `pluginData` as `{key: value}` when it has any, so a tool that keeps state in a file with `node.set {props: {pluginData}}` can read it on its next run (`plugin_data_tests`). Numbers unchanged |
| 2026-10-20 | minor | Measuring while nudging (#894): with Alt held over another layer, the arrows nudge the selection (Shift: the big step) and the distances follow every nudge (Alt+arrows used to do nothing); while measuring, the selection's handles and size label hide and both layers keep their outlines (`measure_nudge_tests`). Numbers unchanged |
| 2026-10-20 | minor | Release fonts (#885): the web release embeds craft-fonts again (Inter), so the browser version draws text and lists fonts; it had none since releases moved to the shared workflow. The craft-fonts pin goes back to the commit with the Latin manifest (the Georgian change had moved it to one without Inter, which would have failed the next release). Numbers unchanged. |
| 2026-10-17 | minor | Community batch #753 (10 PRs): Selection colors for mixed selections (#403); smart-selection spacing fields and canvas handles (#402); a Present tab with device and scale controls (#537); adding auto layout to an empty frame keeps its size (#369); `node.set` applies auto layout keys to frames only and warns about font styles a family lacks (#603); canvas images on whole device pixels (#359); Adwaita window controls on Wayland without server decorations (#592, not run on GNOME or KDE by the maintainers); the keyboard shortcuts panel closes on the home screen (#548); macOS paste modifiers in tests (#750); credits column tests (#347). `cargo xtask parity` re-run: 212 / 212 menu items (100%; the catalog gained View ▸ Full screen with #227 on 2026-10-14), 561 commands. Language table re-measured from the catalogs (2,272 entries): the drafts read 95–99% (strings added since 2026-10-12 show English), Indonesian has a draft (99%; it read none), and Russian, Swedish (its command labels from #599), Turkish and Czech are listed with their counts. Per-area parity rows for the day's landings in `docs/gaps.md`, `target-app-parity.md` and the area docs. Numbers unchanged |
| 2026-10-17 | minor | Control channel while the window is minimised (M8, #720): a minimised or covered window answers every request that acts on the app's state (edits, `ui.inspect`, `ui.pointer`, `ui.key`, `ui.type`, `ui.set`, `app.quit`…) without drawing, and what needs a drawn frame (`ui.screenshot`, `ui.input`, `window.new`) is an error at once; before, those waited up to 60 s and held up every later request on the connection (an attached agent's `inspect_ui` or window screenshot stalled until the window came back). Measured on Windows 11, minimised: requests ~100 ms each (eframe wakes a minimised window at most ten times a second; ~2 ms on screen), a window screenshot fails in ~100 ms instead of 60 s, idle CPU 0. Numbers unchanged. |
| 2026-10-17 | minor | Community fixes from #524 up (batches #727, #728, #733, #735, #736 and #729; 55 PRs) and a crash guard. `.fig` import: explicit instance identity, overrides by key, text style precedence, grid track sizes, slot content, connector labels, skip ink, paragraph options and hanging lists, list start numbers, vertical trim, per-point stroke joins, text on a path, image adjustments and embedded images; large files open within budgets from the file's own size; an import report after opening; imported bytes stay out of the command journal. Files and paste: save a copy as `.fig`; Cmd/Ctrl+V pastes images copied in other apps and layers copied from another design app (`edit.pasteDesignLayers`); paste over and paste properties; clipboard reads on Linux and the BSDs time out after 2 s. Editor: live font preview on hover; Shift+I and Resources open Assets; Shift-drag moves the selection; thin strokes take a nearby double-click; dropped points merge; Flatten keeps open paths; the Actions palette gets tabs, ranked search, a keyboard highlight and recents; the colour picker opens on a new fill or stroke; text strokes follow their alignment; selection highlights after line breaks; a draggable pages/layers divider; variant properties reorder by drag; submenus keep their hover with a safe triangle; menu actions that need a value ask for it; File ▸ Rename edits the name in place. Engine: bad paint indexes, non-finite geometry and pages in copy/cut are refused; a param naming a layer says so when it isn't one; a CLI script is one undo step and resolves `$step`. Server and collaboration: message limits follow `max_file_mb` (client capped at 1 GiB); a damaged generation field no longer drops flushed edits; the web build joins a session from a `#join=` link. Also Swedish command labels, Nix CI through the shared workflow, paint bounds once per render, and a guard that turns a panic inside the boolean-op library into bad input (the stroke draws unclipped). Numbers unchanged. |
| 2026-10-17 | minor | Number variables on text content (M3, #629): the Typography section's variable picker also lists number variables for a text layer's content, as the engine already accepted (the text shows the number, e.g. 42.5, and follows the variable). Other text fields keep their own types. Numbers unchanged. |
| 2026-10-17 | minor | Frames and sections convert into each other (M1, #719): `object.convertToSection` turns a frame on the page or in a section into a section in place (same box, name, fill and layers, which stay where they are; what only a frame has goes: auto layout, clipping, layout grids, guides, scrolling), and `object.convertToFrame` now also turns a section into a frame that keeps its fill and doesn't clip, so the look stays. Nested frames, instances, main components and sections holding sections are refused with the reason; one undo step each. Both are in the canvas context menu (Convert to section / Convert to frame) and in Spanish. Numbers unchanged. |
| 2026-10-16 | minor | Flutter code (ours; the incumbent generates CSS, iOS and Android only): `inspect.code`, `inspect.componentCode` and `codeConnect.set` take `language: flutter` (`dart` is an alias): a `StatelessWidget` for a component (properties as constructor parameters with their defaults, instances as calls, variants as a Dart enum, swaps a `Widget`, slots a `List<Widget>?`) and a Dart widget tree for a layer (`Container` with a `BoxDecoration`, `Row`/`Column` with `spacing`, `Expanded`, `Stack` + `Positioned`, `Text` with a `TextStyle`, `Image.asset`, `Opacity`), plus Dart `import '…';` lines in Code Connect. The Dev Mode language list includes it. Numbers unchanged |
| 2026-10-16 | minor | Adobe Fonts found on macOS (M7, #298): fonts activated through Adobe Fonts rendered in the fallback font without a warning, because Creative Cloud keeps them in its own folder as hidden files named by number, which the font scan never read. The macOS scan now includes `~/Library/Application Support/Adobe/CoreSync/plugins/livetype`, and files without an extension are indexed when they start with a font signature. Fonts are read in place. Regression test in the font database tests; docs/cli.md lists the folder. Numbers unchanged |
| 2026-10-16 | minor | Community fixes from #478 up (batches #591, #598, #602; 69 PRs) and a control-channel fix: a line that isn't JSON now ends a `--control` connection, so a web page can't drive the app through a browser request (#589). Highlights: page copies keep navigation, flows, prototype settings and guides; Find and Replace keeps line endings and case-insensitive matches; text hyperlinks survive import and export; recent colours record only what was picked; MCP screenshots respect `maxSize`; Swedish catalog (UI strings), Czech plurals; bounded canvas overscan while panning; Korean fallback at the run's weight; many vector-edit, guide, page and .fig round-trip fixes. Numbers unchanged |
| 2026-10-15 | minor | Drop files anywhere to open them (M7): dragging files over the window shows a "Drop to open" (home) or "Drop to import" (canvas) overlay. On the home screen each supported file opens as a document of its own and other files are turned away with a toast; on the canvas the files import at the drop point. Both wait behind a small card that is on screen before a large file blocks the window. A canvas import stays with the document and page that received the drop even if the tab or page changes meanwhile, and is cancelled with a message if that document or page is gone. Numbers unchanged. |
| 2026-10-15 | minor | M7: system font discovery skips entries observed as non-regular files before opening them, avoiding the reproduced named-pipe stall. Regular-file symlinks remain supported; directory depth, font limits and priorities are unchanged. Numbers unchanged. |
| 2026-10-15 | minor | Canvas retains a bounded margin while panning so newly exposed strips can reuse the previous render. Numbers unchanged. |
| 2026-10-15 | minor | Dev Mode tail (M6, gap 16): import lines for connected components inside generated layer trees (definitions, slots and swapped-in instances too); `dev.compare {version}` and the compare view's Compare with menu compare with any saved version; behaviour suite (5 tests; 4 failed before the fixes): an inspect-only Dev Mode canvas with menus, shortcuts, nudges, opacity keys and paste refusing design edits, toolbar drawing tools returning to the design, Changed in the ready list (`inspect.readyForDev {status: changed}`), statuses only on sections and top-level frames; read-only plugins' panels from the Inspect tab. Dev Mode 96%/72% → 97%/76%, 0–1 h → 2–4 h (measurements kept on the canvas and a read-only layers panel counted as missing); header totals not recomputed |
| 2026-10-15 | minor | Typography spacing icons scrub with the shared number-field gesture, modifier rates and one Undo step per drag (#373). Units and selected-character targeting are preserved. Numbers unchanged. |
| 2026-10-14 | minor | System theme on first run (M7): preferences without a theme default to System; the first frame reads the OS appearance directly, including when returning from a manual theme. Saved Light/Dark choices stay explicit; unavailable OS appearance falls back to Light. Numbers unchanged. |
| 2026-10-14 | minor | Home screen recent files as cards (M7): each card shows the picture its file carries on its first page's colour, the file's icon, name and "Edited ... ago"; sort by name, creation or last change either way; a table view; a click selects, a double-click or Enter opens, arrow keys move; Open, Show in folder and Remove from recents on right-click. Numbers unchanged. |
"{out} wasn't written: {images_left_out} of its images weren't loaded (together they pass what FigCraft reads from one file), and writing over it would delete them. Write to another path; {input} keeps them."
json!({"id": id,"ok": false,"error": "timed out waiting for the UI: screenshots and input need the FigCraft window on screen (is it minimised or covered?)"})
}
Err(WaitFailure::Disconnected)=>{
json!({"id": id,"ok": false,"error": "the FigCraft window closed before answering"})
Loaded 100 of 1105 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
bflatastic