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).
* 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.
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/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]>
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]>
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]>
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]>
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.
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.
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.
- 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.
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).
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]>
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]>
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]>
* 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.
From review:
- The check ran after any file.open/openRecent/import result, so a
pasted or dropped image, reopening a file that was already open, or
Clear recent brought the window back. It now runs only for results
with a `summary` (a file opened, or pages imported), and never in the
browser.
- Text > Missing fonts... before the startup scan finished listed
installed fonts as missing. It now says it is checking the installed
fonts and lists them once that is done
(`figcraft_text::installed_fonts_known`, also true when system fonts
are opted out, where checks used to wait forever).
- ROADMAP progress log and file-format-parity status line.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Text in a family that isn't installed was drawn in a fallback font
without a word. Now:
- `text.missingFonts` lists the families a document names that no
installed or built-in font provides: text layers and their styled
ranges, text styles, and string variables bound to a font family
(every mode, through aliases), with the number of text layers.
- The desktop app shows a Missing fonts window after the user opens a
file that has any: once installed fonts are indexed (before that every
one of them would count as missing) and after any import report is
closed. Files named on the command line count. Programmatic opens and
the browser build never open it by themselves. Text > Missing fonts...
shows it for the file in front; Escape closes it.
- `figcraft-cli info` prints `missing font` lines (`missingFonts` with
--json).
- file-format-parity: the row claimed REST import warnings that no code
produced.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
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.
* fix(ui): Rotation gets its own icon, not Corner radius's (#256)
The Rotation field showed the corner radius icon, so the two fields
looked alike. Rotation now shows an angle: two rays from a vertex with
the arc between them, drawn in code like the rest of the icon set.
Refs #256 ("The icon for rotating objects is the same as for rounding
corners").
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
* fix(ui): the angle icon is drawn unhinted and centred; its test sits with the icons (#256)
From review: the rays and the arc are both unhinted, so the arc meets
the rays at the 14 px field size; the glyph moves up half a unit to
centre it like the other leads; the test moves from a11y_tests.rs to
the icon tests.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
* fix(ui): SVG and PDF export rows offer no size (#256)
The size picker (1x, 2x, 512w…) showed on every export row, but SVG
and PDF are vector and their export ignores it (export_bytes_with). It
now shows for PNG, JPG and WebP only, and the suffix field takes its
width.
Refs #256 ("exporting to SVG is offered in different resolutions,
although this should not be the case").
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
* fix(ui): export size field as an Option; test WebP too (#256)
From review: the size field's width is an Option instead of a negative
width, and the test checks WebP as well as JPG.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
Unrelated to the security changes in this branch; `main` already failed
`clippy --workspace --all-targets -- -D warnings` without them. Eight
clippy::nonminimal_bool sites across seven files, each taking clippy's own
suggested rewrite.
Every rewrite is pure boolean algebra with the negation kept outside, so the
float comparisons and their NaN behaviour are untouched: !A && !B becomes
!(A || B), and !opt.is_some_and(f) becomes opt.is_none_or(!f). Two sites needed
explicit parentheses the suggestion omitted, where the rewritten term is the left
operand of an && and || binds looser.
designcraft cleared the same lint in bc840130; this is figcraft's equivalent.
clippy::needless_borrow; `url` is already a reference. Missed on review because a
workspace clippy run aborts at the first failing crate, which was figcraft-format.
A document carries things the app acts on for whoever opens it. Three of them were
taken as given.
Prototype Open link and Dev Mode dev resources both reached the platform's opener
with any scheme. In the web build the opener is window.open, so a saved document
could run script in the app's own origin; on the desktop any scheme went to the
system's handler. Both now pass through the new engine module `doc_urls`, which
allows http, https and mailto and nothing else. It is an allow-list on purpose: a
list of known-bad schemes is outgrown by every handler a platform adds.
A figcraft:// invite link joined the session at once. The system hands the app
whatever such link was followed, which may come from a page or a message, so the
Join dialog now opens with the link filled in and the person confirms.
A library link stored in a document was loaded from wherever it pointed. Only a
path on this computer is followed now, so opening someone else's file cannot make
the app contact their server (which on Windows would also send credentials).
Tests: each change has one that fails without it, verified by reverting. The
prototype test fires through an afterDelay trigger, since no click is needed, and
keeps a positive case so a refusal cannot pass vacuously.
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.
- Text editing on the canvas now tells the platform it wants the IME on **every** frame of the
edit. Before, it only did so on frames that brought input events.
- The candidate window opens at the caret. Before, it opened at the canvas corner: the desktop
integration places it by `IMEOutput.rect`, which was the whole canvas.
- Keys that reach the editor while a composition is in progress belong to the IME. Enter, Escape,
Backspace, arrows and shortcuts don't act on the text then, so Enter picking a candidate doesn't
add a newline.
- New test `crates/ui-egui/src/ime_tests.rs`.
- Docs: the IME rows in `docs/ui-parity.md` and `ROADMAP.md` now say it hasn't been checked on a Mac.
Fixes#747.
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]>
Escape or Enter while the pen draws ends the path and keeps its vector in
edit mode with the pen, so the next path joins the same layer; another
Escape leaves edit mode. Entering edit mode on a layer that already is a
vector no longer records an empty undo step.
Current macOS keeps PingFang and other on-demand fonts in its asset
store, which the font scan never walked, so Chinese text showed as
boxes. The scan now also walks the com_apple_MobileAsset_Font* folders
of /System/Library/AssetsV2 (and the older /System/Library/Assets) with
the same depth, size and file checks, and Hiragino Sans GB and Heiti SC
(preinstalled) join the Chinese UI faces and the text fallbacks.
Tested with a folder laid out the same way; not run on a Mac.
gridColumnsSizing and gridRowsSizing are CSS grid-template strings: px
tracks are fixed, fr tracks fill, auto/min-content/max-content hug,
minmax(_, Nfr) fills and repeat(n, ...) expands; other tracks stay equal
(1fr). Tracks past the axis's own count are left out. Files from .fig
keep reading the tracks from their source.
The format's SPACE_EVENLY is its older name for equal gaps without edge
space, so files keep reading it as space between; true space evenly is
SPACE_EVENLY_CSS, which the export now writes.
- The import report translates an importer note when the catalog has it, but 30 notes added by the
`.fig` import series (#626–#639) had no catalog entries, so they showed in English: connector
labels and snapshots, grid track sizing, image filters, slots and their content, vector segment
styles, and text on a path.
- All 30 are now in `en.toml` and in every full catalog: es, de, fr, pt-BR, ru, ja, ko, zh-Hans, cs,
id, sv and tr (30 each). `es-419` overrides only the 3 notes about settings (configuración rather
than Spain's ajustes), as the overlay does elsewhere. The wording follows each catalog's existing
terms for connector, layer, frame, stroke, grid, slot and text on path. The translations are drafts,
like the catalogs, and wait for native speakers' review.
- `every_key_in_the_code_is_in_the_english_catalog_and_no_other` now counts each importer note (the
`warn("…")` literals in `crates/format/src/fig` and `crates/format/src/rest`) as a required key, not
only as an allowed one. A new note that isn't in `en.toml` fails the check, and because Spanish is
in `COMPLETE` it has to be translated there too, so the report can't fall behind again.
`docs/translating.md` says so.
- The importers' texts are unchanged. Notes built with `format!` (layer counts, type names, field
names) still show in English, since the report matches whole notes.
A frame's name above its top edge (or a section's pill) only selected and
dragged it; a double-click there did nothing, since the label lies outside
the frame and hits no layer. With the Move or Scale tool, a double-click on
a label now selects that frame and asks the UI (`canvas.renameLabel`) for a
field over the label with the name selected, as the layers panel and pages
list do. Enter or a click elsewhere renames it (`object.rename`, one undo
step) and Escape leaves the name as it was. Rotated labels work too.
Tests drive the canvas: a double-click on a label opens the field, typing
and Enter rename the frame without moving it, undo restores the name, and
Escape cancels on a rotated frame.
From review:
- The check ran after any file.open/openRecent/import result, so a
pasted or dropped image, reopening a file that was already open, or
Clear recent brought the window back. It now runs only for results
with a `summary` (a file opened, or pages imported), and never in the
browser.
- Text > Missing fonts... before the startup scan finished listed
installed fonts as missing. It now says it is checking the installed
fonts and lists them once that is done
(`figcraft_text::installed_fonts_known`, also true when system fonts
are opted out, where checks used to wait forever).
- ROADMAP progress log and file-format-parity status line.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Menus and tooltips showed boxes for ⌃ ⌥ ↵ ⌫ ⇥ ⇞ ⇟ (e.g. Present ⌥⌘↵):
egui's default face has ⌘ and ⇧ but not these. The desktop app now adds
a system face that has them after the UI fonts.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Brings the progress docs up to date with what landed on 2026-10-17: community batches #727, #728, #733, #735, #736 and #753, plus #729, #745, #746 and #751. Docs only.
Text in a family that isn't installed was drawn in a fallback font
without a word. Now:
- `text.missingFonts` lists the families a document names that no
installed or built-in font provides: text layers and their styled
ranges, text styles, and string variables bound to a font family
(every mode, through aliases), with the number of text layers.
- The desktop app shows a Missing fonts window after the user opens a
file that has any: once installed fonts are indexed (before that every
one of them would count as missing) and after any import report is
closed. Files named on the command line count. Programmatic opens and
the browser build never open it by themselves. Text > Missing fonts...
shows it for the file in front; Escape closes it.
- `figcraft-cli info` prints `missing font` lines (`missingFonts` with
--json).
- file-format-parity: the row claimed REST import warnings that no code
produced.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
- A minimised or covered window now answers every control request that acts on the app's state
(`engine.execute`, `document.inspect`, `node.get`, `ui.inspect`, `ui.pointer`, `ui.key`, `ui.type`,
`ui.set`, `ui.menu.invoke`, `ui.render`, `window.list`, `window.focus`, `app.quit`), in that window's
file, from the app's `logic` (no frame drawn). Before, only six methods were answered there.
- What only a drawn frame can answer (`ui.screenshot`, `ui.input`, `window.new`) is an error at once
for a hidden window: `window 0 is minimised or hidden: `ui.screenshot` needs it on screen
(`ui.render` draws the document without it)`. Before, these waited up to 60 s.
- `crates/ui-egui/src/control.rs`: the `NO_FRAME` allow-list becomes a short `NEEDS_FRAME` list;
`drain_hidden` answers the rest and fails the frame-only ones at once.
- Docs: `docs/control-protocol.md` (hidden windows, `ui.screenshot`) and `docs/mcp.md` (attached
mode while the window is minimised). ROADMAP progress log and revision history.
Part of #720.
- `figcraft-cli mcp --automation-read-root DIR --automation-write-root DIR` (each repeatable, `--flag=DIR` works too). When either is given, every file an agent's tool call reads must lie inside a read root or a write root, and every file it writes inside a write root. Anything else is an ordinary tool error (`automation path rejected: … is outside the folders that may be written: …`); nothing is written and the server keeps serving. Without the flags nothing changes.
- The desktop app takes the same flags (`figcraft --control PORT --automation-read-root DIR …`) and confines every control request and every frame of every window to them, so `figcraft-cli mcp --attach` drives a confined app. `--attach` together with the flags is a usage error, because the app runs the commands and only the app can confine them.
- New engine module `crates/engine/src/file_access.rs`: `Roots` (resolution and comparison), a thread-scoped `confined(roots, f)`, and the file primitives (`read`, `read_capped`, `open`, `metadata`, `read_dir`, `write` (atomic), `create_dir_all`, `exists`, `is_dir`). Every engine file access now goes through it. `crates/engine/clippy.toml` denies `std::fs::*`, `File::open/create`, `OpenOptions::open`, `Path::exists/is_file/is_dir/metadata/read_dir/canonicalize` and `figcraft_format::{write_file_atomic, save_file, load_file}` everywhere else in the engine, so a command added later is confined without a check of its own.
- `ui.screenshot` and `ui.render` return the PNG as base64 `data` when no `path` is given (before, `ui.screenshot` wrote `figcraft-screenshot.png` in the working directory and `ui.render` required a path). The attached MCP backend uses that, so it no longer writes and deletes temporary PNGs, and a confined app needs no exemption for them.
- `initialize`'s `instructions` name the folders the agent may read and write.
- `figcraft-cli` options may be given more than once (`Args::values`); single-valued options keep the last value, as before.
- Docs: `docs/mcp.md` › Confining file access (options, semantics, coverage, resolution, limits), `docs/control-protocol.md` › Confining file access and the optional `path` of `ui.screenshot` / `ui.render`, `docs/cli.md`, `figcraft-cli --help`, and the Automation line plus a revision row in `docs/target-app-parity.md`.
Fixes#715, fixes#481.
An auto layout frame a .fig records no primary-axis sizing for hugs on
that axis (auto layout sized it by its content before that axis could be
fixed) instead of turning fixed; an instance that records no auto layout
takes its main's; REST frames whose sizing modes are left out hug (the
documented default); text hugs by its resize mode (auto width both axes,
auto height its height) instead of coming in fixed.
A layer inside an instance can now override its resizing (hug, fill or
fixed, with min and max sizes; override group `resizing`) and its size.
They were refused, and imported ones were dropped at the first relayout.
Kept through rebuilds, undo, save, .fig and REST import, .fig export and
detach; reset per group and with Reset all changes. Position, constraints
and absolute or grid placement stay the main's.
Adds the progress rows for the community PRs landed today: batches #727, #728, #733, #735 and #736, plus #729. That's 55 PRs, and the batch-13 boolean-op crash guard.
- `ROADMAP.md` and `docs/roadmap.md`: the header's "Last updated/Change" line and one revision-history row each.
- Headline numbers are unchanged. The per-area rows for the `.fig` import and component overrides come with the #731/#732 PR.
Found while testing crates/engine/src/cmd/edit.rs:499 (copy_to_clipboard, called from copy at :536 and cut at :542)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
One value per two bytes of the compressed canvas was too tight for a
public design system: a 39.5 MB canvas holding 29.7 million values got
a budget of 19.7 million (reported in #701). Real files hold 0.37 to 0.75
values per compressed byte; the budget is now one per byte, at least the
old 20 million and at most 128 million. A small file still gets main's
20 million, however far it unpacks.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
`document.json` was read under the 512 MB cap for any archive member, so a
large imported document saved but didn't open again: 900 000 layers save
as a 1.48 GB document.json, packed into 244 MB. It may now unpack to 16
times its packed size (counted up to the archive's own size), within
512 MB and 4 GB; images and other members keep 512 MB. A small archive
still can't unpack into gigabytes.
That saved file reopens in 13 s with a 5.9 GB peak.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
A real 705 MB .fig file (401 644 nodes, 894 004 layers once instances are
expanded) didn't open: "message: too many values". Lifting that cap alone
made it import only its first 200 000 layers, and lifting that one took
41 GB of memory.
The global caps stay as they are. The .fig import gets budgets from the
size of its compressed canvas instead (counted up to the file's own size),
not from the message it unpacks to, since chunks inflate up to 512 MB:
- the kiwi value budget is one value per 2 bytes of the canvas, at least
the old 20 million (kiwi is only read for .fig files);
- the layer cap is one layer per 32 bytes of the canvas, at least the REST
import's 200 000 and at most 2 000 000. A small or crafted file can't
expand its instances into millions of layers, however far it unpacks; the
REST JSON import keeps 200 000.
Memory:
- The mapper no longer passes the document through `json!`, which serialises
its arguments: the whole tree was deep-copied twice.
- Vector networks are built once per distinct network and handed to the REST
importer beside the JSON (`vectorNetworkRef`), not written into it as JSON
objects for every instance that shows them.
- Decoded kiwi objects are trimmed, and each import stage's input is dropped
as soon as the next stage is built.
On that file: peak footprint 41 GB -> 11.3 GB, 67 s -> 25 s.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
- A search covers every command again (the engine's and the UI's own), grouped by
menu where one lists it and under Other otherwise, best matches first (prefix,
then shortest), at most 60. Without a query the menus' items show as before.
- Up and Down move a visible highlight (the pointer moves it too); Enter runs
the highlighted row, not an invisible first one.
- Only runs from the palette are kept as Recents; menus, shortcuts and context
menus no longer write the preferences. Recents match the translated label.
- The Plugins menu's items (run last, manage, import) show with the plugins.
- The search keeps its hint, so every catalog keeps its translation.
- The panel sits above the toolbar pill's own rect (toolbar::pill_rect); one
set_palette/toggle_palette opens it from every entry point.
- Clippy: needless borrows in the toolbar.
A Wayland clipboard read waits on the app that copied with no timeout of its
own. The read now runs on a helper thread with its own clipboard handle, so a
stuck app can't hang the window.
#639 lands first and reads networks with decode_styled; once #631 drops the count it calls
network::decode. Needed whenever #631 is rebased onto a main that has #639.
Image filter warnings name fields of the file, so a crafted schema could add any number of
distinct warnings (each checked against all earlier ones). The mapper keeps at most as many as
the REST importer does.
The Typography section's variable picker also lists number variables for a text layer's content (shown as the number), as the engine already accepted; other text fields keep their own types (#629).
Both carry the same image filter value, which the public docs describe as green when lower and
magenta when higher; a positive native tint adds green. Both readers flip it now (the .fig path
did, the REST path didn't). The .fig test message no longer names the product.
The test schema extended the image scale-mode field that only #634 adds; it now adds the
filter containers to the paint message itself, so it works with and without #634.
Image paints that carry a current filter container now import their exposure, contrast, vibrance (as saturation), temperature, tint, highlights and shadows as native image filters. Neutral values leave the paint untouched. The older color-adjust container alone, both containers together, filters that have no native equivalent (brightness, detail, vignette and unknown fields) and malformed or non-finite values are skipped and reported as import notes. A positive tint in .fig data moves toward magenta, so it is flipped once at the .fig boundary; public REST image filters keep their sign and are not reported as approximated. Imported filters use the approximate native renderer, which the import says once.
Supported TEXT_PATH layers import as editable native text on their vector path, including mixed text styles, a start inside a later segment and reversed text direction. The importer converts the segment parameter stored in the file to the native distance along the complete path.
The mapping accepts a connected, unbranched path with one recognized start-segment marker and the default placement above the path. Damaged or unsupported path data keeps the text readable off its path and adds an import note. Other vertical placements, multiline text and ambiguous or disconnected networks are not converted. Underline and strikethrough remain a native path-rendering limitation and are reported. Public REST TEXT_PATH layers and the experimental .fig writer are unchanged.
The vector network decoder now keeps each segment's style id. Import reports a segment style that names a row of the layer's style table as metadata that was not interpreted, and one that names no row as a reference to a missing style entry, each at most once per file. The earlier note claimed that stroke styles were left out, which overstated what is known about the metadata. Segment geometry, paint and width are unchanged, and the style ids are not kept in the native vector model.
The public docs don't give a default for skip ink; they say the value is unset when the text
isn't underlined, and the exporter always wrote it on. So the export keeps writing on for text
without an underline and the real value under one (that's what the canvas draws), and the import
leaves a value on text without an underline out of the decoration look, so plain text keeps the
default look and round-trips unchanged. The native default stays off, so existing documents
draw as before.
Plugins can no longer reach the disk or network through path parameters (variable.export/codeConnect.export/component.playground writes, codeConnect.import/variable.import reads) or through collab.* and window.*; found and fixed by @lucasmartins-ai, split out of #696.
One value per two bytes of the compressed canvas was too tight for a
public design system: a 39.5 MB canvas holding 29.7 million values got
a budget of 19.7 million (reported in #701). Real files hold 0.37 to 0.75
values per compressed byte; the budget is now one per byte, at least the
old 20 million and at most 128 million. A small file still gets main's
20 million, however far it unpacks.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
DocState::new already settles every opened document (native files, branches and live joins
included); the .fig/REST path keeps its settlement in link_instances for file.import into an
open document.
Frames on the page or in a section convert to sections (object.convertToSection) and sections back to frames (object.convertToFrame), in place, one undo step each, from the canvas context menu; nested frames, instances, main components and sections holding sections are refused.
Every shell reports a modifier change before the key it goes with, so key
events needn't update the held modifiers too. The old test that pressed
Shift+Cmd+V and Alt+Cmd+V expecting "nothing" now covers plain V only:
those chords run their commands, covered by their own tests.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
egui-winit swallows every Cmd/Ctrl+V press, Shift or Alt held or not, and
sends a paste event only for text. Shift+Cmd+V and Alt+Cmd+V became a
plain paste (text) or did nothing (image). The keymap now decides what a
swallowed chord means, from the modifiers tracked in the raw events: its
paste event or its release runs paste over or paste properties instead.
In the browser the key press runs the command, and the paste event no
longer pastes its text as well.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A 1 px Line or stroke-only path only counted as hit within 2 screen
pixels, so a double-click slightly off it cleared the selection and
never entered vector edit mode, and Cmd/Ctrl could not bend it.
Strokes with nothing filled now reach 5 screen pixels; filled shapes
keep the plain tolerance.
Found while testing crates/engine/src/cmd/structure.rs:575
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The Actions button toggles (re-click and Cmd+K close), reads on in
accent blue with a white icon while open, and a press anywhere outside
closes the panel in one click. The opening press is spared once (the
spare expires on press-free frames); presses on the button itself never
dismiss, so the release toggles cleanly shut instead of reopening.
Rows drop the per-family icons (labels + right-side hints only);
plugins and widgets keep their letter tiles. The panel anchors to the
canvas centre at the toolbar pill's exact width (toolbar::main_width),
at a fixed size, with inset hover pills and larger section titles.
Tests: one-click dismiss pinned (palette_tests).
Tabs (All, Assets, Plugins & widgets), Recents kept in the preferences
(newest first, capped at 5), commands grouped by top-level menu, this
file's components (placed at the view centre) and styles (applied to
the selection), installed plugins and widgets with letter tiles. Rows
carry right-side shortcut/kind hints; Enter runs the first visible row.
The panel floats 8 px above the toolbar pill with its exact width and
centre (toolbar::main_width, canvas-centred), at a fixed size so tab
switches never resize it. Tabs keep their frames when idle so the
active pill never moves its neighbours.
Engine: Prefs.recent_actions (persisted, skipped when empty). UI:
invoke/menu/plugin picks record recents; gone entries drop out.
Locales: new keys in en + es, retired the old search hint everywhere.
Docs: ui-parity row describes the new surface (palette_tests).
The .fig export now writes a numbered list's start on each of its lines,
so a reader that takes it per line or per list counts the same. Bullets
no longer keep a start. The field is named start_offset (it holds the
offset: 1 shows "2."). get_node reports listStartOffsets. The parity doc
lists what's still not read from .fig lists.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Found while testing crates/engine/src/cmd/props.rs:361 (strokeWeight is only clamped with .max(0.0)); the panic is raised when the renderer draws the stroke (render crate / tiny-skia stroker)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
winit sends no press for a key held across a focus change (X11's
synthetic presses are dropped by egui-winit), so letting go of V after
coming back looked like a swallowed Cmd+V. After a focus gain, a V
release only pastes with Cmd still held.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Figma stores where a numbered list starts counting on each text line
(listStartOffset; a list shown as "2." has offset 1). The import dropped
it, so every list counted from 1.
A paragraph's list settings now carry a start number (written only when
set, so older files open and save unchanged). Numbering counts on from
it where a list starts; items that go on counting ignore it. The .fig
import reads it, the .fig export writes it on a list's first line, and
plain paragraphs drop it. Toggling a list off ignores differing starts.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
With Shift let go before the button, the release took the layer out of
the selection and also ran the widget's click. The take-out now skips
the widget click, as a Shift-click always has.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
egui-winit swallows the Cmd/Ctrl+V press and sends a paste event only
when the clipboard holds text, so with an image on the clipboard only
the key's release arrives and the image fallback, which waited for the
press, never ran. Paste from a V release whose press never came and
brought no paste event; the press and paste event are tracked from the
raw events at the start of each frame.
On Wayland arboard read the X11 clipboard, which the compositor does
not fill for a native Wayland window, so every system paste found it
empty. Enable arboard's wayland-data-control: it uses the Wayland
clipboard when WAYLAND_DISPLAY is set and falls back to X11 otherwise.
Right-click > Paste here now follows the same newest-wins rule, so a
newer image beats copied layers there too, and lands at the point.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Shift on a selected layer took it out of the selection on press, so the
drag that followed had nothing to move, and with several layers selected
the grabbed one stayed behind. It now leaves the selection on release,
and only when the press didn't become a drag. Where no move can start
(a version being viewed), it still leaves on press.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A log whose generation isn't its snapshot's was skipped as a leftover of a
compaction cut short, but neither generation is under a checksum: one flipped
bit made the server serve the older document, report nothing and overwrite the
log. Replay such a log over the snapshot instead: replay skips what the
snapshot already holds, so a genuine leftover changes nothing, while edits the
snapshot lacks are recovered and the damage is reported.
Found while testing crates/engine/src/cmd/edit.rs:499 (copy_to_clipboard, called from copy at :536 and cut at :542)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
`node.create` read its `parent` as `id(p, "parent").unwrap_or(page)`, so a `parent` that isn't a
layer id — a `{"$step": n}` nothing had resolved, a name, a negative number — quietly meant "no
parent" and the layer landed on the page, with the call reporting success. The same shape sat on
`image.place` and `instance.create`, on `index` (a string index appended where it was asked to
insert), and on the `ids` of every command that targets layers (an entry that wasn't an id was
dropped and the rest ran). A number that names nothing was reported; anything that isn't a number at
all was not.
`params::opt_id` and `params::opt_index` now say what the commands mean: missing and `null` are
"none", and anything else has to be a layer id (or a whole, non-negative number) and is named when it
isn't. `req_targets` refuses an `ids` entry that isn't an id, before falling back to the selection. A
number that names no layer still reports `no such node`, and a boolean is still a number to `f64_of`,
as it is for `x`, `width` and the rest.
This turns a silently wrong result into an error on purpose, so a caller that passed something that
isn't an id and relied on the fallback now hears about it.
`figcraft-cli run` ran each step of a script on its own, so a script's `{"$step": n}` reached
`parent` unresolved and the layer landed on the page while every step reported success — the same
commands through `command.batch` (the MCP `batch` tool, the control channel, plugins) put it inside
the frame. One `edit.undo` took back one step of a script rather than the script, and a failing step
left the earlier ones applied.
`Session::batch_step` is now the one path both take — one step of a batch, its `$step` references
resolved against the results before it — and `run_script` wraps its steps in `Session::grouped`: the
whole script is one undo step, and a step that fails takes the script back. The result lines come out
once the script is through, so none of them reports work a later failure took back; a failure prints
one line with `rolledBack: true` and exits 1. `docs/cli.md` documents the reference and the output.
Duplicate repeats the last duplicate's offset, but of the ways to move the copy
before the second Cmd+D only `node.set` and Alt-drag were covered. Dragging it
with the mouse is how almost everyone does it, and is what #567 describes; an
arrow-key nudge is how a small offset is usually made exact. Both work, because
the step is measured from where the copy ended up rather than recorded from the
gesture, but nothing held them there.
The drag test also asserts the drag moved the copy rather than making another
one, left it selected, and left the original in place: the ways the feature
could break without the step arithmetic breaking.
Both fail if the step lookup in `edit.duplicate` is stubbed out.
`cargo xtask ci` was red on a clean main on macOS: the test's last shutdown
runs after the server has answered an invalid request and closed the
connection, and `shutdown(2)` on a socket whose peer has already closed is
ENOTCONN on macOS and the BSDs where Linux returns success.
The test is right about what it checks; it checked a return value that is not
portable. Closing the client is what the worker thread waits for, and dropping
the socket does that whether or not shutdown reported anything. The file's
other six shutdowns run while the peer is still open, which is well defined on
both platforms, so they stay checked.
With no CI running the tests on pull requests, the local gate is the only
signal a contributor gets, and it could not be green on macOS.
Node lookups in the document's B-tree were still a third of a render on a
900 000-layer file: a small subtree was walked for its bounds, again to
see whether it could be splatted, and again to splat it, and each child
of a drawn layer was looked up by paint_runs, by the mask check and by
draw_node.
- The render's table also keeps the layers it looked up, keyed with a
multiply hash instead of SipHash (ids are counters), and so do the
bounds it keeps.
- draw_children looks each child up once, works out the paint runs from
those (Document::paint_runs_of) and draws the child it has.
- Whether a layer can be splatted is kept like its bounds: layers with
children only, answers the nesting cap decided aren't kept, and kept
answers are read only where their levels fit under the cap.
Ctx gets a second lifetime, for the document the kept layers come from.
Single-threaded cycles per 1440x900 render fitted to the page: 0.39 ->
0.34 G on a page of 246 204 layers, 0.26 -> 0.22 G on one of 118 358.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
Culling a layer asked for its paint bounds, which walk its whole subtree,
and then did the same for each child: every layer's subtree was walked
again at each level above it, and once more in each parallel band. The
bounds worked out during a render are now kept for it (layers with
children only; bounds the 64-level cap cut short are not kept, and kept
ones are read only where their levels still fit under the cap, so culling
is exactly as before).
Every render also looked through every layer of the document for the
largest background blur, to size the overlap of its bands: 900 000 layers
on one large file, on the main thread before the bands start. The
document now works it out when first asked after a change, like whether
it has masks.
A 1440x900 render on a 900 000-layer file, page of 118 358 layers: at
100% 13.0 -> 0.6 ms, at 25% 16.5 -> 3.4 ms, fitted to the page 56 -> 31 ms;
page of 246 204 layers: 12.6 -> 2.5 ms, 12.1 -> 2.0 ms, 69 -> 62 ms.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
Hovering a family in the Typography font dropdown previews it on the
canvas via text.preview; closing without picking restores the original,
clicking applies it as one undo step.
In vector edit mode a point dragged onto another point of its path
snapped there but stayed a separate point, so a path couldn't be closed
or joined by dragging. A single dragged point dropped within the point
hit radius (6 px) of another point now merges into it, in the same undo
step as the move, and that point stays selected.
Closes#401
Copies written by plugins and tools carry their image bytes in the
message: a paint's image.dataBlob indexes the message's blobs. A bare
canvas (what the clipboard holds, see #647) has no images/, so these
fills came in as grey placeholders. The importer now falls back to the
embedded bytes, keyed by image hash, after the archive's images/.
Blobs over the image size cap and indexes past the blobs are skipped and
stay placeholders. Tests: a bare canvas carrying its image in a blob, and
a dataBlob past the blobs.
Found while testing crates/engine/src/cmd/edit.rs:1180
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The pages list took a fixed 40% of the left panel (#272). The line under
it now drags to make the list taller or shorter: at least three rows, and
always leaving the layers their header and three rows. A double-click puts
back the default. The main window keeps the height in the preferences, as
it does the panel widths, saved once the drag is over.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
Flatten (Cmd+E, Alt+Shift+F) and Join selection (Cmd+J) built the result
from each layer's fill outline, so open paths, which have none, were
dropped: two pen paths flattened to "nothing to flatten", and next to a
closed shape they vanished. Vectors now bring their own network (open
paths, unfilled loops and each region's winding stay), mapped into the
parent's space; other layers still flatten to their outline.
Closes#377
`document.json` was read under the 512 MB cap for any archive member, so a
large imported document saved but didn't open again: 900 000 layers save
as a 1.48 GB document.json, packed into 244 MB. It may now unpack to 16
times its packed size (counted up to the archive's own size), within
512 MB and 4 GB; images and other members keep 512 MB. A small archive
still can't unpack into gigabytes.
That saved file reopens in 13 s with a 5.9 GB peak.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
A real 705 MB .fig file (401 644 nodes, 894 004 layers once instances are
expanded) didn't open: "message: too many values". Lifting that cap alone
made it import only its first 200 000 layers, and lifting that one took
41 GB of memory.
The global caps stay as they are. The .fig import gets budgets from the
size of its compressed canvas instead (counted up to the file's own size),
not from the message it unpacks to, since chunks inflate up to 512 MB:
- the kiwi value budget is one value per 2 bytes of the canvas, at least
the old 20 million (kiwi is only read for .fig files);
- the layer cap is one layer per 32 bytes of the canvas, at least the REST
import's 200 000 and at most 2 000 000. A small or crafted file can't
expand its instances into millions of layers, however far it unpacks; the
REST JSON import keeps 200 000.
Memory:
- The mapper no longer passes the document through `json!`, which serialises
its arguments: the whole tree was deep-copied twice.
- Vector networks are built once per distinct network and handed to the REST
importer beside the JSON (`vectorNetworkRef`), not written into it as JSON
objects for every instance that shows them.
- Decoded kiwi objects are trimmed, and each import stage's input is dropped
as soon as the next stage is built.
On that file: peak footprint 41 GB -> 11.3 GB, 67 s -> 25 s.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
A library file is cached by its length and modification time. File systems keep the time in coarse ticks, so a same-length rewrite right after a read can keep the stamp, and library.updates then reports no change. A stamp now proves the file unchanged only once its modification time lies at least two seconds before the read; otherwise the file is read again.
Figma puts copied layers in the clipboard's HTML as a base64 .fig canvas
between (figma) markers. edit.pasteFromFigma reads it with the .fig import,
turns the layers into the session clipboard and pastes them as edit.paste
does (selected frame, in place or view centre), with the import's warnings.
The desktop app reads the clipboard's HTML (arboard) and Paste prefers a
Figma copy; the browser's paste event carries plain text only.
Tested on copies written by our own .fig export; not yet on real copies.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
ROADMAP.md, docs/localization-parity.md and docs/translating.md said
Swedish had 120 of the 588 command labels and menu wordings; with the
previous commit it has all of them. Worded without counts, so the docs
stay right as commands are added.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The [commands] and [menu] rows held their command id as the text
("align.bottom" = "align.bottom"), which the menus and the command
palette would show as is. They now translate the English labels the
code gives those ids (spec! and m! in crates/engine), e.g.
"align.bottom" = "Justera nederkant". Also the interface strings that
are new on upstream main (23bdebd).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Straight connectors saved with a manual routing snapshot (connector and derived-frame version 1) now come in as an editable vector plus their label text layers, instead of an empty frame. Label placement composes the saved background and text transforms, and label paints and text styles use the existing import path. Connectors of other versions or line styles keep the existing fallback with their compatibility notes. At the layer depth or node limit the saved vector is kept without extra label layers, and skipped labels are reported once. Live endpoint bindings and routing are not retained.
Text layers, text styles and instance overrides in .fig files can carry list spacing, hanging list, hanging punctuation and leading trim. The importer now reads them from the highest layer with a valid value, so an invalid instance override row falls through to the valid value below it, while explicit zero, false and NONE values stay resets. Both text layers and text styles share one mapping. The native text data stores these options once per text node, so differing values on separate runs are not represented, and the experimental .fig exporter keeps its existing note that these settings are left out.
Text layers, styled runs and text styles in .fig files carry a skip-ink boolean for their decoration. The importer now reads it into the native decoration look: an omitted run value inherits the layer's, a local value overrides a linked shared text style without changing the style itself, and instance overrides keep an explicit false over a true main. A malformed value counts as unset. The exporter writes the actual value of the layer and of each styled run instead of a constant true, and a decoration look that differs from the default only in skip ink no longer triggers the unsupported-decoration note, so default native text exports its real false value.
Supported TEXT_PATH layers import as editable native text on their vector path, including mixed text styles, a start inside a later segment and reversed text direction. The importer converts the segment parameter stored in the file to the native distance along the complete path.
The mapping accepts a connected, unbranched path with one recognized start-segment marker and the default placement above the path. Damaged or unsupported path data keeps the text readable off its path and adds an import note. Other vertical placements, multiline text and ambiguous or disconnected networks are not converted. Underline and strikethrough remain a native path-rendering limitation and are reported. Public REST TEXT_PATH layers and the experimental .fig writer are unchanged.
The vector network decoder now keeps each segment's style id. Import reports a segment style that names a row of the layer's style table as metadata that was not interpreted, and one that names no row as a reference to a missing style entry, each at most once per file. The earlier note claimed that stroke styles were left out, which overstated what is known about the metadata. Segment geometry, paint and width are unchanged, and the style ids are not kept in the native vector model.
Slot properties of a component, the frames that reference them and the content that instances assign to them now come across from .fig files. A slot definition becomes a native slot property. The referencing frame of an expanded instance keeps following the frame of its main, and when the instance assigns slot content (including an explicitly empty slot) the frame carries that content as its own, independently owned children. Nested instances inside assigned content keep their own source links. An assignment on an outer instance wins over one made by an instance nested in the main, and a slot without an assignment keeps the default content of its main.
Assignments that are malformed, use the empty-GUID sentinel or point at a layer that is not slot content are skipped with an import note. Preferred component keys cannot be mapped to nodes of the file and are dropped with a note, as are slot settings the editor cannot represent. Public REST snapshots, the experimental .fig writer and property resources are not changed.
Grid columns and rows keep fixed pixel sizes, Hug sizing and fractional Fill weights when the minimum and maximum sizing definitions agree. Track GUIDs join the sizing records to the ordered tracks, so reordered records do not change cell placement. Unsupported min/max pairs, invalid sizes and duplicate records fall back with a reported note, and imports are bounded to 1,000 tracks per axis. Public REST snapshots ignore the private track metadata.
Found while testing crates/engine/src/cmd/arrange.rs:132
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A text layer that links a shared text style keeps its own, possibly stale, copy
of the typography. The style's typography now beats that copy, a case,
decoration, paragraph spacing or paragraph indent set on the layer itself beats
the style (so "original", "none" and zero clear the style's value), and an
override row of an instance beats both. An explicit "no style" on a text layer
unlinks it even when the older style field still names one.
Letter spacing given as a percentage of the font size stays a percentage. A text
property value on an instance no longer inherits the character styles written
for the main's text, which count other characters.
Override rows of an instance can name a layer of its main by the key the layer
had before the main was edited. The key now resolves to the layer's current id
when it is unambiguous and the current id does not exist already. Rows of an
enclosing instance win over those of the instances nested in it, whatever order
the wire lists them in, and an explicit "no style" on an override detaches the
style instead of falling back to the older field.
Paints bound to variables take the colours of the active modes when a document
loads, before the first edit, without touching other pending changes.
A stroke-only vector that imports as the outline of its stroke now moves the
stroke's shared style and colour variables to the fill it becomes, and vectors
whose single segments carry their own stroke style say so in a warning.
A .fig import now links the layers of an instance to the layers of the
component they came from by the ids the file stores, instead of leaving it to
the engine to guess from names and places. Renamed slots and layers, siblings
that share a name and swapped mains keep their source link and the override
they carry.
Components that live only on the file's hidden internal pages (cached copies of
library components) are kept in the hidden library store when a visible layer
uses them, so their instances stay instances. The read-only guard for library
copies now compares the copies themselves rather than the dirty set, which an
import leaves full until the first edit.
Rebasing an instance (after detaching an outer instance or restoring a main)
follows the source chain inside the right enclosing instance, so contextual
links survive it. The last change of a GUID wins and a removed change deletes
it. Ordinary file JSON never reads the private provenance.
Importing such a file into an open document, or pasting its layers into another one, copies the cached components along (import) or leaves them behind as a missing main (paste), so no link points at a layer of the target.
Clicking the + button to add a fill or stroke in the Design panel now
immediately opens the colour picker dialog for the newly added paint,
matching Figma's behaviour.
The canvas renderer and PDF export stroked text with the alignment
forced to centre, so Inside and Outside on text did nothing (SVG export
already honoured them). Glyph outlines are closed, so text strokes now
go through stroke_area like other shapes' strokes, with the same fill
rule: inside clips to the glyphs, outside to around them.
Opening or importing a .fig or REST JSON file, or an SVG that has notes,
from the app (File > Open or Import, a recent file, a drop, a paste, the
system's open event) now shows an Import report window with the
document's name, its page, layer and image counts and the compatibility
notes the importer collected, instead of leaving them only in the command
result. Commands sent over the control channel or MCP never open it.
file.open and file.import return the counts as `summary`; ui.inspect
reports the open report as `importReport`. The fixed notes the importers
emit are translated (German and Spanish) and shown with a capital first
letter; other notes, such as SVG's, show as written. The report goes
away with its document. Escape or Close dismisses it.
Save as now offers .figcraft and .fig as separate file types. A .fig
destination, chosen in the dialog or passed to file.saveAs and file.save,
writes the existing experimental .fig export of the document that is shown
and reports how much was left out. The file's own path, unsaved changes,
versions, branches and undo history are not touched, and a shown earlier
version stays shown while the copy is written or a Save as choice is
pending.
Plain Save no longer writes over a file whose saved path ends in .fig;
earlier builds could write native data under that name. It asks for a
destination instead.
Figma's schema has leadingTrim on NodeChange (NONE, CAP_HEIGHT), and
the REST importer already reads style.leadingTrim, but the .fig mapping
never copied it, so text set to "Cap height to baseline" imported
untrimmed in its trimmed box. text_style now copies it with the other
text fields, from the layer or its text style.
Vector points carry their own stroke join in a .fig file. Import now keeps each point's join as the point's own override instead of dropping it for the layer's join.
Joins equal to the layer join no longer force styled stroke runs, and an absent layer join counts as miter. Unresolved point styles, unknown joins, and joins that dashed or variable-width strokes cannot draw each produce a warning.
Pasting or dropping a file runs file.import with the whole file as Base64
data, and every successful command was journaled with its parameters, so
each imported file stayed in memory a second time for the session. Byte
imports, including ones wrapped in command.batch, are no longer journaled.
Imports by path and all other commands are unchanged.
- Several ids: auto layout keys go to the frame-like layers, the others get the remaining keys
and are named in `warnings`. Refused only when no target can have auto layout.
- fontStyle is never refused: it draws in the nearest face, as a text layer resolves it.
`warnings` says so when that face's weight or slant differs from what the name asks for.
- set_properties passes `warnings` on to the agent.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A home-screen drop that carries only bytes queued file.open with data and name, which file.open rejected because it requires a path. file.open now also takes base64 data with a name and opens it as a new document without a save path.
Found while testing crates/engine/src/cmd/file.rs:1283
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Preserve the upstream behaviour and regression tests alongside this feature. Align the control-server connection-limit test with main closing invalid JSON connections: assert EOF instead of shutting down an already closed socket on macOS.
Preserve the upstream behaviour and regression tests alongside this feature. Align the control-server connection-limit test with main closing invalid JSON connections: assert EOF instead of shutting down an already closed socket on macOS.
Regression test for #586: with the colour picker open on a red rectangle or
pen path, duplicating it (edit.duplicate, or copy and paste) and recolouring
the copy leaves the original red. The cause was the picker staying bound to
the original after the selection moved to the copy, fixed by #523; without
that fix this test fails. The duplicated paint is already independent data.
- Files dragged over the window show a "Drop to open" (home screen) or
"Drop to import" (canvas) overlay.
- On the home screen each dropped file the app opens becomes a document of its
own; any other file is turned away with a toast.
- On the canvas dropped files import as before, at the drop point.
- A small card ("Opening as a new file", "Adding to {document}") is on screen
before a large file blocks the window; the files then open a frame later.
- A canvas import stays with the document and page that received the drop,
keeps its placement, restores the page and selection chosen meanwhile, and is
cancelled with a message if that document or page is gone.
- ui.inspect reports pendingOpens; every new string, including the cancel
messages, goes through tr() with en, de and es entries.
page.duplicate copies every layer with duplicate_all_into (navigation between
copied frames points at the copies) and builds the old-to-new list for flows
from that copy, instead of copying each layer separately.
Main's case-insensitive replace loop (from an earlier PR) took the place of the
loop this change edited, so the replacement text is normalised once, where
the parameter is read.
Czech counts take three integer forms: one (1), few (2-4) and other
(0 and 5+). Since #257 catalogs can supply `few`, so `plural()` gets a
Czech rule and the 38 plural entries in cs.toml now read naturally
("1 vrstva", "3 vrstvy", "5 vrstev") instead of the count-neutral
"Vrstvy: {n}" that #469 used for everything but 1.
Test czech_integer_plurals covers the rule (1; 2-4; 0, 5, 11, 12, 21,
22, 100, 1 000 000), a real key through counted(), and that every Czech
plural entry has one and few. Docs no longer say Czech has only one/other.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The control server answered a line that wasn't JSON with an error and read
on. A web page can make the browser send an HTTP request to the loopback
port; its request and header lines were skipped that way and the JSON lines
of its body ran as commands. The first line that isn't JSON now closes the
connection before anything after it is read. Clients that send JSON lines
(MCP --attach, scripts) are unaffected.
Main's file dialogs answer asynchronously now (they no longer re-enter the
event loop on macOS); the menu item takes the path in a callback, and the test
draws a frame to let the dialog answer.
468 of the 588 command and menu rows had the command id as their text
("align.bottom" = "align.bottom"), which the app would have shown as is.
They are left out, so English shows until they are translated. Swedish
isn't in the complete list (that would make every new UI string need a
Swedish translation in the same change); the docs say what it covers.
A text edit holds an interaction too, so treating any open interaction as a
drag dropped the colours clicked while editing text. Recent colours dedupe by
8-bit value: the colour picked and the one the document stores differ by a
rounding fraction.
The message now comes from the UI crate (dialogs::window_failed) through
tr() and trf(), with entries in every catalog, and the app picks the
language from the preferences before the event loop starts, so the box
reads in it even though no frame was ever drawn. The commands to type stay
as written.
Release builds on Windows have no console, so an eframe::run_native error
(no usable graphics adapter, a device the driver refuses) went to an unseen
standard error and the app quit silently. The error is now logged and shown
in a message box with a hint: unset WGPU_BACKEND when it is set, otherwise
update the driver or try another graphics API (OpenGL on Windows), and
where to report it. With --control the box is skipped so agents aren't
blocked.
With drop shadows the backdrop was read twice: before the content and
again after the shadows. The first read is only what backdrop effects
inside the content see, so it now runs only when the layer has such
descendants, and the content is copied for the shadow caster only when it
gets a layer blur, noise or texture. The output is unchanged.
A layer with a background blur (or glass) drew its blurred backdrop into
its own pixels before casting its drop shadows from them, so a translucent
fill cast a shadow as dark as an opaque one, and with "show behind
transparent areas" the shadow was cast from the backdrop's pixels. The
backdrop was also read before the shadows were drawn, so a shadow showing
behind was never seen through the blur.
When a layer has both, it now casts its shadows from what it paints (with
the layer blur, noise and texture it will get), draws them, and only then
reads the backdrop, which goes under its content; the content still sees
the first backdrop under it while it's drawn. Layers without both draw as
before.
A click or Tab into a one-line field selects all of its text, so typing
replaces the value, as in Figma: hex and colour fields, inline fields
(a fill's hex, variable values, renames), names, the font size field,
search boxes and dialog fields (widgets::select_all_on_focus). A drag
from the press still selects only what it covers; multi-line fields
keep the caret where clicked.
The first click into an inline field (a fill's hex) now takes the
keyboard: egui dropped the focus of the editor created by that click,
so it took a second click.
Text in the default Inter (or any Latin font) drew the missing-glyph box for every Hangul
syllable on macOS and Windows: no Korean family was on the fallback list, so the Korean
fonts those systems ship were never tried, and there is no default-font setting to work
around it. Apple SD Gothic Neo (macOS), Malgun Gothic (Windows) and Noto Sans CJK KR /
Noto Sans KR (Linux) now come after the Chinese and Japanese entries, so Han characters
resolve as before.
Fallback faces also follow the run's style: each fallback family gives its face nearest the
run's style instead of always its Regular, so Hangul (or kana) in a bold run comes out bold.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_015MAPjgJo17KasfbzMTbV65
Where the 512 px cap held the light short of half a level, it dropped to 0
in one pixel: a ring 512 px inside a deep pane (46 levels at depth 300,
splay 1, 200%). The falloff now reaches 0 at the reach. Where the cap
doesn't bite this moves the light by at most half a level.
Cut tests now also cover rounded corners and a look under the cap.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Every hex field (fills, strokes, text ranges, effects, layout grids,
colour variables, page and prototype backgrounds, the colour picker)
now takes any of the 148 CSS named colours. Case, spaces, hyphens and
underscores are ignored; hex wins where both could match; names are
opaque and keep the colour's opacity. Color::parse_css (the picker's
CSS field and command colour params) takes the full list too.
Require resolved font entries to be regular files before indexing them. Preserve normal font symlinks and cover named pipes in a bounded child-process regression.
Assisted-by: OpenAI Codex (GPT-6)
Signed-off-by: IanHollow <[email protected]>
Where nothing bends (past `depth`, or with no refraction) the three
colour channels sample the backdrop at the same point: one sample
instead of three. The rim light's facing and powers are worked out only
where the light reaches. Same bytes: an ignored `glass_digests` test
prints digests of 30 renders (5 looks, square and rounded, 3 scales),
identical before and after this change (and to main).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The glass light falls off with the distance to the edge, read past
`depth` (up to ~3 × depth at full splay), but the coverage window around
a cut layer was only `depth + 2` wide. With frost and refraction at 0
the backdrop margin is 0 too, so a band or canvas patch edge running
through the pane changed the light near it: up to 12 levels between a
banded and a whole render.
The window now covers the light's reach (where it falls under half a
level, clamped at 0 and capped at 512 px) plus the span the normals are
smoothed over, and past that reach the light is 0 instead of whatever
the window happened to hold.
Tests: a cut 50–80 px inside the pane equals the whole render byte for
byte (fails on main by 12 levels); light_reach edge cases. Adds an
ignored release timing test (glass_bench_tests).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
W and H ignored scrubs. They now scale live in one undo step. The factor
field applies on release, because it is relative to 1.0 on every frame.
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Delays, scroll offsets, overlay offsets and the overlay background alpha
were set on every frame of a scrub, an undo step each. They now set once,
when let go, like the custom size field.
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The RGB/HSL/HSB fields committed every frame of a scrub, so one drag left
about ten undo steps, and the alpha field ignored scrubs. Both now edit
live and commit when let go.
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Restore the original cell spacing and check the table geometry across text sizes instead of requiring every font to fit without scrolling.
Assisted-by: OpenAI Codex (GPT-6)
Signed-off-by: IanHollow <[email protected]>
The browser-worker test assumed a fixed 2400x2400 glass scene renders
in at least 12 ms; a faster machine or renderer (the glass speedups)
would make it quick and the test fail. The scene now gets more and
more blurred rectangles until renders from scratch of the views the
test uses take three times the threshold.
Refs #295
A hyperlink longer than 200 characters came back as a different 200-character
URL -- one that still looks like a working link -- and the import reported
nothing. It applied to a text layer's base style and to a character style
override alike.
The destination went through `clip_str`, which exists so that names and labels
from a file stay readable when they are quoted in a warning. Every other caller
is a name, a description, a style id or a layer type; a hyperlink is data, and
clipping it silently rewrote it. Two things say so: the FIG importer already
keeps the URL whole, so the two importers disagreed, and `rest::props` bounds
numbers and list lengths but no string beyond that label helper -- the text
content itself is not clipped either.
The reporter's integration test is included, with its assertions unchanged. Its
helper is restructured to return what it found rather than unwrap: the crate
denies `unwrap_used` and `panic`, and `clippy.toml` allows them only inside a
`#[test]` function, so the original shape failed the gate before it ran. Its
failure message now reports the input and output lengths.
StyleRun start/end are usize taken straight from a .figcraft document.json
(nothing clamps them on load; only the REST and SVG importers normalize),
so TextData::replace_range's p - removed + inserted panics in a debug or
test build and wraps to zero in release, where normalize then drops the
run and the layer silently loses its formatting.
Shift the runs with saturating arithmetic; normalize drops whatever is
left. Regression test: text_runs_past_the_text_never_overflow — panics
at crates/doc/src/text.rs:605 on main, passes with the clamp.
The message now comes from the UI crate (dialogs::window_failed) through
tr() and trf(), with entries in every catalog, and the app picks the
language from the preferences before the event loop starts, so the box
reads in it even though no frame was ever drawn. The commands to type stay
as written.
With drop shadows the backdrop was read twice: before the content and
again after the shadows. The first read is only what backdrop effects
inside the content see, so it now runs only when the layer has such
descendants, and the content is copied for the shadow caster only when it
gets a layer blur, noise or texture. The output is unchanged.
The wrapped hint label duplicated widgets::note (a wrapped secondary note
inside the panel's side padding); the two hints use it, and the copy and
the extra spacing after them go.
Design is already the mode while the pen draws, so re-pressing it leaves
nothing: only the Pencil goes back to Move there, as before. Leaving for
Prototype or Inspect still ends the path.
Where the window has less room the font list is shorter than its tallest,
and stepping down left the highlighted row cut off at the bottom. The
scroll now uses the list's height as last drawn.
A page drag released over the Layers below could still reorder against a
row scrolled out of view, and a layer dropped there could move to that
page. The drop now counts only inside the list's visible area. The guard
is ported from #366 by Skyfuller1303 (@Skyfuller1303).
A toast's clock started from the time of the frame that first drew it. That
time is taken when the frame starts, before whatever raised the toast ran:
the Open dialog and a failed open blocked that frame for longer than 3 s, so
the error expired before it was ever seen and the open failed silently. The
clock now starts on the frame after the first one that draws the toast.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The hub kept the addresses removed from every file it ever had open. The
entry now goes when the file is closed (everyone left) or removed from the
server, so a long-running public server doesn't keep collecting them.
The outer batch passes a command.batch step on as written, so the inner
steps' `$step` references point at the inner batch's steps. The docs say
that `$step` is a reserved key in step params.
Nothing is collected while auto-save is off. The wait restarts whenever an
unsaved file changes, so continuous editing doesn't save every few seconds,
and the file is encoded on a thread of its own, then written.
FileSnapshot takes what saving writes, encodes it (thumbnail and all) on any
thread, and writes it back as of the snapshot: edits since stay unsaved, and
bytes from before another save never overwrite the newer file.
Each frame checks the active document's names for CJK scripts, to install
faces for them. The check was keyed on the revision, which pans and zooms
bump too, so every frame of a pan read every layer name. In a 900 000-layer
file that was ~70 ms a frame: panning ran at ~14 fps whatever the canvas cost.
It now keeps the document version it read: the same version (a pan, a zoom)
is skipped, and a new one reads only the layers that changed since
(`Document::changed_ids`, sized by the edit), or every name past 10 000
changes. On that file the UI frame while panning went from ~71 ms to ~3 ms.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YQxjkNSUJhTkTfY74aNeki
Adds crates/ui-egui/locales/id.toml, a draft Indonesian catalog covering
every UI string (1,577 keys), command label (557) and menu wording (31),
machine-drafted and marked "draft — needs native review".
Indonesian keeps a noun unchanged after a count, so plural entries give
only `other` and plural_one needs no change. Preferences ▸ Language lists
Bahasa Indonesia; System maps id / id_ID to it (new assertion in
i18n_tests.rs) and `id` joins the cjk can-show list. Docs updated.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/engine/src/cmd/page.rs:50
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Adds crates/ui-egui/locales/cs.toml with all 2165 entries (ui, commands,
menu), marked "draft — needs native review" and kept out of COMPLETE.
Plurals use one/other only; the other form is worded count-neutral so it
reads correctly for 2-4 as well. Lists cs in the pref.language help text,
adds a cs_CZ.UTF-8 detection test and updates the localization docs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
On macOS without craft-fonts the default face is Helvetica Neue, whose
kerning comes out of the shaper partly as an x offset (e.g. on the "e" of
"here."), so a glyph's x + advance is not its pen end. The layout was
right (the full stop's pen sits exactly on the edge); the test assumed no
offsets. Each hanging line is now compared with the same line set flush:
every glyph must move by exactly the hanging mark's advance, and the flush
right line must end on the edge.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/engine/src/cmd/arrange.rs:327-330
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Inside command.batch, `{"$step": n}` anywhere in a step's params is
replaced by the `id` in step n's result before the step runs
(`{"$step": n, "field": "..."}` picks another field). A nested tree,
such as a frame with children, now takes one batch and one undo step,
not one round-trip per level with ids copied between calls by hand.
Malformed references (forward or self, negative or fractional index,
missing or non-string field, extra keys) fail the batch with BadParams
naming the step, and the existing group machinery rolls it back. The
params walk is depth-capped. Steps are journalled with resolved params.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Found while testing crates/engine/src/cmd/edit.rs:256 (effect.set), :305 (export.update), :317 (export.remove)
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The colour of new text in the app was chosen from the page alone, so
on a dark page text drawn in a white frame (a new frame's default) was
white and invisible. readable_text_fills now looks at what the text is
drawn on: the nearest layer from its parent out whose top visible fill
is a mostly opaque solid colour, else the page. The Text tool and
text.create / node.create pass the text's parent.
The two tests that asserted the page-only rule in a frame now expect
black text in a white frame; new cases cover dark, unfilled, nested,
faint and stacked fills, and the Text tool.
The web build's Services never wrote to the clipboard, so Cmd+V after
Cmd+C got the paste event with whatever text the clipboard held before
and made a text layer of it instead of pasting the copied layers.
WebServices now writes copies (text and PNG) through egui's output,
which eframe hands to the browser clipboard during the copy event.
The web build has no save dialogs, so Export… ▸ Variables JSON / CSS
custom properties returned silently. Services gains downloads() and
download(name, bytes); the browser hands the file over through a Blob
link, and the variables panel uses it where the platform saves by
downloads.
The Text tool made its text on the press and ignored the drag, so a
drag gave auto-width text like a click. The new text now stays in a
TextBox gesture while pressed: past the drag slop it becomes a box as
wide as the drag, its top-left at the drag's, with a fixed width and a
height that follows its text. A click, or a drag within the slop, still
makes auto-width text.
Found while testing crates/engine/src/cmd/edit.rs:962-977
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/edit.rs:975-981
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
On macOS egui never sends `ctrl` together with `command`; Ctrl comes alone there
and Ctrl+R isn't a shortcut (rename is Cmd+R or F2), so case 1 failed on every
macOS run. The keymap is right; the test now presses Ctrl+R only off macOS.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Up and Down now move a highlight through the family list, starting
from the current family and scrolling to keep it in view. The text
panel previews the highlighted family on the canvas, as it does for a
hovered font style, and Enter applies it. Escape leaves the text as it
was.
opacity_from_digits multiplied and added on every character of its input
string before the one-or-two-digit check, so "9999999999" overflowed u32 on
the way (panic in debug builds, wrap in release). The scan now steps n with
checked_mul/checked_add and stops (None) at the first character the number
cannot take, so a long digit string is refused like any other bad input.
Test: opacity_digits (engine keymap) covers the ten-digit string and the
command's BadParams refusal.
Document::next_name summed max + 1 over numbers parsed from existing layer
names, and bump_name did the same for the duplicate-rename path: a document
(or peer, or node.set) holding a layer named "Frame 18446744073709551615"
overflowed both (panic in debug, wrap to a wrong name in release). next_name
now collects the numbers it sees and, when the counter is spent, picks the
lowest free number (mirrors new_style_id); bump_name only bumps when the
number can step, else falls to its existing " 2" branch.
Tests: next_name_never_overflows (doc) and
duplicating_a_layer_named_past_the_counter_does_not_overflow (engine panel
behaviour).
The welcome's seq and processed.last_batch are the host's word and were used
unchecked: accepted computed self.seq + 1 and queue bumped next_batch, so a
host sending u64::MAX overflowed the client (panic in debug, wrap in release).
Both counters now step with checked_add/saturating_add, and a sequence that
cannot advance is refused as a protocol error (the established rejoin path).
Test: hostile_sequence_numbers_do_not_overflow_the_client.
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 hub closes their connection as the end of the session for
them (close 1001: they keep their copy and don't reconnect) and bars their
address from that file while it runs, so the code or link they hold no
longer gets them in; everyone else stays connected.
Protocol addition within version 2: `Admin::RemovePeer {file, client}` on
an open hosting connection with the file's owner token (or the server
key). Older servers say they can't.
Tests: loopback (removed, can't rejoin, others stay), engine (host can,
guest can't, bad ids refused), codec round trip; the collab junk sweep
covers the new command.
Requests reach the UI through the main window's logic. A request for
another window is answered in that window's frame, and the main window's
frame (`drain`) asks that window to draw. While the main window is
hidden, only its logic runs and nothing asked: an idle window on screen
draws no frames, so the request waited until something else repainted
the window, or timed out after 60 s.
`drain_hidden` now asks a window on screen that a request waits for to
draw. A hidden window isn't asked: it can't draw, and the main window's
logic already answers what it can.
The Small caps button in Type settings labelled itself "Aɢ", but the
small-capital G (U+0262) is missing from egui's default fonts, which the
UI uses whenever Inter isn't built in, so the button showed "A" and a box.
The sample is now a capital followed by smaller capitals laid out from
ordinary letters every UI font has. A test checks the UI fonts draw every
letter of the sample.
`drain_hidden` answered only requests for the main window, and only
while that window was hidden. A request addressed to another window
(`window.new`, `window` param) that is minimised or covered still waited
for that window's frame and timed out after 60 s, whether the main
window was on screen or not. An agent that works in its own window lost
it as soon as the window went behind others or to another Space.
`drain_hidden` now looks at the target window of each request: if that
window is hidden (`ViewportInfo::visible() == Some(false)` for its
viewport) and the method needs no frame, it is answered from the main
window's logic with the target window's state swapped in, as while it
draws, so it acts on that window's file. `draw` and the new
`windows::as_window` share the swap (`enter`, `adopt`, `leave`). The
main window's logic runs on every request (the control server asks for
a repaint), so this works while the main window is on screen too. What
needs a frame keeps waiting, as before; a request for a window that
doesn't exist is an error at once instead of waiting for a frame.
Recovery copies were written with plain `save_bytes`, so a file
recovered after a crash came back without its version history and
branches, even though `file.recover` reads both. Copies now go through
`save_bytes_with_history` like a normal save, with each open branch
tab's latest edits (`branches::current`), since `sync` only puts those
in before the next command.
A copy was also only redone when the main file's revision changed, and
a file with edits only in a branch tab didn't count as changed. Both
checks now include the file's branch tabs.
FigCraft's RPM declares no glibc floor, though the deb has `libc6 (>= 2.35)`. On Rocky Linux 9
(glibc 2.34) the RPM installs, and then `figcraft-cli` fails with ``/lib64/libm.so.6: version
`GLIBC_2.35' not found``.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The first-run test settled the home screen with a fixed 12 frames, but
the recent files are looked at on a thread of their own and a card shows
no name until that thread has answered. On a busy CI runner it hadn't
yet, so "ridgeline" was still a grey placeholder and the test failed
(seen on this PR and on codex/issue-250-async-dialog). settle() now waits
until every recent file has been looked at, as home_recents_tests does.
Checked with a reader slowed to 200 ms: the old test fails at the same
line as CI, the new one passes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Recent files are cards since #249, so the row test now opens two
documents with synthetic paths and checks the Open files list: the row
that fits shows name and path in full; the long one shows the name cut at
its end, then the path cut at its start, at least 12 px apart, and both in
full on hover. Without the row fix the name shows as "Quar…" under the
path.
The Open files and Recent files rows painted the detail (the file's path)
right-aligned at its full width and only then elided the name into
whatever was left, with a floor of 40 px. A path wider than the row ran
straight over the name, and the name itself shrank to "Mae…".
The row is now laid out name first: the name keeps its full width when
both fit, otherwise at least 60% of the row; the detail gets the room
left after a 16 px gap and is cut at its start ("…/02-technik/KNX/
file.figcraft"), which keeps the file and its nearest folders. A detail
with less than 48 px of room is left out. Rows of Recents' own lists show
name and path in full on hover when either was cut.
The new widgets::elide_start_font mirrors elide_font, keeping the end.
With no points picked, the Design panel's hint ("Select points to edit
their position, radius and mirroring") didn't wrap and made the canvas
49 points wider, so entering vector edit mode moved the view by about 25
points and picking a region moved it back. The jump came in the middle
of the click that picked the region: with the smallest pointer movement
the release landed elsewhere on the page and the region was dragged by
that distance (about 47 units at 70%). Both edit mode hints now wrap.
Regression test: the canvas rectangle stays the same entering edit mode,
picking a region and switching to the width tool.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The recent files' dates were only read again when the home screen came
back into view, so a file deleted or moved in the meantime kept its card
until then. They are now also read again every 2 seconds while in view,
and at once when the window gets the focus back; a file that is gone
shows "File not found" without leaving the home screen.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review of the zoom previews found three problems:
- The browser worker never previewed: wasm builds timed every render
as 0 ms, below the 12 ms threshold. Renders are now timed with
web-time's monotonic clock on every target (already in the tree via
the engine), and the inline worker is built everywhere so a test runs
it with the production threshold: preview one frame, refine the next,
a newer request skips the old refinement.
- A preview was refused whenever the texture on screen was refined at
the same zoom, even when that texture showed another part of the page
(back at 100% about another point after 800%: 1/64 of the view). It
now has to show the whole view asked for, not just its zoom.
- The latencies read the frame's start time before the inline render
and before the texture was set. They now run on the same clock from
the request to the image handed to egui as the canvas texture.
Refs #295
A layer with a background blur (or glass) drew its blurred backdrop into
its own pixels before casting its drop shadows from them, so a translucent
fill cast a shadow as dark as an opaque one, and with "show behind
transparent areas" the shadow was cast from the backdrop's pixels. The
backdrop was also read before the shadows were drawn, so a shadow showing
behind was never seen through the blur.
When a layer has both, it now casts its shadows from what it paints (with
the layer blur, noise and texture it will get), draws them, and only then
reads the backdrop, which goes under its content; the content still sees
the first backdrop under it while it's drawn. Layers without both draw as
before.
A minimised or covered main window draws no frames, and control requests
were only answered inside a frame, so every control-channel and attached
MCP call (save_file among them) timed out after 60 s while the app was out
of sight. eframe still calls the app's logic then whenever a repaint is
requested, which the control server already does for each request.
The desktop app's logic now answers, while the window is hidden, the
requests that need no frame: engine.execute, engine.commands,
document.inspect, node.get, window.list and ui.render. Screenshots, input
and the interface's own state still wait for the window, and their
timeout now says that the window has to be on screen.
Closes#269
Creating a collection (and adding a mode or a variable, double-clicking
a row, or Rename in its menu) asked egui to focus the new name's field in
the same frame, but that field is only drawn from the next frame on. For
one frame the AccessKit update named a focused node that wasn't in the
tree, and the platform adapter aborts on that, outside the app's frame
guard; Windows keeps AccessKit on through UI Automation far more often,
hence the silent close there. A rename now leaves a flag and its field
takes the focus in the frame it's drawn, through widgets::rename_field,
as the layer rows do.
A new zoom can't reuse the retained canvas image, and the previous one,
re-projected, covers little of the view after a zoom out (1/64 of it
from 800% to 100%), so the page background showed until the full render
finished.
When renders from scratch have been slow (a frame or more), the worker
first renders the whole requested view with a quarter of the pixels
each way, in a Retained of its own so pans and edits still build on the
full image, then refines it unless a newer request is already waiting
(that one gets its own preview). The web build renders the preview and
the refined image in consecutive frames. Pixel preview gets none.
Finished images carry the file, page and whether they are coarse: the
canvas never shows an image of another file or page, and a late preview
never replaces a refined image already at the zoom asked for. ui.inspect
reports the canvas quality and the latency from request to the frame
showing the preview and the refined image.
Refs #295
page.add gave every added page the interface theme's colour. Figma's
documentation says added pages inherit the current page's background;
the theme only decides a new file's page. Read the active page's colour
before creating the new page, in the command, so the app, CLI, control
channel and MCP agree.
Refs #295
"Placeholder rectangles" was a claim about Apple Color Emoji's glyf
outlines that no source I can show supports: the font's sbix chapter
documents that most emoji are bitmaps while the flags are drawn from
outlines, and it is Noto Color Emoji that has no outlines at all. The
comments, the ROADMAP row and the parity row now say that instead.
No behaviour change: comments and docs only.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <[email protected]>
Preferences > Save files automatically (pref.autoSave, off by default;
hidden in the browser, which has no files to save to). With it on, each
open file that has a path is saved 3 s after it changes, never during a
gesture, a live text preview or while a version is shown. A failed save
says why in a toast and tries again a minute later.
Version history checkpoints (version.autosave, every 30 minutes) only ran
for files with unsaved changes, so a file that saves itself would stop
getting them; with auto-save on, edits made since the file was opened
count.
Closes#296
The placement formula is Skia's own fontations port
(src/ports/fontations/src/bitmap.rs) reading the sbix spec's
lower-left-corner rule; the earlier comment named CoreText, which that
source does not mention. Apple Color Emoji measures ~180 MB on current
macOS (the 190 MB figure is a hosted copy of the file), so the comments
say ~180 MB.
No behaviour change: comments only.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <[email protected]>
Release builds on Windows have no console, so an eframe::run_native error
(no usable graphics adapter, a device the driver refuses) went to an unseen
standard error and the app quit silently. The error is now logged and shown
in a message box with a hint: unset WGPU_BACKEND when it is set, otherwise
update the driver or try another graphics API (OpenGL on Windows), and
where to report it. With --control the box is skipped so agents aren't
blocked.
Colour emoji came out as the placeholder rectangles their outlines carry
(Apple Color Emoji) or as .notdef when the face has no outline at all:
nothing read sbix, CBDT/EBDT or COLR. A glyph now asks its face for
colour ink before its outline: an embedded bitmap strike decoded to
premultiplied RGBA and placed by the strike's bearings the way Skia (and
CoreText) places it, scaled to the size the run asks for, or a COLR v0
glyph's layers filled from the font's palette.
Faces that carry only colour tables are indexed again: Apple Color Emoji
is no longer skipped for being 190 MB (the scan still reads headers and
name tables only), and "Segoe UI Emoji" and "Noto Color Emoji" join
"Apple Color Emoji" as colour fallbacks, read only for a text that
nothing else covers. The png crate decodes the strikes; skrifa keeps
shaping, metrics and outlines.
Left for later: COLR v1 paint graphs, alpha-only strikes and the SVG and
PDF exports (text is outlines there).
Tests: 8 cases in the text crate (placement, PNG and raw-BGRA decoding,
scaling, COLR layers and bitmaps of the machine's own colour font) and a
render case gated on that font; the font-free ones run everywhere.
Verified on Windows with Segoe UI Emoji (COLR) and with Noto Color Emoji
(CBDT) read from a downloaded file; the sbix path runs through the same
code and needs a macOS run to be seen.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <[email protected]>
Windows' preferred-language list is also the input-language list: a language
added for a keyboard sits in it, and taking the first entry with a catalog
turned the interface German on a Polish system (a German keyboard) while the
system and the other apps stayed Polish. Only the first preferred language now
decides; when it has no catalog the interface reads English. macOS'
AppleLanguages and gettext's `LANGUAGE` still take the first language with a
catalog. Fixes#291.
Generated with Codebuff 🤖
Co-Authored-By: Codebuff <[email protected]>
The corner-radius handle sits 10 px in from each corner and answers within 6 px, so a press on the inner half of the 8 px corner square picked it: the drag rounded the corners or, dragged outward, did nothing. Inside the drawn square the corner now resizes both sides (in ratio with Shift or the constrain-proportions link); further in the radius handle still answers.
Saving and exporting wrote a truncating std::fs::write to a deterministic
staging name (<path>.tmp-figcraft, <file>.figcraft.tmp) before renaming, so an
unrelated file or a planted symlink at that name was clobbered or written
through, and separate writers of one destination shared a staging file they
could corrupt. Both paths now go through figcraft_format::write_file_atomic: a
uniquely named sibling staging file is created exclusively (create_new) and
written through its own handle, renamed over the destination, and removed again
on any failure, so only the staging file a write created is ever touched. Tests
cover pre-existing staging files and symlinks, concurrent writers, and
injected rename and permission failures.
- A copy Alt-dragged off a layer snaps to that layer too: its edges,
centres and equal spacing (Targets::add_box with the originals' box).
Before, the originals were left out of the snap targets, so a copy
couldn't line up with or space evenly from the layer it came from.
- Duplicate (Cmd+D) repeats the last duplicate's page offset while its
copies are what it duplicates: Alt-drag a copy, or Duplicate and move
the copy, then each Cmd+D steps out the next one as far again. The
step is measured when Duplicate runs. It lasts as long as the
selection (undo reuses ids), copies placed by an auto layout set no
step, and explicit dx/dy win.
- measure_tests: an Alt-dragged copy between its original and another
layer now snaps to equal spacing (45/45 instead of 50/40).
Tests: duplicate_spacing_tests (snap to the original, equal spacing,
repeat after moving a copy and after an Alt-drag, only for the last
copies, undo then a new layer, rotated frame, auto layout).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
A command that fails part-way (node.set applies geometry keys before
values like visible, which can fail validation) mutated the document
even when an interaction was under way, because the rollback in
Session::execute was gated on interaction.is_none(); drop the gate so
the command-entry snapshot is restored during interactions as well,
which is the interaction's last valid preview and leaves the gesture
itself intact.
A collection a saved file references by its asset key (the form library
collections keep) instead of a guid resolved to nothing, so every
explicitVariableModes entry was dropped silently and bound layers rendered
in the collection's default mode (#267). The .fig import now resolves such
references against the file's own nodes (the collection node carries the
same key) for mode entries and the variables' collection link, and reports
modes of collections it can't resolve.
The Pen showed only its anchors while a path was drawn; handles appeared
only in vector edit mode afterwards. Now:
- The point being dragged shows its in and out handles, and the point
before it its out handle (the handles shaping the segments at the pen),
drawn like vector edit mode's (one shared canvas helper).
- Moving on to the next point, the rubber band is the segment a click
would make: it leaves the last anchor along its out handle instead of
going straight.
- A press no longer leaves the previous straight rubber band on screen
over the segment being dragged.
Overlay gains pen_handles and rubber_handle (shown in ui.inspect).
Tests: pen_overlay_tests (handles during and after a drag, none and a
straight band for corner points, cleared on finish); canvas_fixes_tests
(handle lines and dots painted mid-drag, curved band after release).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
One layer directly on the page has nothing to align to (one layer aligns
to its parent frame), so the six Alignment buttons looked active and did
nothing. They are now disabled there, with a hover explaining why.
- cmd/arrange.rs: align_target (unchanged rule) and pub can_align.
- right.rs: the align rows in add_enabled_ui + on_disabled_hover_text.
- The hover text in every catalog.
Tests: a_layer_alone_on_the_page_has_nothing_to_align_to (engine),
the_align_buttons_are_off_for_a_layer_alone_on_the_page (AccessKit).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The colour picker's paint type row borrowed unrelated icons for the
gradient kinds (auto layout's bars, a circle, the rotate arrow, the
instance diamond). Each kind now has its own icon: the Solid swatch as
an outline with that kind's fade inside (linear ramp, radial, a conic
sweep from the top, diamond), drawn with a new per-vertex-alpha mesh
helper.
Test: gradient_icons_fade (each paints a mesh from opaque to clear).
Refs #256
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Drops the draft status from tr.toml (the Preferences window no longer
calls it a draft) and marks Turkish reviewed in ROADMAP.md,
docs/localization-parity.md and docs/translating.md. Turkish stays out of
COMPLETE, so new strings don't have to be translated into it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A slice paints nothing itself, so thumbnail() rendered an empty pixmap of the slice's own box: expanding Preview for a slice showed only the transparency checkerboard while exporting the same slice wrote the artwork correctly.
Give the preview the export's slice rule (drawn_layers + export_area in cmd/file.rs): draw every top-level layer of the slice's page over its world bounds. Regression tests cover a slice over artwork, a slice with nothing under it, and a plain layer.
Signed-off-by: Ahmed Mahmoud <[email protected]>
crates/ui-egui/locales/tr.toml translates every UI string (1,577 keys),
command label (557) and menu wording (31), machine-drafted and marked
"draft — needs native review". Buttons and menu items use bare
imperatives (Kaydet, Sil), instructions the polite -in/-ın (seçin,
tıklayın); the percent sign goes before the number (%50). Turkish keeps
a noun singular after a count, so plural entries give only `other`.
Preferences ▸ Language lists Türkçe; System maps tr and tr-TR to it.
The catalog passes the i18n check even when listed as complete.
docs/translating.md, docs/localization-parity.md and ROADMAP.md list
the new draft. Adds my own contributors/people.toml entry.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
`{"type":"linearGradient","stops":["#hex",…]}` was only expanded by the MCP
tools, so `figcraft-cli run` and the control channel's `engine.execute`
refused it ("`fills` must be paints or colours"). The expansion now lives in
the engine (`params::normalize_paint`, used by `paint_of`), so node.set,
node.create props, fill.add/set, strokes and paint styles all take it; MCP
calls the engine's copy instead of its own.
Tests: engine (every paint command, MCP-equal defaults, junk still refused),
control channel `engine.execute`, and `figcraft-cli run` end to end.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
tar -tzf dist/release/figcraft-*-freebsd-x86_64.tar.gz >"$HOME/freebsd-contents.txt"
for f in '/bin/figcraft$' '/share/applications/ai\.storyteller\.figcraft\.desktop$' '/share/doc/figcraft/LICENSE-MIT$'; do
grep -E "$f" "$HOME/freebsd-contents.txt"
done
# Intentionally tracked at @main (not SHA-pinned) so org-wide CI improvements, security fixes, and build optimizations propagate without PR churn across all craftapp repos.
# Intentionally tracked at @main (not SHA-pinned) so org-wide CI improvements, security fixes, and build optimizations propagate without PR churn across all craftapp repos.
assert m["id"] == "ai.storyteller.figcraft" and m["command"] == "figcraft", m["id"]
for key in ("runtime", "runtime-version", "sdk", "finish-args"):
assert src[key] == bundle[key], f"Flatpak manifests disagree on {key}: {src[key]} vs {bundle[key]}"
print("Flatpak manifests ok")
EOF
# Intentionally tracked at @main (not SHA-pinned) so org-wide CI improvements, security fixes, and build optimizations propagate without PR churn across all craftapp repos.
run:cargo run --locked -q -p xtask -- version set "$FIGCRAFT_VERSION"
- name:Build and package (tar.gz)
run:packaging/server/package.sh
- name:Check the archive
run:|
tar -tzf "dist/release/figcraft-server-$FIGCRAFT_VERSION-linux-${{ matrix.arch }}.tar.gz" | tee "$RUNNER_TEMP/server-contents.txt"
for f in '/figcraft-server$' '/figcraft-server\.service$' '/Caddyfile$' '/nginx\.conf$' '/install\.sh$' '/figcraft-backup\.timer$' '/server\.md$' '/LICENSE-MIT$'; do
grep -E "$f" "$RUNNER_TEMP/server-contents.txt"
done
ldd target/release/figcraft-server
- uses:actions/upload-artifact@v4
with:
name:release-server-linux-${{ matrix.arch }}
path:dist/release/*.tar.gz
if-no-files-found:error
flatpak:
timeout-minutes:90
name:Flatpak ${{ matrix.arch }}
needs:[version, linux]
# Repackages the Linux job's tarball (the same binaries as the AppImage/deb/rpm); no Rust
# build here. 24.04 for a newer flatpak-builder (1.4) than 22.04 ships.
runs-on:${{ matrix.runner }}
# No `environment: release`: this job signs nothing, so it gets no release secrets.
echo "Draft release **$title** at \`${GITHUB_SHA::9}\`:"
echo
sed 's/^/ /' SHA256SUMS.txt
} >>"$GITHUB_STEP_SUMMARY"
# Intentionally tracked at @main (not SHA-pinned) so org-wide CI improvements, security fixes, and build optimizations propagate without PR churn across all craftapp repos.
if (Test-Path (Join-Path $dir 'figcraft.exe')) { throw 'uninstall left figcraft.exe behind' }
Write-Output 'ok uninstalled'
windows-arm64:
# Intentionally tracked at @main (not SHA-pinned) so org-wide CI improvements, security fixes, and build optimizations propagate without PR churn across all craftapp repos.
@@ -25,6 +25,7 @@ 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 |
| 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, Source Sans 3, Source Serif 4, 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) |
<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 +20,9 @@
<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 +65,7 @@
<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>
@@ -227,8 +230,8 @@ 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 and Spanish (a draft waiting for a native speaker's review), 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 two 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 and Georgian 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).
- **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 +265,7 @@ 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 +351,6 @@ 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:** 2026-10-14 · **Change:** minor (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:** 2026-10-17 · **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)
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
@@ -13,7 +13,7 @@ start with [`docs/gaps.md`](docs/gaps.md).
| Indonesian | id | 99% | draft (needs native review) |
| Japanese | ja | 99% | draft (needs native review) |
| German | de | 96% | draft (needs native review) |
| Korean | ko | 95% | draft (needs native review) |
| Vietnamese | vi | 0% | none |
No other languages. Figma ships English, Japanese, Korean, French, German, Spanish and Portuguese
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 and Simplified Chinese drafts (Preferences ▸
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,
IME) already works (macOS IME composition fixed for #747, not yet checked on a Mac). Detail: [`docs/localization-parity.md`](docs/localization-parity.md).
## Upcoming
@@ -131,6 +131,10 @@ Ranked; detail and milestones in [`docs/roadmap.md`](docs/roadmap.md).
## Progress log
- **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.
- **2026-10-14** · 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.
- **2026-10-14** · SVG/PDF export depth (M0.9, #208): visible descendants deeper than 256 levels below an exported root are refused before bounds or drawing, including masks and raster fallbacks. SVG has a checked API; the engine and CLI report the limit and suggest selecting an inner layer. Existing destination files survive refusal. Boundary, inner-root, hidden-subtree, multi-root/page and command/CLI regressions cover the change. Numbers unchanged.
@@ -312,6 +316,23 @@ Ranked; detail and milestones in [`docs/roadmap.md`](docs/roadmap.md).
| Date | Change | Summary |
|---|---|---|
| 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 | 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 | 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. |
| 2026-10-14 | minor | Lost images reported on loading (#207): a `.figcraft` file whose image member is damaged, too large or missing no longer loads in silence; each lost image (up to 20, then a count) is listed in `repairs` on `file.open` with its hash and the reason, and the app shows it as a toast. Saving still writes the file without the lost images. Numbers unchanged |
| 2026-10-14 | minor | SVG import budgets (#190): gzipped SVG is unpacked by FigCraft within 64 MiB before parsing (it was unpacked without a limit), so the nesting check sees the real document; SVG documents embedded in `<image>` data URIs share one 32 MiB budget and at most 4 levels, and one past a budget is left out. Linked files stay unread. Numbers unchanged |
"max_file_mb is {}: the app opens files of up to {} MiB, so a file that large can't be opened",
self.file_mb(),
client_max-1
));
}
if!self.open_publish{
returnout;
}
ifself.file_mb()>64{
out.push(format!(
"open_publish is on with max_file_mb = {}: anyone may send messages that large, and the memory set aside for reading and queueing them grows to eight times it; keep it at 64 or less on a public server",
self.file_mb()
));
}
ifself.operator_contact.trim().is_empty(){
out.push("open_publish is on but operator_contact is empty: people can't reach you about abuse; set it (and terms_url)".into());
}
@@ -168,7 +182,7 @@ impl Config {
returnErr("`idle_days` must be from 0 to 3650".into());
json!({"id": id,"ok": false,"error": "timed out waiting for the UI (is the window visible?)"})
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?)"})
Loaded 100 of 479 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.
Dylan M. Taylor