request, execute, commands and screenshot move onto the shared crate (storytold/craft-libs#413): a timeout on every call, one data field for the screenshot's PNG, and docInspect and uiInspect beside them. A call the page gave up on cancels its request if the app has not started it, so the app skips it. openFile, captureDownloads, files, readFile and inspect stay FigCraft's own, on the same object.
About > Contributors still showed an old snapshot, because nothing
regenerated contributors/contributors.json before a release. Opt in to
the shared refresh from craft-common-actions: a read-only credits job
regenerates the snapshot for the release commit, and the shared release
workflow installs it before every source build. The snapshot is not
committed back.
A session hosted with `bind` 0.0.0.0 (or ::) kept listening on every
interface but put 0.0.0.0 in its edit and view links, URLs, join codes
and shown address, which nobody can connect to. They now name this
machine's local network address of the listener's family (loopback
without one), on the same port.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6DxbDYxuPEA2pJirCLQ
Splay bends away from the centre of the layer's bounds, as the reference
tool's does: on a ring or an L that centre lies outside the shape, so an
inner edge bends towards the hole. Checked side by side on a ring and an
L at Splay 0% and 100%. The module doc still said splay 0 bends along the
edge's normal; it bends towards the core since #1889.
`MAX_SHAPE_PX` now says it is 4M px on the desktop and 1M on the web, and
that a bigger shape gets no core, so a large glass text or vector can bend
differently once zoomed into, in a 2× export, or on the web.
Docs only.
Review follow-up: `effects/x/radius`, `effects/0` and `effects` name no
effect's field, but `list_binding` doesn't parse them, so the layer's
check took them for some other field and kept an inert binding. A key
under `effects` that names no effect's field is now refused as well.
The dispersion window samples the backdrop 18 times a pixel wherever the
curve bends it, even where the move is so short that the channels split
by a sliver of a pixel. Under 1/256 px every step lies within that of
the moved point, so one sample is within one level (of 255) of the window
over any backdrop, the hardest being 1 px columns of black and white or
of white and nothing. Those pixels now take one sample.
A larger split doesn't qualify: at 1/4 px one sample would move red and
blue by 1/8 px, up to 32 levels across a hard edge.
While a press that went down on the canvas is held, the view centre now follows the canvas rect's own move as well as its size change, so a still pointer keeps its page point when the left sidebar or rail appears mid-press.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
`Glass::set_bound` clamped a bound light angle to ±180, so a variable of
270 drew the light from 180, a quarter turn off, while a typed 270 draws
from -90. A bound angle now turns round into -180–180 (270 is -90, -0 is
0), the same direction the angle draws in.
`variable.bind` on a layer checked only the variable's type against the
field's scope, so `effects/0/refraction` on a drop shadow, or a binding on
an effect the layer doesn't have (`effects/9/radius`), was kept and did
nothing. An effect style already refuses both. The layer now uses the same
check: effect `i` must exist and have the field, and the field's type
comes from that effect. A binding an older file has on such a field still
does nothing, and still unbinds.
egui lays an area out inside the size it had on the previous frame, and a
scroll area shrinks to fit its content, so once a search or a short list
had made the style or variable picker small, it stayed small, also on the
next open. Both now offer their content the screen's height on every
frame. Their lists still stop at their own caps, and placement keeps them
on screen.
The previous fix made a local share's `link` follow the code shown, but its
top-level `url`, the session's WebSocket URL, still carried the edit token:
`collab.join {link: url}` from a view-only share joined with edit access and
the edit reached the host. The link layer now hands over the view way in as a
link and its URL (`SessionInfo::view`, a `WayIn`, in place of `view_link`), and
`collab.peers` gives a local share's `link` and `url` from the way in for the
code shown (`Live::invite`). Online shares are unchanged.
Regression test: `code_tests::a_local_view_share_s_link_and_url_only_let_people_look`
joins through the `link` and then the `url` over loopback; both are welcomed as
viewers whose edits never reach the host.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2VLhC9d5v8VSsf4ng
The view keeps its centre when the canvas is resized, so the right panel appearing (minimized UI, on the press that selects) shifted the pointer's page point by half the width over the zoom and the move handling took it as a drag. While a button that went down on the canvas is held, a resize now keeps the top-left instead.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
`collab.share {via: "local", access: "view"}` showed the view-only code but
gave the session's own invite link as `link`, and that link edits: whoever
joined through it could change the host's file. A local share's `link`, top
level and in `sessions`, is now the one that goes with the code shown, and it
follows `collab.share {access}`. Online shares keep the owner's own link there,
as before.
Regression test: `code_tests::a_local_view_share_s_link_only_lets_people_look`
joins through that link over loopback and is welcomed as a viewer whose edit
never reaches the host.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2VLhC9d5v8VSsf4ng
The session the app hosts for a local network share now has two secrets:
the edit code and a view-only code (same address and port, a 44-bit secret
of its own). Whoever joins with the view code is welcomed as a viewer: they
see every change and presence, the app makes the file read-only for them,
and the session refuses their batches, as a server does for view links.
Share > Local network offers Can edit / Can view before sharing starts (the
edit code then never shows) and while sharing. collab.share {via: "local",
access: "view"} does the same; share in collab.peers lists both codes and
links. There is still no comment code on the local network.
Next::Stop boxes its event: SessionInfo grew past clippy's
large_enum_variant threshold, and it only ever carries a status or an end.
Tests: loopback_tests::a_view_token_lets_people_look_and_not_edit,
code_tests::a_local_view_code_lets_people_look_and_not_edit,
live_share_tests::a_local_share_can_start_for_looking_only, and the local
share tests that asserted edit-only codes. Docs: collab-protocol.md,
collaboration-parity.md.
Edit ▸ Set default properties put the selected layer's properties into the Paste properties
clipboard and did nothing else: a rectangle drawn afterwards was still the default grey, and what
Copy properties had copied was replaced.
The defaults are now kept apart (`Session::default_props`: one set for shapes, set from any layer
but text, and one for text). Rectangles, ellipses, polygons and stars drawn or clicked with their
tools take the shape defaults: fills, strokes and their settings, effects, opacity, blend and
corners (only where the shape has corners), through `apply_copied_props`, the rules Paste
properties already used (now shared). New text from the Text tool takes a text layer's fills and
font and is laid out at that size. Frames, sections, slices, lines, arrows, vectors and layers made
by commands keep their own look; the help center is silent on those. `edit.setDefaultProperties`
answers `{for: shapes|text, properties}`.
Tests: `default_props_tests::new_shapes_drawn_with_the_tools_take_the_default_properties`,
`setting_defaults_leaves_the_copied_properties_alone`,
`new_text_takes_a_text_layers_defaults_and_shapes_keep_theirs`. Docs: ui-parity.md, ROADMAP.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2LxvJ6MuFmiLA6Zqm
`proto.settings` took any string as the device ("Pixel Nonexistent", "iphone 16") and saved it in
the file: the presenter then drew no device frame while its picker showed the name as if it were
one. An unknown colour was already refused with the device's colours, but a typo in the device
went unnoticed, and an agent couldn't read the settings back (`proto.settings {}` answered
`{ok: true}`, and neither `inspect_document` nor `get_node` shows a page's prototype settings).
An unknown name, or a value that isn't a string, is now refused with every name the picker offers
("None", "Custom size" and the devices), and nothing changes. A known name is taken in any case and
with spaces around it, and stored as the picker spells it (`known_device`). Every call answers the
page's settings as they are after it (`settings_json`): device, background, rotation, the colour
shown, the screen as presented (null without a device), the custom size and responsive, beside the
`ok: true` it answered before; `proto.settings {}` reads them.
Test: `proto_present_tests::unknown_devices_are_refused_and_settings_read_back`. Docs: mcp.md,
prototyping-parity.md, ROADMAP.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2LxvJ6MuFmiLA6Zqm
Closing a tab whose file has unsaved changes, by its X, a middle click, File ▸ Close tab or its
shortcut, only showed a toast ("save first, or close again with Alt to discard"; from the menu
"close the tab with Alt-click") and left the tab open, so closing a tab seemed not to work. A
`file.close` refused for unsaved changes asked the UI with `dialog.unsaved`, which toasted too.
Closing such a tab now asks the unsaved-changes question the window asks when it closes
(`windows::unsaved_question`, shared): Don't Save closes the tab and drops the changes, Cancel or
Escape keeps it, Save saves the file next frame and then closes the tab. A file without a path, or
saved as .fig, asks where with Save as first; the tab closes once that saved it and stays open if
it is cancelled or fails. `UiState::close_tab` (`windows::TabClose`: Ask, Save, AfterSaveAs) keeps
where it is; `ask_close_tab` starts it for the tab strip, the menu and `dialog.unsaved`. Alt+close
still drops the changes without asking. The three toasts are gone from the code and the catalogs.
Reordering tabs by dragging, the issue's other half, already works on main (landed 2026-10-19,
`dragging_a_tab_along_the_bar_reorders_it_without_detaching`).
Tests: `titlebar_tests::closing_a_tab_with_unsaved_changes_asks` (the X asks, Escape keeps the tab,
Don't Save closes it, Alt+click closes without asking, File ▸ Close tab asks),
`titlebar_tests::save_in_the_tab_question_saves_then_closes` (Enter saves then closes; no path:
the tab stays). The title bar test harness can hold modifier keys. Docs: ui-parity.md, ROADMAP.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2LxvJ6MuFmiLA6Zqm
Under Wayland an app can't tell that its window is minimised (no `minimized`, no `occluded`),
and the compositor stops asking it to draw. eframe keeps running the app's logic for a window
whose redraw is overdue, but `control::background` still left every request for "the window's
frame", which never came: with the window minimised on GNOME or KDE Plasma each control-channel
or attached-MCP request waited and timed out after 60 s ("it was not applied").
`control::background` now treats a window that drew no frame for `NOT_DRAWING_AFTER` (0.2 s)
while a request waited for it as hidden: requests that act on the app's state are answered from
the logic passes, as they are for a window the platform reports minimised. `FigApp::control_waits`
keeps, per window, its pass count and the time the wait began; a frame drawn since starts the
wait again, and a pass is asked for 0.2 s later while a request waits. Screenshots and `ui.input`
still wait for a drawn frame. Measured on GNOME Wayland with the window minimised by its own
button: 0.54 s a request (before: 13.7 s for one, then 60 s timeouts).
Test: `idle_tests::a_window_that_stopped_drawing_answers_from_its_logic_passes`. Docs:
control-protocol.md, mcp.md, hardware-parity.md, target-app-parity.md, ROADMAP.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2LxvJ6MuFmiLA6Zqm
A padding changed on both sides of a branch gave a conflict for every size and place auto
layout worked out from it: the component's, each instance's and their hugging parents' width
and height, each instance's layout, and every child's position (10 in a file with one instance,
138 with 15). The merge was refused until each had a side, yet the merged file laid itself out
again from the padding whatever they got. A `layout` conflict also took all auto layout fields
together, so a gap changed only on the branch was lost when main won the padding.
`merge::plan` now leaves out properties both sides changed to values both files work out for
themselves (`Plan::derived`): a hugging auto layout frame's or a filling child's size, text
sizing to its characters, an auto layout child's place (when both sides turn and scale it the
same way), and a value a layer in an instance has as its main's layer has it. Merging keeps the
file merged into's own value and normalising lays it out from the merged inputs.
`layout` and `childLayout` merge field by field: a field changed on one side (or the same way on
both) merges, and each field changed both ways is a conflict of its own (`Key::Field`,
`<id>/layout.padding`), written whole from the merged fields and each picked side's value. The
review shows it as "auto layout padding" with both sides' values ("8, 8, 8, 8").
Tests: `branch_tests::one_conflicting_auto_layout_edit_is_one_conflict` (the issue's file: one
conflict, either side laid out as if set directly), `auto_layout_fields_changed_one_way_survive_a_conflict`
(merge and update from main), `sizes_auto_layout_works_out_never_conflict`; the merge fuzz test
leaves worked-out values out of what each side must show. Docs: collaboration-parity.md, mcp.md,
ROADMAP.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2LxvJ6MuFmiLA6Zqm
plan/ is an optional notebook, the calendar exception and the Rust-only line now live in the Rust rules, and the quality gates no longer say to push to main. The Figma iconography ban stays. CLAUDE.md imports AGENTS.md.
A file imported from another app can hold library copies in its library
store with no record of the library they came from. On an instance of
one, the command took the local path and failed in selecting it with
"isn't on a page (it is a library component)", which says nothing about
what to do. It now says the copy has no link to its library and that
enabling the library reaches it.
Found on a real imported file, where copies of the same library set were
cached more than once and only one of them could be linked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
with_pixel_size read only the PNG's header, so a PNG with a whole IHDR and
a cut-off body passed as the one to paste. Since #1724 image.place refuses
it as damaged, and nothing was pasted: the decoded clipboard image, which
paste falls back to for an unreadable PNG, was never tried. The PNG is now
decoded in full before it is accepted, so the paste falls back as it does
for any other unreadable PNG. The zero-size and pixel-count checks run
ahead of that decode, so an image past the limit is refused before it is
decoded.
Raised by CodeRabbit on #1360 after the merge.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On an instance of a library component the command refused with "open
that file to edit it". The help center describes it opening the file that
holds the main ("Edit main components"), so it now does: the linked
library file opens in a tab, or its open tab comes to the front, on the
main's page with the main selected and zoomed to.
The library keeps the file's layer ids, so the copy's id map
(ImportedComponent::ids) leads from the local copy back to the exact
layer: a nested instance reaches its own main and a variant that variant.
The new tab takes the canvas size of the tab it came from. A tab still at
the default 1200x800 is refitted to the whole page on its first frame,
which would replace the zoom to the main.
A main deleted from the library file since it was published leaves the
file open with a note. A moved library file, a team library on a server
and a path off this computer are refused before anything opens; the last
uses the same check that guards library reads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
menu_icon_item drew the check mark as a text prefix, so a checked entry's label sat about 14 px left of the others, and Flatten (no icon) started further left still. The check now has its own fixed column ahead of the icon slot, and Flatten uses the same row with an empty icon slot.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A press on a connection handle or noodle returns from the canvas input before the ordinary press reset, so the pick (and keyboard box) stayed up. Clear them on that press too.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The previous commit saved the file with CRLF endings; only the 30 changed lines are a real change.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
bflatastic