The face's own style (Figtree's Light) is its default instance, so it
resolves without axis values and lays out as before, on the static path.
docs/agents.md says how named instances are selected.
In the browser editor a .jsx script couldn't open any file, even one the
page holds next to the project: File.open("r") returned false with "File
not found" and File.exists was false. The footage check after File > Open
also marked imported media as missing.
Both asked whether a file exists on the local disk (std::fs), which has no
files on wasm32:
- scripts read files through Session.services but checked existence with
std::fs (File.open's access check, File.exists, Folder.exists); they now
ask services.exists / services.is_dir, after the folder check;
- WebServices kept the default exists (std::fs); it now looks in the file
table.
Desktop and CLI are unchanged (FsServices keeps the std::fs defaults).
Tests: script tests_services (a host with in-memory files), web smoke
step hostFiles (a script reads a stored file, footage check finds stored
media).
Clicking keyframes on Position, Rotation and Scale highlights those
properties, and Copy then added every key of a highlighted property that
had none selected, so deselected keys came back on paste. The keyframe
copy branch also left a stale value clipboard that Paste preferred over
the new keys.
Copy now takes whole animations only when no key is selected or the
Transform group is selected, and every copy branch clears the clipboards.
Also collapses a nested `if` in the STL importer into a match guard,
which Clippy 1.95 flags under -D warnings (no behaviour change).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The font scan read only whether a face has an fvar table, so a variable
family offered one style (its default instance), fontStyle = "Black"
picked that face with synthetic bold, and an instance's PostScript name
(Figtree-Black, what After Effects scripts use) found no face at all.
The scan now reads the named instances (style, axis values, PostScript
name, or the one Adobe Technical Note #5902 derives). They are styles of
the family in text.fonts and the font menus, a style, PostScript name or
full name selects the instance, and layout uses its axis values as the
base variations; axes set on the text still win.
The roi quad is stored raw from the control channel (crop.useRoi passes
serde_json's [f64;4] straight through, unlike the sibling edge fields,
which are clamped to plus/minus 20 000) and nothing repairs it on load.
Crop::edges converted it with round() as i32 and subtracted raw: an x of
-1e12 saturates to i32::MIN and w as i32 - l overflows in debug builds,
wrapping to a wrong-but-bounded crop in release.
Clamp each converted edge to the same plus/minus 20 000 the sibling
fields use and subtract with saturating_sub in edges() and frame_size().
Regression test: hostile_roi_cannot_overflow_crop_math — panics at
crates/project/src/render_queue.rs:774 on main; with the fix the hostile
quad degrades to a 1x1 frame and the sane-ROI behavior is unchanged.
inline_len computed (w * n * bpc).div_ceil(8) * h with Width/Height taken
straight from the content stream and no size validation — unlike decode()
and mask_image, which bound their geometry first. /W 5000000000000000000
overflows the multiply in debug and test builds; in release it wraps to a
small length and the caller slices the wrong bytes, misparsing the image
and shifting the EI resync.
Chain the arithmetic through checked_mul and return None, which the
caller already routes to the EI-scan fallback. Regression test:
inline_image_with_absurd_dimensions_falls_back_to_ei_scan — panics at
crates/pdf/src/image.rs:85 on main, passes with the chain.
mask_image read /BitsPerComponent from the attacker-controlled /SMask
dict with no validation and fed it straight into 1u32 << bpc — the main
image path rejects anything but 1|2|4|8|16 before its own identical
shift, the mask path did not. /BitsPerComponent 32 panics in debug and
test builds and wraps the shift in release, mistranslating every sample
into alpha.
Mirror decode()'s depth check: an unsupported mask depth means no usable
mask, so the image draws with the mask treated as absent. Regression
test: smask_with_absurd_bpc_is_ignored_not_fatal — panics at
crates/pdf/src/image.rs:184 on main, passes with the check.
`pen_press` placed each new vertex where it was clicked (after snapping)
whatever the modifiers, so Shift did nothing and a vertical, horizontal or
45° segment had to be lined up by eye. With Shift held, the new vertex now
moves onto the nearest line at a multiple of 45° through the vertex before
it, as far along that line as the click lies (`constrain_45`). Masks and
shape paths both go through this path.
`exec_on_keys` put the named keys in the key selection, then `exec` made
the keys' comp active. Since comps keep their own selections (#1037),
activating the comp stored the injected keys with the comp that had been
active and brought back the target comp's own key selection, usually
empty, so `keys.interpolation` and `keys.velocity` had nothing to act on:
setInterpolationTypeAtKey and setTemporalEaseAtKey were silently ignored
in any comp other than the active one. The keys' comp is now made active
before the keys are selected, and the comp that was active comes back
afterwards, as `exec` does for its own target (now `leave_comp`).
Entries whose layer buffer was freed were pruned only once the cache went over its budget
(1.5 GiB), so an animated layer, which makes a new buffer every frame, left a dead 4K texture
per frame allocated until the cache filled up.
Retry only explicitly read-only control queries after connection failures. Report ambiguous edit outcomes without resending mutations, reconnect on the next explicit call, and reject request-ID exhaustion without overflow.
Add loopback regressions for lost and malformed replies, read-only retries, playback queries versus edits, and unknown methods.
Fixes#1247
Use checked division and remainder for EPS integer operators so i64::MIN with a divisor truncating to -1 cannot panic. Preserve the existing zero-divisor fallback and normal truncation behavior.
Add public EPS parse regressions covering extreme, fractional and ordinary operands.
PR #1223
The PDF / Illustrator import dialog now offers Composition - Retain Layer
Sizes. Each file layer's footage is cropped to its artwork bounds (fills,
stroke outlines, images, clips) and positioned to keep its place on the
page. Decoding, continuous rasterisation, shape conversion and reload share
one crop (effectcraft_pdf::layer_doc_with_sizes), so they agree.
Also collapses a match guard in the STL reader for current stable clippy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Black & White offset each pixel's luminance by (hue weight - 0.5) times
its HSL saturation. HSL saturation is near 1 for any tinted near-black
pixel, so the code or two of colour noise in compressed footage's
shadows came out as bright and black blocks, and a 32-bpc overbright
such as (1.5, 0.5, 0.5) divided by zero and went black.
The gray is now the pixel's smallest channel plus its chroma (largest -
smallest channel) times the same hue-interpolated slider weight. Neutral
pixels keep their value, a pure hue becomes its slider's percentage
(Reds 40% turns pure red into 40% gray), noise moves the gray by no
more than its own size, and nothing divides. The CPU effect and the GPU
kernel change the same way.
Merge each offloaded property-group result against the snapshot the job started from. Preserve disjoint live edits, group identity and child changes; refuse conflicting results without changing the document or undo/redo history.
Record an apply rejection on its originating job so completion logs and waiting control commands report failure. Keep concurrent jobs' outcomes separate and retain the correct completion position when the bounded log rolls over.
Add actual-worker regressions for conflicts, disjoint edits, concurrent analyses and undo/redo, plus control completion/error regressions.
Fixes#1249
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Create each atomic write through its own exclusively created temporary sibling and keep its open file handle until writing finishes. Concurrent writers publish complete files without truncating or following an existing temporary entry; failed writes or renames clean up their own staging file.
Keep the temporary basename bounded independently of the destination name so previously supported long project filenames still save. Retain the concurrent-writer and symlink regressions and add a 251-byte filename overwrite regression that checks exact bytes and staging cleanup.
Fixes#1233
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem: the built-in animation preset library had no Behaviors category
(gap 7: 54 presets against After Effects' 621).
Fix: 22 original Behaviors presets in crates/engine/presets/behaviors.json:
Soft Wiggle, Shake, Float, Sway, Orbit, Figure Eight, Bouncing Ball, Drift
Right, Drift Diagonal, Pendulum, Wobble, Spin Slow, Spin Fast, Pulse,
Heartbeat, Breathe, Rubber Band, Jitter Scale, Blink, Strobe, Flicker, Fade
Pulse. Each puts a time-based expression on one transform property or
opacity (no effects), built on the property's own value so it keeps running
for the layer's whole duration. Listed under * Animation Presets > Behaviors.
Sidecar, ATTRIBUTION.md, gaps.md row 7 and parity docs updated.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Problem: effect colour labels (26.5) were drawn only on the Effect Controls header; the Timeline's effect rows looked unlabelled.
Fix: an effect group row in the Timeline paints the label colour as a 4 px bar on its left edge (same look as the Effect Controls header) and registers `timeline.group.<uid>.label` with the label's name for agents. Docs: gaps.md row 21 and ui-parity.md updated.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
With several layers selected, typing a value into one dimension (say Z
Position) sent the edited layer's whole value with selected: true, and
prop.set gave every selected layer that same value: they all took the first
layer's X and Y and stacked on top of it. Scrubbing was fine (it sends an
offset).
prop.set takes dims: with selected and a typed value, the other selected
layers get only those components and keep the rest of their own value (Position
with Separate Dimensions included). The three panels send the components the
edit changed (both for linked Scale). Without dims, agents still set the whole
value on every selected layer.
Fixes#1434
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N9kwMcw2pYQ5fSjKa3ZPnh
Adds Georgian (ქართული, code ka) to Settings > General > Language, with
Match System for ka / ka-GE / ka_GE.UTF-8 locales:
- i18n.rs: the menu catalog, the same 628 (command, English label) keys as
the other languages, plus the Undo/Redo entries, the workspace and 3D
view shortcut submenus, the Reveal labels on Windows and Linux and the
macOS Services items.
- i18n/ui.rs: GEORGIAN, every row of the Chinese panel catalogs (687), and
GEORGIAN_PANELS, 195 rows for text the panels send through tr() that no
catalog has yet (form dialogs, drop messages, screen-reader names,
context menus, Project and Timeline column headers). The tests hold
them to the same rules (unique, same placeholders, used by a converted
panel) and keep them from shadowing a Chinese row; prefs_live.rs joins
the converted files.
- i18n/effects.rs: Georgian names for all 307 effects and 21 categories;
the CC series and brand names stay in English.
- Properties > Text Animation: the Add Animator and Presets buttons grow
with their translated labels (English keeps 116 / 84 px).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
Match System (the default) stayed English in the browser build, which has
no system-locale probe (#229). The page now passes ?lang= from the address,
then navigator.languages, and Match System takes the first of them with a
catalog (an English tag counts), else English. A language chosen in
Settings still wins; native builds are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
(cherry picked from commit 73c064e3858068ed19fbfafbdd0b3c8cc229a42e)
Text drawn outside the menu bar stayed English in every language:
- i18n::command_label: the Timeline / Composition layer and property
context menus now take the menu bar's row for the same command (… or
...), else the panel catalog's row; the layer menu's Effects submenu
draws effect and category names through the effect catalog.
- The dock's tab titles (and the dragged tab), the Composition, Effect
Controls and Properties tabs, and the workspace bar and menus use the
window.panel and window.workspace rows every menu catalog already has.
- The Project panel's search hint, details hint and column headers, and
the Timeline's column headers and Columns menu go through tr().
The Simplified and Traditional Chinese rows for "Type" read 文字 (text, the
Type tool); the string is the Camera Settings field and now the Project
column, so they read 类型 / 類型. dock_ui.rs joins the converted files. Languages without a row for a
string keep the English, as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YMTAbGNRxnMEzRvYnQGphF
(cherry picked from commit b335f1e2764929b78d7b678ec026f680b1a8129f)
The EPS interpreter counted one operation against its budget for each
name it ran, whatever the operator did, and paths had no length limit.
Parsing one arc at a radius of 10^300 passed 3 GiB of resident memory
within 43 seconds. A loop of arc at a radius of 2.2 x 10^18 was still
running after 60 seconds, and a loop of pathbbox on a path of 100,001
elements and { {} loop } loop did not finish within 20 seconds. PDF
content had no path limit either: parsing a page that repeats
0 0 1 1 re 4,000,000 times peaked at 2.3 GiB of resident memory.
A path now holds at most 1,048,576 elements (MAX_PATH, the limit of
VectorCraft's EPS interpreter). In EPS, moveto, lineto, curveto,
closepath, arc, charpath and the rect operators check it, and an
operator that would take a path past it stops the program with
"PostScript (path limit)". In PDF, m, l, c, v, y, h and re (each adds
at most five elements) stop the content stream with
"content (path limit)" when the path already holds more than 1,048,571.
Each path element an operator makes, reads or copies (arc, rectangle
arrays, charpath, pathbbox, clip, eoclip, gsave, save) now counts
against the EPS budget. Each pass of repeat, for, loop and forall
counts too, and so does each procedure item that is not a name. A
moveto right after a moveto replaces it, as the PostScript Language
Reference describes, so text placed with moveto and show leaves one
element.
Five tests in tests_import.rs cover the loops, the path limit in both
interpreters and runs of moveto and show.
EffectCraft's RPM requires Fedora package names: `libxkbcommon`, `(vulkan-loader or libglvnd-egl)`,
`alsa-lib`. openSUSE has no `libxkbcommon`, `vulkan-loader` or `libglvnd-egl` package, so zypper
refuses it with `nothing provides 'libxkbcommon'` (exit 4, which the ArtCraft Launcher shows as
"error 4"; storytold/craft-launcher#341).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two German menu labels described a different command:
- `layer.timeReverse` was "Ebenenreihenfolge umkehren", which reads as reversing
the stacking order rather than time -- now "Ebene zeitlich umkehren", matching
`keys.timeReverse`;
- `layer.style.convertToEditable` was "In editierbare Formate umwandeln", and
"Formate" are formats, not layer styles -- now "In bearbeitbare Stile
umwandeln".
Where the same English term had more than one German rendering, the one already
in the majority wins, so this settles rather than re-translates:
- "Expression" is "Ausdruck" in seven entries and "Expressions" in the two
`layer.expressions` ones, which now read "Ausdrücke aktivieren/deaktivieren";
- "Reveal" is "anzeigen" in eleven of twelve entries, so `layer.revealInFinder`
joins `file.revealInFinder` at "Im Finder anzeigen" -- in every other
catalogue both commands already share one string;
- "Track" as a verb is "verfolgen" in the other four entries, so `track.mask`
is "Maske verfolgen" rather than "Maske tracken";
- one spelling of "nach vorne", matching the neighbouring arrange entry.
German also had no platform variants for the file manager. The engine puts the
platform's name into the label before the catalogue sees it, so on Windows the
source is "Reveal in Explorer" and on Linux "Reveal in File Manager"; only
Ukrainian and Czech special-cased those, and a German UI showed the English
string. Added, with a test mirroring `czech_…` and `ukrainian_…`.
Two of the reported six are left alone: "Essential Graphics" is a 3-3 split and
a panel's proper name, and "Null" partly mirrors the English, which itself says
both "Null Object" and "Nulls". Both are translator's choices with no majority
to follow, not corrections.
Phones and messaging apps store frame times rounded to whole milliseconds
(33/34 ms at 30 fps, 16/17 ms at 60 fps). FilmCraft's reader takes the
median duration, so such clips imported at 30.303 or 58.824 fps, and New
Comp from Selection and renders followed that rate. When the median is no
standard rate, the durations near it are averaged and snapped to the
standard rate within 0.5 %; exact and variable rates keep the reader's
estimate.
draw_bounds padded every stroke by width × miter limit / 2 on each side, joins or
not. A straight 1080 px wide wipe used as a track matte became a 4326×6366 layer
buffer (~440 MB as RGBA f32), made anew every frame while its trim animates, then
cached on the CPU and uploaded to the GPU. With 12 frames exported in parallel a
16 GB Mac ran out of memory and froze.
Now the miter limit applies only to miter-joined paths with a join (two or more
segments, or closed); square caps reach √2 half widths, other strokes one; Taper
and Wave keep the widest padding. Rendered pixels are unchanged.
A full-res frame of the affected ad: 1.7 GB → 0.7 GB peak.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Gradient Editor's Preview now covers the other settings dialogs that edit
something that exists: Solid Settings on a solid layer, Camera Settings, Light
Settings and Composition Settings. A change of the draft reaches the project at
once, so the canvas shows it while it is made.
- One shared mechanism (`panels::live_preview`): each step goes back to a
checkpoint of the project and undo history taken when the dialog opened, then
runs the dialog's command once. So the project is always "as it opened, plus
one command": OK leaves one undo step, Cancel and Escape restore the
checkpoint, Preview off applies nothing until OK, and switching it off takes
back what was applied.
- A previewing dialog opens at the lower right, without the dark backdrop, and
the window behind it is no longer greyed out; it can still be dragged by its
title. The same dialogs creating something new (New Solid, New Camera, New
Light, New Composition) are unchanged and have no Preview.
- `Placement` is public, and `EffectcraftApp::dialog_state` gives tests and
automation the drafts.
Claude-Session: https://claude.ai/code/session_013fTnVxnXTv2MBW6uj2PPyU
Marco Arnulfo