Review follow-up. The file(1) check for binaries that cannot run here now needs an executable
(file says "executable" or "pie executable") or an ELF with an interpreter, which is how an older
file(1) still shows a PIE it calls a shared object. FreeBSD libraries, object files and core files
for the target architecture are refused.
Part of storytold/craft-common-actions#39.
With the keyboard in a Design panel field (X, W, a colour…), the
shortcut handler returned before looking at any key, so Cmd/Ctrl+S and
Shift+Cmd/Ctrl+S did nothing: no save, no message, the tab kept its
unsaved dot. Those two now work from a field: the field lets go of the
keyboard, which commits what was typed as clicking away does, and the
save runs the next frame.
Review: `mcp --bogus --help` stopped at --bogus with exit 2. The parser
now keeps the first bad option and reports it only when no --help or -h
follows.
Review: " . .name. . " kept inner dots at the edges; trim spaces and dots
in one pass. docs/cli.md and safe_file_name's comment give the 100 and
200 character limits.
Review follow-up. CARGO_BUILD_TARGET must now be a *-unknown-freebsd triple, so a Linux or macOS
target can no longer be packaged under a FreeBSD name, and it is read before uname -m, so an
unsupported host does not block a valid cross build. When the binaries are for another
architecture and cannot run --version, file(1) must report each as a FreeBSD ELF executable
for the target architecture.
Part of storytold/craft-common-actions#39.
`figcraft-cli mcp --help` failed with "unknown option --help" and
exit code 2; only a leading --help worked. The argument parser now
returns CliError::Help for --help or -h among a subcommand's options
(not as an option's value), and main prints the usage on stdout with
exit code 0. docs/cli.md says so.
export into a folder replaced every character outside letters, digits,
'-', '_', space and '.' with '_', so a frame called
"Homepage - desktop (live + verbeteringen)" became
"Homepage - desktop _live _ verbeteringen_.png". The app's
file.exportLayers keeps those names and only replaces what file systems
refuse. The CLI now uses the engine's safe_file_name, and docs/cli.md
says which characters are replaced.
Also adds my own entry to contributors/people.toml.
The release workflow now cross-compiles FreeBSD aarch64 inside the x86_64 VM
(storytold/craft-common-actions#50) and sets CARGO_BUILD_TARGET=aarch64-unknown-freebsd.
package.sh now reads the binaries from target/<triple>/release when CARGO_BUILD_TARGET
is set, names the archive after the target instead of uname -m, and skips the --version
run when the binary is for another architecture than the machine. x86_64 packages are
unchanged.
Part of storytold/craft-common-actions#39.
Found while testing crates/text/src/layout.rs:1070 (shape_segment calls TextCase::apply_after one char at a time; crates/doc/src/text.rs:124 apply_after)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
node.create, node.set and the other prop appliers map fill/stroke
"none" (either case) and null to an empty fills/strokes list, as
fills: [] already did (undocumented), so an outline-only shape is one
call - agents write "none" for no paint. KEYS and docs/mcp.md
document it; the error names the accepted forms. fill_none_tests
covers create, set, undo, text and the error.
Change to rebuilt the instance from the new variant, and set variable
and set variable mode re-applied bindings, but the player never laid the
result out: no text sizing, no auto layout. A hugging button with a
longer label than its main showed the other variant at the main's width,
its label clipped, and a text bound to a variable kept its old size.
These actions now run the engine's normalize on the player's document
afterwards, the pass that follows every edit on the canvas: text resizes
to what it shows, instances follow and auto layout reflows.
Found while testing crates/svg/src/lib.rs:153 (esc, used for tspan text at :598, :608 and :613)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A file or library saved before "modes" was an override group has a mode
chosen on a layer inside an instance without its marker, so the next
rebuild or library import still put the main's mode back.
Copies follow their main layer's modes when the main changes them, so a
layer inside an instance on other modes chose them itself.
mark_layer_modes marks those when a file opens and when a published
library is read from a file or a server. Layers on their main's modes
and top-level instances stay unmarked.
A variable mode chosen on a layer inside an instance (the icon inside a
Close button, recoloured through an icon colour collection) wasn't an
override group. A full rebuild of the instance starts each layer over
from its main and carries only override groups, so the mode fell back
to the main's. Library import always rebuilds, as do main edits such as
a new auto layout direction, so a modal's recoloured close icon came
into other files in its default colour.
"modes" is now an override group: setting a mode inside an instance
records it, rebuilds carry it, rebasing a nested instance finds it, the
component panel lists it, and resetting it brings back the main's mode.
A top-level instance's own modes stay its own and record no override,
like its resizing.
Narrows the range-extension wording: the date library computes the
calendar fields, and splitting seconds into hours, minutes and seconds
is ordinary arithmetic, which is what the code does.
The rule said never to hand-roll calendar math, but the range fix
(shifting by whole 400-year Gregorian cycles so dates past chrono's year
range keep their old output) is cycle arithmetic, so the rule and the code
disagreed. The rule now names that exception: the library still computes
month, day and time, and only whole cycles are carried in the year.
`utc_date` (collab) and `date` (ui-egui) took `u64` seconds, which the old
hand-rolled civil-from-days code converted for the whole `u64` range. The
chrono rewrite used `from_timestamp(..).unwrap_or_default()`, so any instant
past chrono's own year limit (262143) silently became the epoch — changing
existing output. `u64::MAX` seconds, for example, used to read
`584554051223-11-09 07:00 UTC` / `Nov 9, 584554051223`, not 1970.
Fix: add a `utc_parts(days, sod, nanos)` helper per crate that moves the
instant into chrono's range by whole 400-year Gregorian cycles (146_097 days),
lets chrono do the month/day/time, and restores the real year afterwards.
chrono stays the only calendar logic; the cycle count just shifts the year, and
every `u64` day count fits in `i64` so the conversion is always exact (no epoch
fallback). Split seconds with `u64` arithmetic first to avoid the `i64` seconds
overflow near `u64::MAX`, matching the old code.
Tests now pin the old output instead of the epoch:
- collab utc_date(u64::MAX) == "584554051223-11-09 07:00 UTC"
(and utc_date(253_402_300_800) == "10000-01-01 00:00 UTC", already pinned)
- ui-egui date(253_402_300_800) == "Jan 1, 10000"
date(u64::MAX) == "Nov 9, 584554051223"
chrono's %Y writes a '+' before years past 9999, whereas the old hand-rolled
{y:04} did not. utc_date now writes the year with {:04} to keep the exact old
output. from_timestamp already reads a clock past chrono's range as the epoch
rather than panicking. Tests cover the 5-digit-year and out-of-range cases.
Replaces the two hand-rolled Howard Hinnant days-from-civil copies with chrono
as a workspace dependency:
- crates/collab/src/message.rs: utc_date (YYYY-MM-DD HH:MM UTC)
- crates/ui-egui/src/home_recents.rs: date (Mon D, YYYY, UTC; keeps the
translated month names and trf formatting)
Output is unchanged. Neither function had a test, so each gains one pinning the
exact strings the old code produced (leap days included). chrono uses only std,
so wasm stays safe. AGENTS.md/CLAUDE.md now say time goes through chrono.
Follow-up to storytold/cadcraft#1785
A Button main's override on its icon instance's glyph (a red fill) was
lost when an instance swapped the icon: the outer rebuild re-copies the
glyph from the main's layer with no overrides of its own, so the nested
rebuild matched it by name and carried nothing. Before that rebuild, each
old layer of the swapped instance now takes the overrides held along the
source chain it had before this rebuild, through layers inside instances
in the mains that place the swapped instance. Geometry, slot content and
groups the layer overrides itself are left out. Matched by name, the
values carry onto the new main's layer as the instance's own: inherited
at the swap, frozen after, and back to the icon main's value on reset.
An icon main's own overrides stay with it.
stroke_area used rect_side_strokes only for square corners; a rounded
rect-like layer with individual weights fell through to a uniform ring at
stroke.weight (canvas, SVG, PDF, shadows, shape builder, outline stroke).
pathops::rounded_side_strokes now draws them as CSS does: the outer edge is
the rounded rect (the fill outline for inside strokes), the inner edge is
inset by each side's weight with elliptical corners max(0, R - weight), so
border-l-4 rounded-r-lg is a straight bar and a weight that differs across
a rounded corner tapers through it. Center and outside grow the outer box
and each non-zero radius by half or all of the weight (CSS has only
inside). The inner edge runs against the outer one, so the ring is a ring
under non-zero too; where it could leave the outer edge (radii shrunk to
fit a narrow inner box, smoothed outer corners) it is cut out instead.
Found while testing crates/doc/src/proto_expr.rs:172 (display writes `not {x}` with a bare var_ref) with crates/doc/src/proto_expr_parse.rs:262 (not() checks variable names before the `not` keyword)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
apply_order gave fillStyle, strokeStyle, effectStyle, gridStyle and
textStyle the rank of plain values, so the JSON map's byte order decided:
fills, strokes, effects, grids, layoutGrids and variationSettings sort
after their link and overrode it (node.set then unlinked the style;
node.create kept a link its values contradicted). Style links now apply
after the values and x/y, just before boundVariables. A deviation set in a
later call still unlinks.
Found while testing crates/doc/src/proto_expr_parse.rs:462 (variable lookup before the true/false literals at :466) with crates/doc/src/proto_expr.rs:151/186 (display writes the literal as bare `true` and var_ref writes such a variable bare)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A third row in open_subpaths.svg: a filled and stroked path that is only
0.001 short stays one layer, and an open triangle with
paint-order="stroke" puts its stroke layer under its fill layer.
SVG fills every subpath as if it were closed; the import filled only
subpaths that close themselves (a z, or an end within 1e-6 of the
start). Logos exported from PDF round relative coordinates to three
decimals, so their outlines end about 0.001 short of where they started
and most of their letters came in empty.
A filled path now has its open subpaths closed before it becomes a
vector network: an end within 1/1000 of the path's size snaps onto the
start, any other open subpath of two or more segments gets a closing
segment. Stroke-only paths are unchanged. A filled and stroked path
with a really open subpath becomes a fill layer (closed) under a stroke
layer (open), so the stroke stays open as SVG draws it.
SaveFiles named each test's folder after the process id and the
clock in nanoseconds. The tests run in parallel and the macOS clock
counts microseconds, so two of them often got the same folder and one
failed with AlreadyExists (14 of 30 runs here). A counter now keeps
the names apart.
Switching an instance to another variant copied the old root's sizing
and min/max back over the new variant's on the rebuild, so an instance
switched from a hugging variant to a fixed-width one whose children
fill kept hugging at the old width. The next edit then turned those
children Fixed, as a child can't fill an axis its parent hugs.
Resizing still equal to the old variant's was never the instance's own,
so on a sibling-variant switch it now follows the new variant; sizing or
min/max the instance set itself stays.
A Shift-click with the Pen already put the next point at a 45° step
from the last one, but the line drawn to the pointer while moving
ignored Shift and followed the pointer, so holding Shift looked like it
did nothing. The line now ends where the click will put the point. Both
use one helper, at_45_degrees.
Pin third-party actions to full-length commit SHAs (with version comments)
instead of mutable tags/branches to prevent supply-chain tampering via
retagged releases. Resolves SonarCloud rule githubactions:S7637.
Gert Pellin