* fix(layers): refuse to ungroup the sole empty layer group (#11118)
* fix(layers): document the ungroup guard and cover the valid empty-group case
Review follow-up: docstring on ungroup() and a control test showing an empty group that has sibling root layers still ungroups (the group simply disappears).
---------
Co-authored-by: Hartmut Nörenberg <hnoerenberg@users.noreply.github.com>
* PSD: keep stylistic sets in Txt2 for PhotoCraft-native documents.
needs_txt2 now covers Txt2-only OpenType flags (ss01–ss20 and non-default TXT2_FEATURES). Fresh documents get a framed point-text object with full style runs so Set 1/5 round-trip on save/reopen (#10308).
* Txt2: seed fill colour and font list on PhotoCraft-native blocks.
Fresh style runs get a /53 paint dict, and an empty font list gets a CoolTypeFont entry so index 0 resolves (CodeRabbit on #10313).
* Txt2: tag SimplePaint and ensure numbered /0 on legacy blocks.
CodeRabbit: write /99 /SimplePaint on fresh /53 dicts, and create the numbered resource dictionary before pad_frames when a legacy block lacks it.
* Txt2: treat a missing stylistic-set mask as 0 when comparing styles
A style run without key 85 used to agree with any ssNN mask, so enabling
ss01 could leave the file's runs in place on save. Align with import's
default of 0. Also restore the G8 gaps note after rebasing onto main.
* style(io): rustfmt psd_kerning stylistic-sets test
* Txt2: keep Photoshop /0 stable when seeding fonts and frames
Only seed a CoolTypeFont list and inject /0/8 when the block is
PhotoCraft-native or already had those keys / needs new frames, so
unedited Photoshop Txt2 saves stay byte-identical (#2374).
* style(text): rustfmt oversized-template assert
* PSD: keep the Txt2 scaffolding to PhotoCraft-native blocks
A file's own Txt2 block (any Photoshop era) is written as before: no
numbered /0 or /0/8 frame list is injected, a layer without a template
object gets its stub, a run without /53 isn't given one, and no
minimal style base is used. Only a block PhotoCraft creates itself
(prev.is_none()) gets fresh point-text objects, minimal runs, a seeded
font list and a frame list.
Fresh objects now write one /5 paragraph run per paragraph (UTF-16
lengths, checked conversion), as edited objects are refitted.
Tests: Photoshop blocks with an empty or missing /0 get no native
scaffolding; the oversized-template test is back to main's assertion;
fresh paragraph runs; the PSD round trip checks ss01 and ss05 on every
run.
---------
Co-authored-by: FredApps <153173854+FredApps@users.noreply.github.com>
- #9979: ui.dialog.set can no longer rewrite an open field picker's __colorPicker
marker (its OK went to the Crop shield instead of the field), and Preferences Apply
waits for an open colour picker as OK does, so it never saves a previewed colour.
Both have regression tests that fail without the fix.
- #9182: the stress test's PC_BUILD_ROUNDS / PC_BUILDS_PER_DEVICE fall back to the
default for 0, which ran no builds and passed.
- #10837: the ROADMAP entry no longer says a Latin hint rules windows-1251 out; the
hint settles a word-initial rare letter and breaks exact ties.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Undo after Reset now brings back the pixels beyond the frame: the box Undo
step also keeps whether the frame is being edited. Reset fits the first
frame to the ratio preset or output size picked in the bar, as a new frame
is, so a 1:1 preset does not end up with the 2:1 canvas.
The bar had only Cancel and Commit, and no way to put a changed box back
to its first frame without cancelling. Add Reset beside them: it frames
the selection's bounds or the canvas again, upright and with the straighten
line gone, and keeps the old frame as a box Undo step. It is disabled while
the frame is still the untouched default.
The position checks of the codepage guesser (letters after a character, a closing !,
the Cyrillic run around a letter) rescanned the rest of the name from every non-ASCII
character, so a hostile .atn or PSD with a 64 KiB step name of one letter kept Load
Actions busy for about three minutes. Each is now found once per name; the three
measured cases take 0.03 s together, and a regression test covers them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Toolbar: Alt/Option-click cycles the flyout tool slot.
Matches Photoshop: plain click keeps the shown tool; Alt-click advances through the slot without opening the menu (including tools Shift+letter skips).
* Toolbar: don't fit-on-screen on rapid Alt-clicks through Hand.
Alt-click only cycles the flyout slot; exclude Alt from the Hand/Zoom double-click handlers and cover inactive-slot cycling.
* Toolbar: ignore Hand/Zoom double-click after an Alt-click.
egui pairs a plain click with a prior Alt-click as a double-click; remember the previous click's Alt so fit-on-screen and actual-pixels only run for two plain clicks.
* Toolbar: read click modifiers from the slot's pointer release.
Match DesignCraft/EffectCraft: bind release mods to the tool rect, treat any modifier as blocking Hand/Zoom double-click, and harden the Alt-cycle test by clearing i.modifiers before release.
* Toolbar: cover 1/2/3-tool Alt-click cycles and hidden-tool slots.
Addresses ReviewerA on #10301: next_slot_tool wrap cases for short flyouts, and visible_sections so Edit › Toolbar hidden tools shrink the Alt-click cycle.
* style(ui): rustfmt after rebase onto main
* Toolbar: next_slot_tool never panics on an empty slot
Index with checked arithmetic and get() (house rule: no panics on
input-derived indices); an empty slot returns the current tool. Test
added.
---------
Co-authored-by: FredApps <153173854+FredApps@users.noreply.github.com>
Main now hides the frame after a committed crop (`hide_frame` /
`show_frame`, #10642) on the same `dismissed` state this branch added,
and keeps it hidden when the Crop tool is picked again, as measured in
Photoshop. Esc now uses that state: `dismiss` is `cancel` then
`hide_frame`, so after Esc picking the tool again also keeps the frame
hidden, and only a press brings it back. The Esc test follows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
While drawing, Make Shape no longer commits the path first: the shape takes
what the commit would make, as one History step. With a vector mask
targeted the shape is the drawn path alone, and the mask stays as it was.
A saved path being drawn into stays as it was.
path.toShape takes an inline `path` (with `consumeWorkPath`), no longer
copies the active shape or vector mask when there is no Work Path, refuses
a path without a segment, and refuses a `layerName` that isn't a name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EZbgszHyuEuWAzxaamteuU