A settings get no longer reads and parses settings.json each time. The
parsed settings are kept per plug-in folder and read again only when the
file's size or modified time changes. A set to the value a key already
has writes nothing, and every other change is written with an atomic
write before the call returns. plugin.remove drops the kept copy.
cacheBytes in plugin.list, plugin.info and plugin.install comes from a
total kept while the cache folder keeps its modified time, instead of a
listing of up to 8,192 items per call. A cache write still lists the
folder to enforce the caps.
docs/plugins.md states what a run counts and that a cache call moves a
whole entry in one step. docs/plugins.md and docs/mcp.md now say when a
later plugin.remove retries a leftover .removed folder.
each_filter_plugin_takes_a_shortcut_of_its_own required the exported row
to end in "+Alt+Shift+F7". Export Text writes each shortcut as the dialog
shows it, which is "⌥⇧⌘F7" on macOS, so the test failed there. The test
now compares the row with menus::pretty_shortcut, as
export_text_lists_the_dialogs_rows_with_their_shortcuts does.
The Warp gradient test now compares the baked bounds with the square's
outline mapped through warp_point, for every style and parameter set. It
requires the gradient to move exactly when those bounds move, and pins
39 of the 45 cases as moving the bounds.
A pattern's or a brush's art was copied into the document with new ids,
but the art of an opacity mask inside it kept the ids of the reader that
read the definition. Those ids could repeat ids of objects in the
layers. The mask art now takes document ids, as a symbol's art already
does.
The test for Crisp and Strong type in a scatter brush stroke at 50% opacity
now checks that every pixel is half of what the same stroke draws at full
opacity. It fails without the coverage fix for Crisp as well as for Strong.
Switching documents, or windows of a document, deactivates the Type tool, which clears the tool its double-click came from, so Escape after a later edit goes to the Selection tool. The field's comment and the two checklist rows now say so.
Spine::at_length computed tangents from flattened polyline chords, which
are piecewise constant and jump discretely between chords on curved
spines. Steps oriented to the spine with Align to Path stuttered at
large step counts.
Fix: store the original PathData in Spine and compute the tangent from
the original cubic at the true arc-length position via ParamCurveDeriv,
proportionally mapping flattened arc-length s onto each segment's curve
arclength using anchor_length values, instead of using piecewise-constant
chord directions.
Updated segment-selection condition in Spine::at_length to skip segments
whose flattened end is not greater than start before checking whether s is
within the segment. Preserve the existing final-segment fallback so
at_length(0.0) selects a nonzero segment and uses its tangent.
Added zero-derivative fallback tangent for blend-step sampling to keep
Align to Path from passing a zero tangent to atan2.
Added missing docstrings for Spine::new and Spine::length to satisfy
docstring coverage requirements.
Co-authored-by: Craft Bot <bot@storytold.dev>
dotCooCoo