craft-svg builds against an older vectorcraft-geom and vectorcraft-pathops by git, so the
graph holds two packages of each name and `cargo check -p vectorcraft-geom` is ambiguous.
The wasm step now passes each workspace crate's package id.
Dragging out a dimension from a segment (Linear) or a corner (Angular)
drew its first preview frame at the wrong distance. The preview placed
the dimension at the pointer before reading the segment's end points or
the corner from the art, and measured the offset from the placeholder
points the dimension starts with. A drag from the top edge of a
rectangle to 40 pt above it showed an offset of 60 instead of -40.
The preview now clears the offset, reads the art's points, places the
dimension at the pointer and checks the result. Clearing the offset
keeps an out-of-range offset from the last pointer position from
failing the read.
Tests: the segment and corner drag-out tests fail without the change.
placement_recovers_after_a_position_out_of_range also passes on main;
it fails when only the offset clear is removed.
warichu_is_a_character_attribute_set_saved_and_exported looked for Tj or
TJ anywhere in the exported PDF. The file ID in the trailer and the XMP is
a base64 hash that changes each second and now and then spells one of
them, so the test failed on main twice (bf3cf422e, 446bf59bc) while
export outlines warichu type every time. It now counts text-object lines
only, as the colour font PDF test already does.
Changing Preferences › Hyphenation › Exceptions bumped each open
document's revision, but the caches of laid-out type outside the text
crate were keyed on the font generation alone. The renderer kept the
glyph geometry, raster-effect pixels, live-effect and envelope art,
pattern tiles and placed symbol art of unchanged type objects, so
hyphenated area type kept its old line breaks on the canvas. The
hit-test art of blends and envelopes, the Fit to Artwork Bounds answer,
the layout caches of the Type and Touch Type tools and the out port's
overflow mark were keyed the same way.
vectorcraft_text::layout_inputs() returns a number that changes with
the fonts and with the hyphenation exceptions. Every one of those
caches now keys on it instead of FontDb::generation(), and envelope ink
uses it in place of the private envelope_ink_inputs(). The hit-test key
hashes it together with the plug-in revision.
Tests: a renderer that drew hyphenated area type, alone, with a drop
shadow, in a group with a Transform effect, in a symbol instance and in
an envelope, draws it as a new renderer does after the exceptions
change; the out port of selected area type shows the overflow that the
new exceptions cause. The render test helpers that run a step again
when the fonts change during it now also run it again when the
exceptions change.
The Parametric Shape tool's center and the Slice tool's corners snapped
with Smart Guides to the bounds of the art only, so View › Snap to Glyph
had no effect on them, and neither did the anchors and segments of
paths that the other drawing tools snap to. The point dragged out from a
Parametric Shape's center did not snap at all, and neither tool showed
Smart Guides while hovering before a press.
Both tools now snap through the drawing tools' shared snapping
(DrawSnap). A Parametric Shape's center (from a drag or a double-click),
the point dragged out from it, and a slice's corners land on glyph
guides and on path anchors and segments. The guides show while hovering
and during the drag. When a ruler guide then takes a slice corner, the
Smart Guides are no longer shown. A click with the Parametric Shape tool
still draws nothing when its center snaps a few pixels from the pointer.
Docs: docs/mcp.md, docs/target-app-parity.md.
egui places an Area by the size it had the frame before. When a family
name measured late (a face added for its script, a font found in the
background) widened the menu on the helper's last frame, that frame drew
it where the narrower menu fit, and a_menu_wider_than_its_field_stays_in_the_window
saw it 0.7 pt past the window's edge (three failed CI runs on 2026-11-13).
The helper now draws until the menu stops moving.
With the Selection tool, a change in the Character or Paragraph panel to a
selected frame of threaded text now changes every frame of its story,
locked and hidden frames included, in one undo step. Before, only the
selected frames changed.
- The panels send the frames of each selected story as the command's ids:
every Character and Paragraph field, the font family menu and its
preview, Bullets and Numbers, anti-aliasing, Variable Font axes, Reset
Panel in both panels, the Tate-chu-yoko, Warichu and Hyphenation
dialogs, and the Tabs panel's stops and Clear All Tabs. A story
selected through several frames or a group is listed once.
- The fields read every frame of the story and show a value only when all
frames share it.
- Clear All Tabs is enabled when another frame of the story has stops.
- Convert to Text keeps acting on the selected frames.
* PDF import: outlined Indic glyphs keep their own stroke and transparency (#5235)
Follow-up to #6906, for two review findings on how a line kept as
outlines is joined into paths:
- Glyphs were joined when their stroke widths matched; dash, cap, join
and miter limit were ignored, so a dashed glyph took the stroke of the
solid one before it. They join only when the whole stroke matches.
- Translucent glyphs were joined into one fill, where overlapping marks
paint once instead of compositing as the file draws them. Only opaque
glyphs join.
Also records the fix in the PDF row of docs/target-app-parity.md and in
docs/roadmap.md, as AGENTS.md asks.
Tests: outlined_glyphs_stroked_differently_stay_apart and
outlined_translucent_glyphs_stay_apart (both fail without the change).
* PDF import: outlined Indic glyphs under a soft mask stay apart; check kept geometry (#5235)
Review follow-ups on #6957 and #6906:
- Glyphs kept as outlines under a soft mask were joined into one path,
so the mask was applied once to all of them where the file applies it
to each glyph. They now join only when no soft mask is active.
- The fallback tests counted outline nodes only. A new test compares the
kept outlines (element count and bounding box) with the same file
opened as outlines, so a dropped or moved glyph fails it.
Tests: outlined_glyphs_under_a_soft_mask_stay_apart (fails without the
change) and kept_glyphs_are_where_the_file_draws_them.
* PDF import: mask outlined Indic glyphs one by one under a soft mask (#5235)
Review follow-up on #6957. Keeping glyphs out of one shared path was not
enough: each path still went into the soft mask's queue and
flush_masked wrapped them all in one masked group, so a mask that is not
constant was applied once to every glyph, where overlapping glyphs
composite one by one in the file.
With a soft mask active the queued art is now drained before the first
glyph and after each one, so every glyph gets a mask of its own.
Test: outlined_glyphs_under_a_varying_soft_mask_are_masked_one_by_one
checks that each outlined glyph carries its own mask (0 of 3 without
the change).
* #6957: clarify soft-mask exception in glyph grouping docs
---------
Co-authored-by: ruchirguitar <info@jeevantirth.org>
After Reset Bounding Box on a turned symbol instance, Break Link to
Symbol and Object > Expand gave the art a box turned with the instance's
transform again. The art's boxes now turn with the instance's own box,
so they stand square when the instance's box did, and the angles the
art has inside the symbol are kept relative to it. The group that an
instance's opacity, blending mode or opacity mask adds around the art
takes the instance's box angle instead of standing square to the page.
A reflected instance can hold a box 180 degrees from the box its
transform gives in one step; the art now takes the instance's box
there too, the same rectangle with its handles where the instance had
them.
Compound shapes and Pathfinder receive an instance's art through the
same rule.
The clip groups and clip paths that 9-slice guides make for gradient
and pattern art take the angle of the cell that places them, and the
outer group keeps the object's angle turned by the instance, instead
of standing square to the page. This also changes the result for an
instance that was never reset: a 20 degree sliced instance gave clip
groups at 0 degrees and now gives them at 20.
Docs: params text, docs/mcp.md, docs/target-app-parity.md.
The Home screen gets a Learn section between Recent Files and the new-file
presets: the Quick skill guide card, which shows the Learn panel's list of
skills (learn.skills), and each tutorial under its skill, which shows the
panel at its steps (learn.show). It works with no document open and over
open documents, opens no document and never hides the panel.
The card is shared with the panel's For you tab and now shows an outline
while it has the keyboard focus.
A straight `(N) *` guide path that runs across the canvas or an artboard
opens as a ruler guide, but a hidden one (`1 Xw`) stayed a guide path.
It now opens as a ruler guide too, and every ruler guide made from such a
path keeps the path's hide, lock and name, so its Layers row shows them.
Before, a locked guide opened unlocked and a named one lost its name.
Hidden or locked guide paths that are short, slanted, grouped or `(S) *`
stay guide paths, and commented copies (`%_`) are still left out.
Docs: docs/target-app-parity.md, docs/file-format-parity.md.
A swatch delete, Swatch Options edit, merge, Edit Color Group, Spot
Colors options change or Default Swatches replace run as a
command.batch step left the fill and stroke new art takes linked to the
old or deleted swatch, because the defaults were skipped whenever an
interaction was open. A batch step now rewrites them as the command run
on its own does; a live Swatch Options preview still leaves them alone,
and a failed batch or a canceled batch preview puts them back.
The same commands now also rewrite the fills and strokes of the
appearance new art takes, and record the change with their undo step,
so Undo and Redo put the paints for new art back with the swatches.
BFlat