- EPS CCITTFaxDecode goes back to hayro-ccitt (main's filters.rs and data.rs):
craft-stream-codecs decodes /K -1 /Columns 70 data to other rows
(eps import::tests_images::ccitt_fax_images_draw). DCTDecode is main's again.
- The AI editing writer decodes through the image crate again: going through
RGBA8 lost 16-bit sources, so their 8-bit warning disappeared
(ai_editing::tests::sixteen_bit_images_are_written_at_eight_bits).
- Crop Image refuses an image over its allowance by header size again, with
main's "an image of W × H pixels is too large to draw" message.
- Test fixes: colorized_images writes its gray PNGs gray again, and the PDF
16→8-bit test expects rounding, as the 8-bit copy is made.
Reuse the panel deletion helper so the clicked swatch is deleted only after confirmation. Cover confirmation text, cancellation, clicked-item targeting and Undo across all five views. All 39 Swatches tests and all 13 local CI gates pass; fresh deep CodeRabbit review reports zero findings.
A frame of type on a path whose first paragraph continues from the frame
before now draws a repeat mark at its start as the character before it in
the frames before, as area and point type do. The mark keeps its advance,
so the glyphs stay where they were, and live SVG text writes the character
drawn.
The engine's turn-direction check requires every rotation param to say which way it turns. shape.editFlare's rotation is shape.flare's setting, which turns the rays clockwise.
`sent` now reports a spooler note as "The spooler replied: ...", so the
cross-language case checks that form instead of the old "(job 12)" one,
which matched the plain row and passed without the note row.
The touch points of a drawn line were found by sampling each target
segment in 24 steps and bisecting where the sign changed. When the line
was exactly square or tangent at one of the samples, the value there was
exactly 0, neither interval changed sign, and no snap was offered: for
example the foot of a perpendicular from a point above a segment's
middle, or a circle's tangent point when it fell exactly on a sample.
A sample whose value is exactly 0 now counts as a touch point when its
neighbors are not 0, so a straight curve seen from its own line still
gives none. The ends of the segment, a cusp and a start point on the
curve itself are left out, as between samples.
Part of #7846
paint.invert and paint.complement given a group whose own fill or stroke
changed, with only graph design art inside it, left the last color and
the recent colors as they were, because the remembered paint was taken
only from the objects inside the group. It now comes from the first
group whose own fill or stroke changed. A recolor that changes no
group's paint, such as the complement of a gray fill, remembers nothing.
A button beside the Help › Search Commands field opens a Recent Commands
list: the last five commands and tools picked from the palette, most
recent first. A command picked again moves to the top, so it takes one
place. Choosing a row closes the palette and runs it the way a pick of
the same result does, so dialogs, a dimmed item's reason and tool picks
work as before. With nothing picked yet, the list shows a dimmed No
recent commands row.
A command that is not available now, such as a filter of a plug-in that
is not loaded, is left out of the list but stays in the saved ids, so it
shows again when it comes back. The saved list keeps the last five
available picks and at most 20 ids in all. It is saved with the UI state
next to Recent Fonts.
While the list is open, its first row takes the keyboard, Up and Down
move through its rows and wrap at the ends, Enter runs the row, and
Escape closes the list before the palette.
Docs: docs/target-app-parity.md (command palette).
Fixes#6376
Vertically aligned area type in a frame that is not a rectangle flows its
text again until it settles, and each pass added the line-start repeat
marks to the layout's records once more, so SVG export could look up the
wrong character. Each pass now starts from empty records, and an undone
pass restores the ones before it.
The progress line said nested symbol instances leave nothing live after
Expand Appearance. It now names the 1,024-image and 256-megapixel
limits, the objects they leave with their raster effects and the
`rasterLeft` warning, and the test that covers them.
The ka and ru catalogs already had a row for `connect end must be
"start" or "end"`, and this branch added a second one, which fails
bundled_catalogs_are_consistent. The added rows are removed and the
existing translations stay.
The Georgian and Russian catalogs from main still carried rows for the
old `{within}{key}`, `{within}settings.{format}` and `{lo}–{hi}{unit}`
messages, which no message matches since their values are joined first.
Both catalogs already have rows for the joined messages.
craft_image_codecs::probe_size only reads WebP, EXR and the GIF/BMP/TGA family;
for PNG, JPEG and TIFF it returns Unsupported. The cutover used it for header
sizes, so DXF export left every PNG out (cad tests), plugin objects lost PNG
pixels, and EPS image-mask and SVG import fell back to decoding whole images to
learn their size. vectorcraft_doc::header_size reads PNG's IHDR and a JPEG's
frame through craft-image-codecs and other formats through the image crate's
header readers, as main does, without decoding pixels. The SVG filter's JPEG
size comes from jpeg_info instead of a full decode.
Each line that started with a repeat mark scanned the paragraph's text from
its start for the character the mark stands for, so a paragraph with marks
at the start of many lines took quadratic time to lay out. The scan now
goes on from the line before.
Main added Spanish rows for ten of the dialog-answer messages this branch also translates. The branch's copies are removed so each message has one row and bundled_catalogs_are_consistent passes; main's translations are identical.
Main's Georgian catalog has a row for `index` must be a whole number
and one for the knockout groups warning, so the copies this branch
added are removed. Main's knockout row had two spaces after the colon,
which neither the program's text nor the message scan matches; it now
has one. The Georgian counts in localization-parity.md are recounted.
The egui crate's header comment and the M14.8 line in
docs/target-app-parity.md called it "the ⌘K palette" and "a ⌘K command
palette". Cmd+K runs edit.preferences, and help.commandPalette (Help ›
Search Commands…) defaults to Cmd+Shift+/. Both now name the menu item,
and the crate comment links the palette module.
Merging main brought in Georgian and Russian rows for the SVG filter messages that this branch also added, so each catalog held them twice and bundled_catalogs_are_consistent failed. The branch's copies are removed and main's rows stay.
docs/mcp.md said that moving a label on its own never moves or scales a
graph's plot. A graph saved without `placedArt`, or with one that has no
width or height, is measured on `placed`, the bounds of all of its art
including the labels. Moving an outermost label of such a graph on its
own changes those bounds, and the next redraw moves and scales the plot.
The paragraph now says so.
The image made of an object's shadows and outer glows took the default
blend mode, so a Multiply shadow (Drop Shadow's default) over colored art
became a plain shadow painted over it. When the shadows and glows all use
one blend mode, the image now paints with it, for the object's own effects
and for nested symbol instances alike.
A thread of frames that hold only repeat marks or soft returns made each
frame look back through every frame before it, so opening or flowing such
a story took quadratic time, with the setting off as well. The leads are
now carried forward in one pass.
The read-root check for SVG output now looks at the same art the
writer draws:
- A placed document's art is checked, with the pixels its file holds.
A placed native file that links an image outside the read roots was
written with that path under Images: Link.
- Hidden objects inside symbols, patterns and graph designs are left
out unless hidden layers are kept, as they are in the layers.
- With profile tiny12 the file leaves out opacity masks, and their
images are not checked.
- An object whose effect SVG has no filter for is checked as the image
the file draws of it, for file output and clipboard.exportSvg alike.
The last three refused files and clipboard SVG that name no file
outside the roots.