When a command chord comes before a key with a hold shortcut in one frame,
and the key's wait ends in that frame, the key goes back to the frame as
with Hold Shortcuts off. Its repeats in that frame, the text they type and
its repeats in later frames now go back too; before, they were held back
until the key was let go.
When one entry holds a key with Shift (Shift+1) and another the character
that key types (`!`), holding Shift+1 runs the hold shortcut of Shift+1.
Before, the typed character was matched first, so the Shift+1 hold
shortcut could not run.
A command that needs a value, such as Transparency, is listed in the
editor only while it has a shortcut. shortcuts.set now refuses a hold
shortcut for such a command without a shortcut, and the command loses
its hold shortcut when its shortcut is cleared or given to another
entry, through shortcuts.set, the Keyboard Shortcuts dialog or an
imported set. Its key no longer runs a command the editor does not show,
and it is free for another entry without a conflict.
When a key with a hold shortcut is pressed and released in the same
frame after a chord with Cmd, Ctrl or Alt, the key no longer waits: it
goes back to the frame, which handles it as it does with Hold Shortcuts
off. Its tap no longer runs before the chord, so Cmd+Z followed by X in
one frame no longer undoes X's change instead of the earlier edit.
When a wait starts and ends in one frame and a key typed before it opens
a dialog, the frame's key events and texts are taken from the position of
that key's first event (its press, or its text when the text came first),
not from the event that ended the wait. An Enter pressed between the two
no longer reaches the dialog.
A settings file can hold a Hold Shortcut Delay outside 0.2 to 1 s. The
dialog showed the value brought into range but kept the saved one, so OK
sent it to prefs.set and failed. The dialog now opens with the delay in
use, and OK keeps it.
- When a key with a hold shortcut is pressed and its wait ends in the same
frame, the single-key shortcuts typed before it in that frame run first,
so the keys act in the order they went down.
- The repeats of a key in the frame that starts its wait, and the text
they type, are dropped like its later repeats.
- Import Shortcut Set started from the VectorCraft defaults clears the
working hold shortcuts; Add to the Current Shortcuts keeps them.
- Export Text's shortcut sheet and CSV file list a row's hold shortcut
after its shortcuts while Hold Shortcuts is on, as the text file does.
main's gate fails french_catalog_uses_nonbreaking_prose_punctuation on
the "{pos} of {total}: {name}" row of fr.tsv, added by #8017, which has a
plain space before the colon. The row now uses U+00A0 there.
Open #8053 fixes three other rows that main already has right, so it
does not clear this failure.
category_fields now returns the rows that carry a description, so the
Hold Shortcut Delay test's UI closure discards the value, as the other
prefs_dialog tests do.
With Link, File > Place now places an EPS file, a PostScript file and a
PostScript .ai as a linked placed document, but the Place dialog still
counted them as other files, so Link's tooltip did not say they become one
locked object, and the dialog test expected a PostScript .ai to be
embedded.
- `is_vectorcraft_file` counts an EPS, EPSF or PostScript file on disk, and
a PostScript .ai, as a document, and leaves out a DCS file, which is
placed as a linked image.
- Where the platform reads part of a file, it reads only the file's start
and tells a DCS file from it with `place::head_is_dcs` (the check Links
uses), binary EPS headers included.
- The dialog test now checks the .ai, .eps and .ps cases, the bounded read,
DCS files plain and behind a binary header, and a file named .eps that
isn't PostScript.
The rows of Select by Artboards and the Variables panel, and every
collapsible section header, reported their state through
`WidgetInfo::selected`. egui writes that value as the node's toggled
state for every role, so the accessibility tree showed these rows and
headers as toggled. AccessKit keeps a list item's selection in its
selected property and a disclosure control's state in its expanded
property.
- `widgets::list_item_info` gives a row the list item role and sets its
selected property, and both lists use it.
- `section_toggle_name` sets the header's expanded property, as egui's
own collapsing header does.
- Tests draw both lists and a section header and check the role and
state of each node in the accessibility tree.
The drag check took any line intersection kurbo reported, and kurbo also reports
one a tiny step before the start or after the end of a cubic. On a long curve that
step reached past the 0.0001 pt join tolerance, so a drag 0.0002 pt short of an edge
stroked it. Only intersections inside the segment count now; the checks on the
edge's ends still cover touches there.
With variation cells selected, the Save Color Group to Swatches button
now saves those cells as one swatch group, in grid order (row by row,
left to right, each cell once), named after the harmony rule, in one
undo step, each color in its own model. With no cell selected it saves
the harmony colors as before.
A cell selection now ends when the guide it was made in changes: the
base, the Color Guide Options, the variation mode, the harmony rule, the
library limit, or the colors the grid shows. Before, changing the
variation mode or the options kept the old cell positions selected, so
Save Colors as Swatches could save colors that were never picked.
With a gray between black and white, an edge pixel of black on white
took the black-and-white blend only when one of its eight neighbors lay
past the gray on each side and the flat black and white areas lay nearer
than the flat gray. On an edge blurred by a pixel or more neither holds,
so gray slivers lined black shapes: a black disc beside a gray square on
white traced up to 35 gray paths at blurs of 1 to 3 pixels.
The check now walks from the pixel along the line between the two outer
colors, up to 5 steps each way, each step to the neighbor that lies
farthest toward one of them while the pixels keep getting darker (or
lighter). The blend of the outer two lies as near as the farther of
their flat areas, or as the larger of the two step counts to a pixel of
each outer color when that is smaller. The first step is the old
neighbor check. The same disc and square now trace with one gray path
at blurs of 1 to 2 pixels. At 3 pixels, with the colors named, a gray of
128 is one path, a gray of 200 leaves 4 slivers with no area, and a gray
of 230 leaves 8 slivers of up to 21 pixels (14 gray paths before).
On edges blurred by a pixel or more, gray near black loses pixels. A
soft-edged gray band up to 4 pixels wide along black traces as black
and white at blurs of 1.5 and 2 pixels, and one up to 3 pixels wide at
blurs of 1 and 3 pixels. A soft-edged gray stroke 2 pixels from black
can lose some or all of its gray, and one 4 pixels away a few pixels. A
soft-edged gray stroke running out of black, blurred by 1 or 1.5
pixels, loses 3 or 4 of the 16 pixels nearest black. Sharp-edged bands
and strokes keep their gray.
Docs: docs/image-trace-adjustments.md, docs/target-app-parity.md.