view.proofSetup.save refuses a setup whose file would be larger than
the 64 KiB view.proofSetup.load reads (a device profile with a very
long name), and writes nothing.
Shift held once an artboard move or Alt-drag copy is under way keeps it
to the level, upright or 45° line from General › Constrain Angle, as a
Selection tool move does. A Shift-press still draws an artboard inside
another or marquee-selects (#2497). Smart Guides, the grid and whole
pixels slide the move along its line, and the cursor stays the move
cursor during a move.
When ten or fewer fonts match, adding more to the exhausted list returns
the same list as a new search, so the null more assertion could not tell
the two apart. The test now adds a font to the list before the null more
call, and only a new search drops it.
The Control bar label's request to scroll the Links panel to the selected
image now carries the document it was made in. A draw of another document
drops it and leaves the Show filter as it is, and every draw that reads the
request clears it, including one with an empty list.
retype.find reads `more` with opt_bool, which treats null as omitted, so
`{more: null}` runs a new search. The params text said any value other
than true or false is an error. The params text, its MCP mirror and
docs/mcp.md now say that false, omitted or null runs a new search and
that a value other than true, false or null is an error. A test checks
that a null more lists the first ten again.
Follow-up to #8400. A Perceptual gradient expands each segment into
stops on its OKLab curve. Between two stops of the same color the round
trip through OKLab rounded the color, so a stretch between two
paper-white stops came out as nearly white RGB. On a PostScript plate
the separator then saw ink there, and an overprinting gradient painted
over the ink below across that stretch.
Gradient::expanded now gives a stop between two stops of the same color
that color, unchanged, for every method. GradientMethod::mix and
Gradient::sample are not changed.
On macOS the temp folder /var/... is a link to /private/var/.... Kept
script file grants hold real paths, and ScriptScope refuses a path by
its names alone when they don't start with a granted path, so the 11
grant tests that name their files under the temp folder were refused
and asked again. folder() now returns the real path of the folder it
makes.
ui_inspect_script_report_leaves_out_files_outside_the_read_root failed
on macOS for the same reason: the report names the included file by
its real path, and the test compared it with the /var spelling. It now
compares with the real path. No change outside the tests.
Fixes#8693
Swap Fill and Stroke on a whole Live Paint group compared the faces'
fills without gradient placement, so faces with freeform gradients
counted as one shared fill and the edges took a freeform gradient
without its points. A freeform gradient on any face or edge now counts
as a paint the parts do not share, and the group keeps its paint.
- docs/mcp.md says that a face or edge given with its whole group follows
the group's rule, and that Default Fill and Stroke replaces the parts'
appearance also when the Default Graphic Style is redefined.
- The progress-doc line states the behavior as VectorCraft's own rule and
lists all eight tests.
- Tests cover freeform faces in Swap, an edge given with its group, and
Default Fill and Stroke with a redefined Default Graphic Style.
RenderOptions::hidden was a Vec<NodeId>, and the renderer checked each
node it visits against it with contains (node_checks::left_out and
Renderer::draw_node), so every check could compare against all k
hidden ids. The Fold Preview print hides every non-printing layer and,
with Show Fold Lines off, every object with a line role.
hidden is now a HashSet<NodeId>, so each check is one hash lookup. The
Fold Preview collects print_hidden's list into the set, and two tests
that built the field with vec! build it from an array. The field's doc
comment now names the Fold Preview print as its example.
New test a_render_leaves_out_every_object_of_a_large_hidden_set renders
2,000 strips with 1,000 of them in hidden and checks every strip's
pixel. It checks that hiding still works for a large set; the old Vec
drew the same pixels, so the test does not fail at run time without
the change.