* CI: FreeBSD builds with two jobs
The nightly FreeBSD run has been killed for memory (exit 137) while
compiling the workspace's test targets, on both architectures, every
night since at least 9 November. Two rustc processes at a time fit in
the VM's 12 GB.
* CI: one FreeBSD build job at a time
With two jobs the engine and ui-egui test builds were still killed when
they compiled together.
The Japanese row in docs/localization-parity.md said every message is
translated, and the Japanese catalog line in docs/target-app-parity.md
said 38 messages were missing, one of them waiting on #7214, which has
merged. On main the untranslated report lists 98 messages without a
Japanese row, 77 of which every complete-message language lacks. Both
lines now give the counted rows and the report's count, and name the
open pull requests that translate part of them.
A block's line took the tallest ascent and deepest descent of its
characters, or the height of the stack's em boxes, whichever was larger.
The top block line's characters reach above the stack by their ascent
past the em box, and the bottom line's below it by their descent, so with
a wide Line Gap the line was too short for them. The line's ascent and
descent now also take in each character's reach from its own block line's
baseline.
A WMF without a placeable box picks its scale from the map mode its
drawing records are played in. Lines and shapes drawn under R2_NOP or
R2_XORPEN, and PATBLT records the reader does not fill, no longer count,
with the mix mode saved and restored with SAVEDC and RESTOREDC.
An open stroke that reaches far outside the picture is left out before
it joins the strokes waiting to be joined, so the strokes drawn before
it are still added.
A fill in a glyph's clip marked the clip as filled before the fill was
added. When the fill painted nothing (a palette index past the palette,
or a radial gradient that draws no circle), the next glyph painted in
the same clip became the clip's only child, and closing the clip kept
that glyph on its own, so it drew outside the clip.
The clip is now marked filled only when the fill adds the clip's first
child, and a fill that adds nothing leaves the mark as it was.
A clip group with a fill or stroke of its own and no geometry effect is
baked through bake_pieces, which baked the clipping path like any other
member instead of applying the geometry effects in it as the canvas clips
by it. A compound clipping path whose member has a Transform effect with
copies then baked into nested compound paths, and SVG, PDF and the other
exports dropped the moved member and its copies from the clip.
bake_pieces now applies with_clip_effects_applied to the clipping path
first, as the plain clip group route already does. Clipping type keeps
its own handling.
A plot of the art's bounds now leaves placed images, objects at 0%
opacity and unpainted paths out of the art of a dynamic symbol, as it
does for other symbols. An object that an instance's fill, stroke or
opacity makes the plot draw is measured for that instance only, and an
instance whose symbol holds nothing else is left out, so a document
with only such instances is refused.
A PDF/A-2b export at PDF 1.4 with Create PDF Layers and a page longer
than 14,400 pt is written as PDF 1.6, which has PDF layers. The export
decided which layers place their instances' symbol art before it raised
the version, so a hidden layer, written as a PDF layer that is off, kept
its instances' art without the art's live effects.
The version a long page needs is now settled before the document is
baked. The pages come from the artboards, which baking leaves as they
are.
oversized_tiff now also refuses a TIFF whose JPEG-compressed strip or
tile (Compression 7) declares a frame wider or taller than the strip or
tile, one with more than MAX_TIFF_CHUNKS strips or tiles, and one whose
JPEG headers take more than 16,777,216 steps to read. It reads BigTIFF
directories, follows the layout and the repeated tags the decoder
reads, and refuses offsets of types it does not read.
The check before a PNG or JPEG is embedded as it is read the profile's
color space only from a profile of at least 128 bytes, so a gray or RGB
profile cut short after its color space reached the PDF writer. The
color space is now read from bytes 16..20 of any profile that holds
them, and valid_icc rejects the short profile.
BFlat