Reword decode16 and DdsCompression::None docs, check the 16-bit buffer
length before narrowing for BC1-BC3, and test typeless DXGI codes, padded
16-bit rows, G16R16 masks and decode16 of BC1.
A no-op v17 -> v18 step so FilmCraft can save per-clip camera raw
Source Settings (Interpretation::raw) and older builds refuse such
projects instead of dropping the settings on save. Library part of
storytold/filmcraft#1078 (Part of storytold/filmcraft#1065); the MXF
CinemaDNG part already landed in #1145.
encode16 writes 16-bit pixels uncompressed as R16_UNORM, R16G16_UNORM or
R16G16B16A16_UNORM with the DX10 header, and fits BC4/BC5 endpoints and
indices against the 16-bit values; BC1-BC3 round to 8 bits first.
Texture::decode16 reads these formats (and legacy FourCC 36, wide channel
masks, BC4/BC5 interpolated at 16 bits) without losing precision.
- A bitmap that the item or byte limit refuses is counted in
dropped_images once, however many pieces a clip cuts it into.
- An EMF+ pen whose texture brush was not read is counted in
dropped_images, as a lost texture brush object is.
- An EMF+ clip path or region path whose figures all lie on a line
encloses no area, so it gives an empty area instead of its bounding
rectangle.
XML readers skip an element, attribute list or notation declaration to its
first `>`, quotes and brackets included. The guard's scans now end these
declarations the same way. A `]` inside `<!ELEMENT svg ] >` no longer ends
the internal subset for `too_wide` and `prescan`, and a `[` in one no longer
lets a later `<!--` hide the entity declarations after it from `declared`.
sniff_delimiter reads the first 64 KiB of the text and leaves out a
record cut off at that point (#1209). When the first record is longer
than the sample, no record is counted and the function returns comma,
so a semicolon or tab file with a very wide first row opens as one
column.
When no candidate delimiter finds coherent complete records with a
delimiter in each, a candidate that finds no complete record now
counts the delimiters read so far in the first record. Delimiters
inside quoted fields are not counted. A quote inside a field that does
not start with one makes that record incoherent, and the coherent
candidate is picked. When one of those candidates ends the sample
inside a quoted field, each of them counts only up to that field's
opening quote. The per-candidate loop moved into a scan_delimiter
helper, which runs a second scan over the text before that quote.
declared() skipped a comment or CDATA section wherever `<!--` or
`<![CDATA[` appeared, including inside an entity's quoted value, another
declaration's literal or a processing instruction. With
`<!ENTITY c "<!--">` before a large entity, the scan skipped to the end
of the file, the large entity was never recorded, and
entities_expand_past let its references through to the XML reader.
- declared() reads a declaration's quoted literals and a processing
instruction as text: a comment or CDATA marker there starts nothing,
and the declarations after it are still read.
- prescan() counts the markup in an entity's value the same way, and a
comment, CDATA section or instruction that closes inside the value
still opens and closes no element.
- too_wide() and prescan() find the end of the internal subset past
quoted literals, comments and instructions, so a `]>` in a value no
longer ends the document type early.
TeX skips the blanks after a control space, `\ ` or `\` and a tab, as it
does after `\` at a line end. The lexer left them to the whitespace branch,
which added a second space, so text taken as written, such as a reference
or an `\item` label, read `one two`. The control-space branch now skips
them the way the line-end branch does, and a blank line after one still
ends the paragraph.
A caller could only tell from skipped_records that part of a file was
not read. Picture now also reports:
- truncated: the parse stopped at a limit that ends it (items, bytes
kept, SAVEDC depth, clip paths), in the file or in an EMF+ metafile
image drawn in it.
- dropped_images: images that were not read: bitmaps past a size, byte
or decode limit or not decodable, EMF+ metafile images that do not
parse or are nested too deep, and pattern or texture brushes whose
bitmap was not read. An image with no pixels, or drawn from a source
rectangle that selects none, is not counted.
- ignored_clips: SELECTCLIPPATH with RGN_OR, RGN_XOR or RGN_DIFF, EMF+
clips combined with XOR, exclude or complement, EMF+ unions drawn as
the rectangle around both areas, and EMF+ clip paths and regions
whose area stands in for the one the file gives, when the clip that
results can differ from the one the file asks for.
- unplayed_raster_ops: bitmaps drawn as plain copies because their
raster operation is not SRCCOPY, SRCAND, SRCPAINT or SRCINVERT, and
operations without a source other than PATCOPY, BLACKNESS and
WHITENESS.
Behavior changes:
- An EMF+ metafile image adds its own counts, skipped_records included,
once, the first time it is drawn. A nested picture without a size
still draws nothing, and its counts are kept.
- A raster operation that takes no source no longer draws the record's
bitmap. PATCOPY, BLACKNESS and WHITENESS fill the destination in
STRETCHDIBITS and STRETCHDIB, as they already did in BITBLT,
STRETCHBLT, DIBBITBLT and DIBSTRETCHBLT. Other such operations
(DSTINVERT, PATINVERT) are left out and counted.
Three ways the lexer lost a blank line, so two paragraphs ran together:
- A comment took its line end with it, so a blank line right after it
read as a single line end (wordcraft#6350).
- A control word took the spaces and one line end after it, so a blank
line right after `\relax` at a line end read as a single line end
(wordcraft#6366).
- Only LF was a line end. TeX ends a line at LF, CR or CRLF. With CR line
ends every paragraph break was lost, and a comment or a verbatim
environment ran to the end of the file and dropped the rest of it; with
CR or CRLF a control word kept the line end after it as a space
(wordcraft#6365, #6373, #6374, #6375). The lexer now reads every line
end as LF.
Each symptom has its own test. The CR fix needs the control-word fix
before it: CRLF files read a blank line after a control word correctly
only because the CR stopped the control word from taking the line end.
Port of storytold/wordcraft#6387, whose files moved here with the LaTeX
reader.
The control symbol branch took the line end or tab after a `\` as the
symbol's name, and the parser drops unknown commands, so `mid\` at a line
end followed by `line` read as `midline`, and a blank line after it no
longer ended the paragraph. TeX reads `\` before a line end or a tab as a
control space, as for `\ `.
- lex.rs: `\` before a line end (LF or CRLF) or a tab gives `Tok::Cmd(" ")`
and leaves the line end in place, so a following blank line still ends
the paragraph and the line end adds no second space.
- parser.rs: an optional argument such as an `\item` label reads a
control space as a space instead of writing `\ ` back as text.
This also keeps apart, on reopening, a paragraph that ends in two spaces:
export writes its last space as a control space at the line end.
Port of storytold/wordcraft#6298, whose files moved here with the LaTeX
reader.
`\\*` is a line break that also forbids a page break after it. The lexer
read `\\` as the control symbol and left the `*` as text, so `a\\*b`
showed `a`, a line break, then `*b`, `\\*[3pt]` showed `*[3pt]`, and in
a tabular the next row's first cell started with `*`.
The lexer now takes the star with `\\`, so the parser and the tabular
reader see the line break they already handle, with its optional
spacing argument.
* ci: run the Windows gate's tests with nextest and link with rust-lld
The Windows leg of the Workspace gate takes 55-100 minutes. About half of that is
cargo test running ~370 test binaries one after another, twice (default and all
features), and another 15-17 minutes per feature set is the uncached test build,
where link.exe links every test executable.
- Windows runs the tests with cargo-nextest (parallel across binaries) and the
doctests with cargo test --doc. The same tests run; .config/nextest.toml puts
the tests that relied on in-process mutexes (craft-ui GPU, craft-video h264
fixtures, craft-print) in one-at-a-time groups.
- Windows links with rust-lld via CARGO_TARGET_X86_64_PC_WINDOWS_MSVC_LINKER.
Linux and macOS legs are unchanged.
* ci: keep link.exe on the Windows gate
The first run with rust-lld built the tests no faster than link.exe at the same
runner speed (14.4 min against 15.1 min on a runner 8% slower at clippy), so the
linker change is not worth its risk. nextest stays.
* svg: floor declared() name window to a character boundary
If the MAX_NAME+1 cut split a UTF-8 character, get(..bound).unwrap_or(rest)
fell back to scanning the entire suffix. Floor the window instead, and
when recording an overlong name extend to the next character boundary so
the key is still longer than MAX_NAME.
* test(svg): assert declared() keys stay bounded after a multibyte cut
The overlong-name refusal alone also passed on the old unwrap_or(rest)
fallback, which recorded the whole suffix. Bound recorded keys to
MAX_NAME+4 so the test fails if the suffix is stored again.
camera_raw_mode read a missing SubIFDs offset as 0, so it walked the TIFF
header as the pointer array and could classify a malformed TIFF as RAW.
Skip the entry instead, matching LightCraft's classifier.
* wordmodel: keep the kashida length of a justified paragraph
ST_Jc lowKashida, mediumKashida and highKashida were all read as
Align::Justify, so the length was lost. ParaProps and ResolvedPara now
keep it; a direct both/justify cancels an inherited length. The DOCX
reader and writer round-trip the three values.
* docx: keep a style's kashida length when writing justify
w:jc encodes alignment and kashida together. A paragraph or style
with Align::Justify and kashida unset still inherits a length from
the style chain; writing "both" would read back as Some(Kashida::None)
and cancel it. Use the resolved length in that case.
A defined name's formula may use a structured reference (PtgList). Names were decoded inside
xlsb::workbook, before Package::workbook indexed the table parts, so such a name was left out
and the cells using it kept only their saved results. xlsb::workbook now returns the names
undecoded (Names), and Package::workbook decodes them with xlsb::read_names after
index_tables. Regression: a_name_using_a_structured_reference_is_decoded_after_tables_are_indexed
fails without the reorder.