crates/metafile is deleted. formats, layout, render and pdf take their
WMF, EMF and EMF+ reader from craft-metafile (craft-libs 9dcbd90c, the
rev craft-docx is already pinned at), so EMF+ records are read: an EMF+
only file's DrawImage of a 32bppPARGB bitmap draws its pixels.
render and pdf draw what the shared model adds: pattern and hatch fills,
clip regions and clip paths, pen caps, joins, miter limits and dashes,
turned, skewed or mirrored bitmaps, and the opaque background behind
text. The text layout and the default font for a file that names none
(Arial, a 16-unit cell) move to layout, next to metafile_text_run.
layout/src/metafile.rs holds the rules both outputs follow: a negative
crop side draws as no crop, as before; a cosmetic pen's shortest dash
or gap is one output pixel; a dash pattern whose period is under a
pixel, or that would split a path into more than 100,000 pieces, draws
solid; and an invalid miter limit is 10. On screen, a pattern tile whose
sides are both under 1.5 device pixels fills with its average color,
computed once a tile, and the vector cache counts each item's clip
region, clip list and dash pattern.
Fixes#6636
Document pictures now decode through wordcraft_geom::picture (behind a
new `image` feature of the geom crate). It reads the header, refuses the
picture before anything is allocated when the decode's peak would pass
the budget, then decodes with the header's size and the budget as
limits.
The peak is the larger of three moments: the decoder's output beside
what it holds while decoding, the output beside its RGBA8 conversion,
and that beside a premultiplied copy. A progressive JPEG's decoder
(zune-jpeg) holds an i16 for every coefficient of every component, in
whole MCUs, for the whole picture; the SOF header gives that size. A
7900 x 7900 progressive JPEG without subsampling needs 9 bytes a pixel
(561 MB), past the 512 MB budget, although two RGBA buffers (499 MB)
fit. With the fix reverted, the test's header-only file of that size
was decoded and the test process peaked at 542 MiB resident.
- Render and PDF export ask the probe in their wrappers (decode_pixmap,
load_image), so a picture past the budget gets the grey box or is
left out, as before.
- The crop dialog and every picture edit decoded with load_from_memory
and looked at the size afterwards. They decode through the probe now,
and so do wrap outlines and the RTF re-encode.
- A picture edit has the drawing budget instead of an 80 M pixel cap
checked after decoding. A picture past that budget (about 67 M RGBA
pixels) can't be drawn either, so no edit that could be seen is lost.
* Use chrono for UTC/calendar conversion instead of hand-rolled copies
Replaces the hand-rolled civil_from_days / days_from_civil copies and the
Excel serial-date, FILETIME and logging/PDF date conversions with chrono as
a workspace dependency (default-features = false, std only, so wasm is
unaffected). Output formats are unchanged and pinned by the existing (and one
new) test. Excel's 1900 leap-year bug and the 1900/1904 systems keep their
explicit special cases; a small proleptic-Gregorian fallback covers years
outside chrono's NaiveDate range so no input changes behaviour or panics.
AGENTS.md now says time goes through chrono.
Follow-up to storytold/cadcraft#1785
* Never panic or reshape date output for clocks outside chrono's range
chrono's From<SystemTime> unwraps, so a clock past year 262143 panicked
inside the logger. logging::timestamp now uses the fallible
DateTime::from_timestamp and reads such a clock as the epoch, as the old
saturating code did.
iso_from_unix_secs fell back to the Unix epoch for timestamps chrono
rejects; it now keeps the proleptic Gregorian date civil_from_days gives,
restoring the pre-chrono result for supported out-of-range u64 inputs.
Years past 9999 are written with {:04} rather than %Y (which prepends a
'+') in logging::timestamp, iso_from_unix_secs, docbin::filetime_iso and
odt::now_iso, so byte-for-byte output is preserved.
filetime_iso gained a docstring; AGENTS.md's chrono bullet now notes the
narrow out-of-range fallback. Tests cover each boundary.
* Don't build an unrepresentable SystemTime in the far-future test
UNIX_EPOCH + Duration::from_secs(u64::MAX / 2) overflows Windows'
SystemTime and panicked before timestamp() ran, failing the Windows
test job. Build it with checked_add and only assert where it exists.
* Route out-of-range dates through chrono's 400-year cycle, not hand-rolled copies
The PR kept hand-rolled Hinnant `civil_fallback`/`days_from_civil_fallback`
copies for dates outside chrono's ±262143-year range, and AGENTS.md carved a
"narrow fallback" exception for them. That left calendar arithmetic outside
chrono after all.
Replace every copy with the Gregorian 400-year cycle (146_097 days): chrono
converts the day/year within one cycle and whole cycles are added back via i128
(so i64::MIN/MAX can't overflow). This keeps chrono as the only calendar logic
and reproduces the pre-chrono proleptic Gregorian result for every value:
- engine cmd/mod.rs `civil_from_days` (reached by the system clock and by
`iso_from_unix_secs` for a u64 seconds count past chrono's range).
- engine cmd/controls.rs `days_from_civil` and `days_in_month` (`days_in_month`
now uses `rem_euclid(400)` since month lengths repeat every 400 years).
- formats sheet.rs `civil` and docx chart.rs `civil` (Excel serials).
Outputs are unchanged: verified the cycle result equals the old formula over a
wide range and at the extremes with a scratch chrono program. New tests pin
accepted inputs past chrono's range on both sides of the epoch, through the
public/lib entry points and the private converters, with values from the old
formula. Removed the AGENTS.md "narrow fallback" exception.
pdf/lib.rs `civil` is left as-is: its only caller is the live system clock and
its year is clamped to 0..9999 by the PDF DateTime, so the epoch fallback past
year 262143 is the accepted live-clock behavior, not a hand-rolled copy.
Tested: cargo test -p wordcraft-engine -p wordcraft-formats -p wordcraft-docx;
fmt --check; clippy --all-targets -D warnings.
* logging/docbin/odt/pdf: keep extended-year dates, no epoch fallback
FredApps (CHANGES_REQUESTED) showed `DateTime::from_timestamp(..).unwrap_or_default()`
changes output: a timestamp the old code converted past chrono's year 262143 range now
collapsed to 1970-01-01 instead of its real proleptic-Gregorian date. The same regression
exists wherever the PR swapped a hand-rolled converter for the epoch fallback. Fix every
such site to keep chrono as the only calendar logic via the 400-year Gregorian cycle
(146_097 days): chrono converts the day within one cycle, whole cycles carry in the year
(i128, saturated into i64). Verified the cycle result equals the old formula across a wide
range and at the extremes.
- apps/wordcraft/src/logging.rs `timestamp`: added `civil_from_days(u64)` (cycle), kept the
pre-1970 epoch clamp (`duration_since`) and the milliseconds/time-of-day format exactly.
Pins: UNIX_EPOCH + 10_000_000_000_000 s -> 318857-05-20T17:46:40.000Z (FredApps' example,
was 1970-...); the checked_add u64::MAX/2 s fixture now asserts 292277026596-12-04T15:30:07.000Z
instead of the epoch.
- crates/docbin/src/metadata.rs `filetime_iso`: restored a cycle-based `civil(i64)` instead
of `from_timestamp(..)?` (which dropped the field past chrono's range). A u64 FILETIME caps
at ~1.83e12 unix s (year 60056), inside chrono's range, so no reachable value changes; pinned
the largest FILETIME (u64::MAX) -> 60056-05-28T05:36:10Z to lock the extended-year path.
- crates/formats/src/odt.rs `now_iso`: reuse the crate's existing cycle converter
`sheet::civil` (made pub(crate)) rather than the epoch fallback.
- crates/pdf/src/lib.rs `civil`: cycle converter; year returned unclamped and the caller's
existing PDF DateTime 0..=9999 year clamp is kept exactly, so month/day/time of a far-future
clock survive. Pins: civil(10_000_000_000_000)=(318857,5,20,17,46,40),
civil(-10_000_000_000_000)=(-314918,8,13,6,13,20), civil(253_402_300_800)=(10000,1,1,0,0,0).
engine `cmd/mod.rs::iso_from_unix_secs` already uses the cycle fallback (last round); unchanged.
* Don't build a year-318857 SystemTime unchecked in the logging test
Windows' SystemTime only reaches about year 30828, so
UNIX_EPOCH + Duration::from_secs(10_000_000_000_000) panicked before
timestamp() ran and failed the Windows test job. Build it with
checked_add and assert where the platform can hold it. The year-10000
case fits everywhere and stays unconditional.
* Name the 400-year range extension as the one exception to the time rule
The rule said never to hand-roll calendar math, but the range fix
(shifting by whole 400-year Gregorian cycles so dates past chrono's year
range keep their old output) is cycle arithmetic, so the rule and the code
disagreed. The rule now names that exception: the library still computes
month, day and time, and only whole cycles are carried in the year.
* Say the library computes month and day; time-of-day is plain arithmetic
Narrows the range-extension wording: the date library computes the
calendar fields, and splitting seconds into hours, minutes and seconds
is ordinary arithmetic, which is what the code does.
`now_ms` in layout, `now_unix` in engine, formats and pdf gated
`js_sys::Date::now()` on `target_arch = "wasm32"`, which also matches
wasm32-wasip1 and wasm32-wasip1-threads. There js-sys has no host to call,
so it panics ("cannot call wasm-bindgen imported functions on non-wasm
targets"), and since `layout()` times itself with `now_ms`, every document
laid out on WASI aborts. `std::time` works on WASI, so these now use the
browser clock only on wasm32-unknown-unknown, the gate wordcraft-docx
already uses.
Windows metafiles (WMF, EMF), common for logos, crests and diagrams in
older documents, couldn't be decoded and showed as grey boxes, on screen
and in PDF.
The new crate wordcraft-metafile (layer 1) reads them from the public
[MS-WMF] and [MS-EMF] specifications into paths and bitmaps, and places
them in a picture's box; render and pdf paint the result as vectors.
It covers the drawing records Office pictures use (lines, polygons,
curves, rectangles, ellipses, arcs, path brackets), pens and brushes,
the mapping modes and world transforms, and bitmaps in DIB form with the
usual raster operations. Input is treated as hostile: every read is
bounded, and a byte budget caps what one picture may allocate.
Not yet: text records (ExtTextOut) and EMF+ records are skipped, so a
metafile whose content is mostly text shows only its lines and fills.
docxide-pdf bench (365 documents, vs main + #135): the twelve documents
with WMF/EMF pictures show them; the pictures are a small part of their
pages, so the scores move little (Jaccard passes 208 -> 209, five
documents up to 3 SSIM points better, none worse).
Co-authored-by: Oliver Zieschang <4963183+naxvog@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
#214 gave now_iso the browser's clock on wasm; the other clock helpers were
still stubs there: Insert Date's weekday was always Thursday, the PDF
creation date was left out, and the layout and UI millisecond clocks
returned 0, so browser AutoSave's timer never fired, the caret didn't blink
and status messages never cleared.
The engine's clock is now one now_unix() (std on native, js_sys::Date in the
browser, as #214 does) that now_iso and the weekday share; the pdf, layout
and ui-egui clocks read js_sys::Date the same way on wasm. Refs #188.
Co-authored-by: Dmitrii Khlapov <19991457+dkhlapov@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
krilla subsets fonts only when the document is finished, so one font its
subsetter rejects ("failed to subset font: malformed font") failed the
whole PDF export. Two of the 365 documents in docxide-pdf's Word
reference corpus hit this on Linux: Caladea Italic and Bold Italic
(ttf-caladea, our metric-compatible stand-in for Cambria) and Noto Sans
New Tai Lue Bold.
When finishing fails on a font, the exporter finds the face krilla
names, logs a warning and writes the document again with that face's
glyphs drawn as filled outlines (FontDb::outline), including in the
watermark; every other font is still embedded as selectable text. At
most 32 faces fall back before the export gives up with the error.
kurbo moves from dev-dependencies to dependencies of wordcraft-pdf (it
was already in the workspace).
Test: a copy of the bundled Source Sans 3 with one glyph turned into a
truncated composite fails the old export exactly as the corpus documents
did; it now exports, with no text glyphs from that font and its glyphs
drawn as paths.
Co-authored-by: Oliver Zieschang <4963183+naxvog@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>