Commit Graph
6 Commits
Author SHA1 Message Date
.CooCoo c419985c54 Tests: the balloon anchor test counts work steps instead of timing (#5608)
anchors_walk_dense_unrelated_markers_in_linear_work compared CPU times of
a few microseconds, so cache and timer noise could push the ratio past its
24x limit on a linear walk (43.52x in a CI run).

wordcraft-doc now counts, in tests only, the characters its paragraph
object lookups walk (objects_before, object_offsets and
objects_with_offsets), through a work::count helper behind the existing
test-support feature. The test asserts that 8x the markers take at most
12x the steps, as the layout linear-time tests do: the walk takes 2001 ->
16001 steps (8.0x), and the object_offsets + object_at walk it replaced
takes 2003001 -> 128024001 (63.9x).
2026-11-05 01:32:10 +01:00
242e850aaa #3660: refuse a package that declares too many zip entries before zip parses them (#5260)
* #3660: refuse a package that declares too many zip entries before zip parses them

zip::ZipArchive::new parses and keeps every central directory header
the end record declares before its caller can look at len(). A ZIP64
end record can declare one per 46 bytes of the file: a 47 MB package of
one stored file listed 1,000,000 times takes zip 3.0.0 about half a
second and a large transient allocation, and len() then says 1, since
the copies share a name, so the docx reader's MAX_ENTRIES cap never
fires. ODT and spreadsheet packages had no cap at all.

wordcraft_doc::zip_guard::check finds the end records the way zip finds
them (backwards; the ZIP64 record from the locator's offset on; a plain
record when it has no entries or a central header before it) and refuses
a package declaring more than the reader's limit. The docx package
reader, the ODT reader, the spreadsheet reader (mail-merge data) and the
template-title probe call it first, with 20,000 entries as the limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* formats: keep the extensions doc line on EXTENSIONS

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
2026-11-02 04:58:18 -05:00
Martin BallandClaude Opus 5.5 bec112a4b8 #3692: review — pasted text boxes skip only the box over the budget, and say so; media_index private; kept markers' order pinned
- A text box whose story doesn't fit what is left of the paste's byte
  budget gets no story, but later boxes with smaller stories are still
  copied while the budget lasts (it used to zero the budget). Each
  source story is costed once per paste, so skipping a big story many
  times doesn't walk it again. The paste logs one warning with the
  number of boxes left empty. wordcraft-doc now depends on log.
- Document::media_index is a private field and MediaIndex a private
  type: a lookup cache, not API. The one struct-update literal outside
  the crate (a formats test) builds the document field by field.
- kept_control_markers documents that delete_range relies on its
  sorted, deduplicated output; delete_range sorts it if it ever isn't,
  and control_markers_kept_are_sorted pins it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:03:06 -04:00
Martin BallandClaude Opus 5.5 17d19b42c5 #3642: review — the DIFAT sample is test-only, valid .doc/.xls files pass the guard, and budget drops are logged
- repeating_difat_sample is built only for tests: this crate's, and the
  readers' through a `test-support` feature they turn on as a
  dev-dependency; it is no longer in the production API.
- Tests write valid compound files with the cfb crate in Word's and Excel's
  shapes (WordDocument + table stream with Data, summary and ObjectPool
  streams; Workbook with summary streams), in 512- and 4096-byte sectors,
  plus an 8 MB one whose FAT sectors spill into a DIFAT chain; each passes
  cfb_guard::check and opens (no Word- or Excel-made files are added).
- The .doc work budget notes what it leaves out (floating objects, link
  targets, alt text, Prm formatting) and logs one warning per kind once the
  document is read (#3692).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 23:47:39 +00:00
asystemoffieldsandClaude Opus 5.5 5795259c45 File: recovery copies and .json files store pictures as base64 (#586 follow-up)
Native JSON wrote Document::media and Document::passthrough with serde's
default for bytes, an array of numbers. With a synthetic 27 MiB picture a
recovery copy (compact) was 96.4 MiB and a .json file (pretty-printed,
one number per line) 285.5 MiB.

Each part is now one standard base64 string (doc::byte_store, using the
workspace base64 crate), encoded straight into the writer. Reading accepts
base64 or the old array of numbers, so recovery copies and .json files
written before still open; anything else is an error, never a panic.

Measured (release, same synthetic picture): compact 96.4 -> 36.0 MiB,
pretty 285.5 -> 36.1 MiB, write 252/614 -> 52/57 ms, read 404/533 ->
46/37 ms. A recovery snapshot of a 6,000-paragraph document with the
picture: 99 -> 39 MB, worker time 2.3 -> 0.26 s, start-up scan 1.03 ->
0.14 s.

Tests: byte_store unit tests (round trip, old arrays, hostile values);
recovery_format.rs native_json_stores_bytes_as_base64,
native_json_with_byte_arrays_still_opens, native_json_with_bad_bytes_is_an_error,
ignored measurement native_json_size_with_a_large_picture; desktop.rs
copies_with_byte_arrays_written_before_still_restore.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-15 10:10:23 -04:00
Brandon ThomasandClaude Opus 5.5 d1e80df4da M0.1–M0.6: workspace, geom, document model, fonts, layout, render
Clean-room foundations for WordCraft: Arc-shared document model with
paragraph/run editing, styles, lists, sections and tables; Word-style
font substitution; first-fit line breaking with tabs, lists, alignment,
pagination, tables, headers/footers and hit testing; vello_cpu page
rasteriser.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-05 22:30:16 -04:00