row_names uses the shared display_name helper, so a built "Language (Region)" name is translated in
parts. The dictionaries-folder comments in prefs.rs and tests.rs say File > Options > Language.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCYmv6UJmY2QcuhfXco
Swedish (Finland) is built at run time, so no catalog row can match it. One helper
(ribbon::display_name) now translates the language, the region and the {language} ({region})
template separately and joins them; the Set Proofing Language list, the Language menu, the status
bar and the dialog all use it. English output is unchanged and matching never uses a name.
gap 8 no longer describes the missing-pack bar, which is in its own branch.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCYmv6UJmY2QcuhfXco
The Dictionaries folder moves here from Proofing with the Dictionary files list (#1063) kept. Set as
default sets only the saved default proofing language (the new pure review.language.preferred command), and
proof.packs returns the languages rows.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCYmv6UJmY2QcuhfXco
With no list of packs yet the bar offers a language with no dictionary and Install fetches the
list, then the pack. The caret's typing language counts, the spelling context menu heading is in
the interface language, and closing the bar is remembered per document, pack locale and tag.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCYmv6UJmY2QcuhfXco
Every published pack is named in English, Valencian is ca-ES-valencia, and rows are matched by
tag (language, script, region, variant), never by the name shown. One region helper
(packs::is_region, packs::without_region) next to pack_for_tag serves the engine and the UI.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCYmv6UJmY2QcuhfXco
- PACK_INDEX: language-packs-2 index.json, sha256 b35c7aef…, built from craft-i18n 20467b63
- craft-proof pinned at storytold/craft-libs 47c9f625 (flag-based compounding, #497; review follow-ups and en-AU, #496)
- Options › Language: ProofLang::EnAu has no Hunspell dictionary name of its own (English uses SCOWL)
- docs: Danish, German, Dutch, Swedish, Norwegian, Hungarian, Estonian, Croatian, Persian and Esperanto are packs; Mongolian and Thai stay hidden
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- craft-proof now names ja, zh and th in missing_dictionary: packs::missing_dictionary hides it
for a pack language when no pack for it is listed (status bar and Options).
- de-CH and pt-PT are pack languages in craft-proof (de-ch, pt-pt): names, hyphenation and
grammar stay German and Portuguese (tests); pack names fall back to the region-aware name.
- About › Language data shows craft_proof::scowl::notice() when SCOWL is embedded (test).
- Flatpak manifest's craft-i18n commit follows craft-i18n.rev.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- proof.packs reports index_override; the CLI prints the test source; the bar's Install tooltip
says it turns on downloading of fonts and Language packs, and names a test source.
- Docs: Install works once language-packs-1 is published (until then only a test index); with a
test index the sha256 only catches a broken download.
- About › Language data keeps its size: a fixed-height scrolling list, licence texts read again
only when the dictionaries change.
- The engine's Cancel is the font store's own flag (no snapshot), so a cancel stops a fetch or a
stalled download too.
- Index text shown without control, bidi or format characters; long names cut; the bar works out
what to offer again when the list changes (index generation, not its length).
- ui_shot's store is a temporary folder (or storeDir); tests reset the backend even on panic; a
fixture pack proves per-run proofing and hyphenation from a pack.
- The default-build check follows craft-proof's carry-over (de, pl, pt, it, ru embedded until packs
ship; CRAFT_PROOF_EMBED=en is English only); the Polish hyphenation check runs again.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- Install, Show available languages and Refresh turn on the font store's downloading only on the
person's own click (input_from_control, as in #959); a faked click says why instead.
- proof.installPack and proof.packs {refresh} run on the engine's thread in the app ("started"),
with progress, results and errors in proof.packs; the CLI and headless MCP stay synchronous.
- A session's layout is keyed on the dictionaries' generation, so parked documents proof again.
- proof.removePack refuses while that pack installs; pack commands that need a locale are left out
of Search; options text names Refresh; licence tooltip says Source; errors go through tl!;
the bar rescans at most once a second while typing; English names for every pack language and
sr-latn; About has one Close button; xtask doc comment back on cmd_i18n_report.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- `review.issues` and `file.options` report a run's language with `ProofLang::tag()`, so a
Pack language reports its locale, not "".
- The test-source warning reads craft-font-store's `language_pack_index_overridden()`.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- craft-proof (benandez/craft-libs 6ae4f03, storytold/craft-libs#485) and craft-font-store /
craft-font-catalog (630e42f, #486) by git rev instead of local paths.
- ProofLang::Pack: the bar offers every language with a pack (zxx/und never).
- Options › Language warns when CRAFT_LANGUAGE_PACK_INDEX supplies a test index.
- Pack `source` (upstream link) when the index gives it, else the archive.
- The bar's close button is the code-drawn close icon: "✕" had no glyph in the interface fonts
and egui panicked drawing it.
- `cargo xtask language` allows our own ru-extra, which stays embedded.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- engine `packs`: the packs Backend (the font store on the desktop and CLI, a fixed list in tests
and ui_shot), the pinned index, installs and removals through craft-font-store's
`language-pack` kind, craft-proof reading the store's `language-packs/` folder at once.
Commands open nothing and never turn downloading on (only an Install click in the app does);
confined automation can't write the packs folder outside its write root.
- ui-egui `language_packs`: File › Options › Language › Language packs (name, size, licence,
source, Install/Remove, progress, errors), the missing-pack bar (dismissible per document,
never for a language without a pack), About › Language data (full licence texts).
- CP4.4: `cargo xtask language` (in `ci`): the default build embeds only en-* Language data.
- CP4.5: docs (help, MCP, control protocol, gaps #8, localization parity, roadmap).
- Path deps to Lane 4's worktrees while developing.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
CodeRabbit on #5205:
- NOTICE: the SCOWL English dictionaries (en_US, en_GB-ise, en_CA) with
Kevin Atkinson's SCOWLv2 notice, Benjamin Titze's notice and Geoff
Kuenning's Ispell BSD licence for the affix files, as their licences ask of
distributions.
- docs/gaps.md: gap 8's evidence says each English checks its own SCOWL
2026 dictionary, with the #40 and #520 regression tests; #40 (closed) is no
longer listed as open. Status line and revision history.
Also: docs/localization-tr.md's craft-proof path change (3110dcf02, #5202)
and docs/architecture.md's no-data wording (b7dad64d9) get their status lines
and revision rows.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
craft-proof moves to 0675d58: an English no longer takes another English's
SCOWL words (en-US flagged colourblind, en-GB colorblind), British English
keeps -ize, and English suggestions drop junk splits and names.
craft-i18n.rev and the Flatpak source move to storytold/craft-i18n#114's merge
commit (059cbf9: judgment/judgement right in every English).
No wordcraft test depended on the dialects crossing
(engine 1120 + 43, layout 578 pass). localization-parity's English row says so.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
craft-proof moves to storytold/craft-libs#483's head (e6e948f, on #482's
9e5862a): each English checks against its own SCOWL rel-2026.02.25 dictionary
through the affix engine (ordinals via COMPOUNDRULE), with en-variants.txt as
the dialect policy. Lookup order: dictionaries folder, [English pack], SCOWL,
system Hunspell, the built-in Moby list; Moby stays for hyphenation.
craft-i18n.rev and the Flatpak source move to storytold/craft-i18n#109's merge
commit (efb1127), which adds the SCOWL files.
Typed-text tests pin British spellings in en-GB (#40), modern words (#520),
Canadian/Australian/NZ/Indian English (#1281, #1045) and ordinals; they skip
without the Language data, as the other data tests do. Docs: typing-parity,
localization-parity, ROADMAP progress log.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
CodeRabbit on #5202:
- AGENTS.md, ATTRIBUTION.md, README.md, docs/architecture.md: without
CRAFT_I18N_DIR English spelling falls back to an installed Hunspell English
or craft-proof's built-in Moby list, and only pattern hyphenation is off
(Bulgarian, Turkish and Russian hyphenate by rule); they said English
wasn't spell-checked. (CLAUDE.md carries the same sentence; left for the
maintainer, as agents don't edit it.)
- NOTICE: craft-i18n's patterns are language-data/<lang>/hyphenation.pat,
not en-us.pat, pl.pat, pt.pat, de.pat, it.pat.
- docs/spelling-dictionaries.md: the Flatpak's dictionaries folder is
~/.var/app/ai.storyteller.wordcraft/data/wordcraft/dictionaries; the host's
folders are outside its sandbox.
- printing.yml: the repository checkout doesn't persist credentials.
- tests.rs no_language_data: also probes the American/British variant table,
so a craft-i18n checkout missing some files skips the dialect tests
instead of failing them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
Polish/Portuguese/Czech review commands, the custom dictionary file and the two typing squiggle
tests pass without CRAFT_I18N_DIR at 48ed6f8. English dialects, regions' English and Hyphenation
Options still need the variant table and patterns, so they keep the skip.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
storytold/craft-libs#482's new head: without Language data English spells from the built-in
Moby list plus en-extra with edit-1 suggestions, and a system English Hunspell loads leniently.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
- wordcraft-proof is now `package = "craft-proof"` from craft-libs (local
path for now; TODO in Cargo.toml for the git+rev pin after the craft-libs
PR merges). `wordcraft_proof::` paths are unchanged.
- crates/proof and assets/{spelling,hyphenation} deleted with their
ATTRIBUTION.md rows; one row now points at craft-i18n's Language data;
NOTICE paths updated.
- craft-i18n.rev (and the flatpak source) moved to the craft-i18n commit that
adds language-data/ (CI sets CRAFT_I18N_REQUIRED=1, so the pin must have it).
- xtask layers: proof removed. Engine tests carry their own copy of the
synthetic Czech fixture (they reached into crates/proof/src).
- Docs: architecture.md, CLAUDE.md/AGENTS.md layer table, README build note
(CRAFT_I18N_DIR), localization-tr path, ROADMAP progress log.
Before/after wordcraft-cli on a 17-language proofing document: issues,
suggestions, manual hyphenation, PDF text and PDF bytes identical.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
`Kmm` with `km/K` suggested `Km`, which the checker flags; `Km` itself got no
suggestion. A suggestion whose dictionary form is only correct through a
KEEPCASE stem now keeps that case (`km`). Other stems are re-cased as before
(`MIXCASX` → `MIXCASE`).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01FypgCX8LsYYE8fsj4byBm2
dpkruse's Genuine Steps Certificate: a bulleted text paragraph closes a
two-column continuous section, then an empty bulleted one closes another.
Word draws no bullet for the empty one; it is now quiet (takes no room) and
draws no label, still counted. A stacked mark that doesn't fit keeps its
line for the caret but draws nothing below the margin (CodeRabbit).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
CodeRabbit (local): keep-lines compared its height with the room below the
mid-page section start, so a paragraph that fits on a new page split instead.
It now compares with the next column, or the next page's body when this is the
last column. A stacked mark whose page break before starts a page no longer
gets the previous page's bottom back.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
A continuous section moves the column top down below the text above. at_top()
then held at every new section start, so a paragraph that did not fit was
placed anyway, as on an empty page. A run of one-line continuous sections ran
off the bottom of page 1 (100 of them reached 2,480 pt down a 792 pt page).
The overflow checks for paragraphs now use fresh(): at the top with nothing
above on this page. at_top() still drives the spacing rules, unchanged.
Empty paragraphs that close a continuous section right after another
section's mark took no room. Measured in Word on synthetic files ("Start", N
such marks, "End"): 10 put "End" about 9 cm down page 1; 50, 100 and 200 all
put it on the first line of page 2, the same with 24 pt marks or no paragraph
spacing. So each takes its line while it fits, then stays at the page's
bottom. A mark right after text stays quiet, as measured before.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
Noto Sans Thai and its siblings have no Latin letters or digits, so on a
machine whose only Thai fonts are Noto no face covers the test run and
fallback_for_run rightly returns None. The guard now asks whether the Thai
substitute covers the whole run, which doesn't go through fallback_for_run.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
Unescaping and escaping again turned and 	 into raw characters that read
back as spaces. The raw value of well-formed markup holds no `<` and every `&` starts
a reference, so only a `"` from a single-quoted value is replaced with ".
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01HUodYU9r4mPHed4pPsjHC2
- A package storing WORD/FONTTABLE.XML (part names are case-insensitive) loaded no
embedded fonts and a save dropped the table and every font: lookups used the exact
constant spelling. They now find the key case-insensitively.
- scrub_private re-pushed attribute values raw (quick-xml push_attribute does not
escape), so a value read from single quotes became malformed XML. The value is now
unescaped and escaped again.
- rel_targets matches _rels/ case-insensitively.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01HUodYU9r4mPHed4pPsjHC2
- docs/zotero.md and peer.rs module docs: a read error, or an open error other than NotFound,
refuses (sandboxes denying /proc/net/tcp: AppArmor, Snap, Flatpak, are refused).
- read_line_capped retries ErrorKind::Interrupted like std's read_line.
- New scan_tables (what owner_check calls) takes the table opener; the 70 MiB test drives it,
so reinstating take(64 MiB) there fails the test. Added tests for EINTR and a denied open.
- Row parsing uses split_whitespace().nth() instead of collecting a Vec.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The Linux same-user check read /proc/net/tcp{,6} through a 64 MiB take(). Hitting the cap
was silent, so a local user binding ~450k loopback listeners pushed the connection rows
past it; both uids came back unknown and the "trust loopback" arm let their fake Zotero
listener on 23116 through. A truncated last line could also glue to the next header.
Now each table is streamed line by line (BufRead, nothing kept, no total cap, lines capped
at 4 KiB) and stops once both rows are found. Any read error refuses. "Unknown, trust
loopback" remains only for a listener absent from a table read in full (WSL2, proxies).
Tests: rows after 70 MiB of filler, and a read error mid-table.
rich.rs: document that the MAX_RUNS cap still allows one first run per paragraph (bounded
by MAX_PARAS).
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Wall-clock bounds of 2 s failed on loaded runners. Each bound is now 10 s, at least 10x the
fixed code's time and well under the old code's on the same input (measured by running the
pre-fix code under the same load):
- many distinct media: 0.5 s now; the old scan of every image took 131 s.
- object walks: under 0.3 s now; the old walks took 360 s (#3614).
- delete across lone control markers: 3,000 markers (was 1,500); 1 ms now, the old delete
took 18 s at 1,500 and 114 s at 3,000.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
- 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) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
Finding 80 (CWE-400): adopting a fragment deep-copied a box's source story
once per text box showing it, including boxes inside the story being copied,
so 4096 boxes (the story budget) pointing at one large story copied it 4096
times. Now: a box showing a story it is inside gets no story (a cycle);
blocks without text boxes are shared with the source (copy-on-write Arcs)
instead of copied; and a paste adopts at most MAX_PASTED_BOX_BYTES (32 MiB,
estimated text + runs + objects) of box content, charged before copying,
after which later boxes keep no story, as when the story budget runs out.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 71 (CWE-407): loops of object_offsets + object_at recounted the
objects before every offset, so walking a paragraph's objects was quadratic
(40k markers: 360 s for the regression test before, 0.2 s after).
Paragraph::objects_with_offsets yields (offset, object) in one pass, and
every such walk in crates/doc uses it: the field scan (field_ranges, run on
every docx load), control events, bookmarks, perm ranges, text_box_anchor,
past_field_ends, typing_pos and the markers a range delete keeps. Also:
perm_ranges keeps open ranges per id instead of searching a list for each
end; kept_control_markers pairs each container with its table wraps once
(was once per marker) and probes a set instead of every range.
Part of #3614: the engine, docx reader and UI callers it lists live outside
crates/doc and still use object_at in loops; they can move to
objects_with_offsets separately.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 48 (CWE-407): P::word scanned and copied the rest of an alphabetic run
for every letter that didn't start a function name, so a 20,000-letter
equation did ~200 million steps outside the parser's step budget (5 parses:
8.3 s before, well under 0.1 s after). The scan now stops one past the
longest name in FUNCTION_NAMES (computed at compile time); a longer run is
never a name, as before.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 45 (CWE-407): Document::add_media compared each new image with every
stored one, so an import with many distinct small images did quadratic work
(40k images: 213 s before, 0.26 s after). A MediaIndex cache (not saved,
ignored by ==) maps bytes to their first key, is rebuilt when media changed
size behind its back, and every hit is checked against media. Generated
image{n} names also skip taken names once instead of on every add.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Folding zip entries whose names differ only in case now keeps the
spelling a relationship part targets, and the first in byte order only
when no relationship names any of them. The fold logs one warning per
package, not one per entry.
Limits that leave data out now say so, once per document or save:
- a chart past the chart limits on save (series or points dropped);
- embedded fonts past the byte budget (one warning with the count);
- a kept glossary that can't be cleaned of personal information and is
written from the model instead.
Tests capture the warnings (crate::test_log) and check each is logged
once.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
Ultrareview finding 1. With "remove personal information" or "remove date
and time" set, the glossary was always written from the model (01c78899),
which holds only the entries the reader took: one without a name, past the
entry cap or with a bad content type was lost, and with none left the whole
glossary part went. The kept parts are now written with the same names and
dates removed that the writer leaves out of the main document (annotation
authors become "Author", comment initials "A", annotation and UTC dates
go; a people part's authors and presence too), rewritten with quick-xml and
matched by namespace. A part that can't be rewritten (malformed, UTF-16)
falls back to the model, so the privacy guarantee holds either way.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 73: Package lookup matched part names case-insensitively, but relationship targets kept the spelling they were written in, so every casing of one part became its own media entry, graphic parse and passthrough key, and a save wrote a full copy per spelling (64 spellings of a 64 KB picture: 64 media parts). Package now keeps a lowercase index, Package::canonical returns the stored spelling, and rels() resolves internal targets to it, so every downstream cache is keyed by the one part; get_shared uses the index instead of a linear scan.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 55: flat.rs xml_fragment inlined kept part bytes into pkg:xmlData after a prefix check only, so a part whose text closed the wrapper (</pkg:xmlData></pkg:part><pkg:part ...>) wrote parts of its own into the saved .xml - e.g. a second core properties part. Now the bytes are inlined only when quick-xml reads them as exactly one balanced element with nothing but whitespace, comments and PIs around it; anything else is written as pkg:binaryData and still round-trips.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 76. chart_xml trusted its ChartSpec to be sanitized, but a .json
document is deserialised without ChartSpec::sanitize, and every series
repeats every category, so a chart past the limits multiplied into a
gigabyte-sized part on save. chart_xml now writes a bounded, sanitized copy:
only the series and points the limits keep are copied before sanitizing.
(Sanitizing on .json open, and SmartArt, are in the engine and not covered
here.)
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 72. An unchanged glossary was carried byte for byte, judged by a
fingerprint of its entries that leaves out the document's privacy settings,
so a save with "remove personal information" or "remove date and time" kept
the authors and dates of the glossary's tracked changes. With either setting
on, the glossary is now written from the model, which applies them.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 51. embedded_fonts copied a part's bytes for every font-table
reference, so the 64 references allowed to one 256 MB part could make 16 GB of
copies on open. A reference repeating one already taken (same part, key and
name) is skipped, and copies stop at MAX_FONT_BYTES (one part's limit),
checked before each copy is made.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
w:id isn't a number against every id mapped so far, so a comments part of
many non-numeric ids took time quadratic in their count (49,000 comments:
2.4 s, now 0.35 s in a debug build). The ids mapped before the part are
collected into a set once; ids given in the part are already in `entries`.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Finding 43. Package::open_limited dropped an entry whose read failed before
adding the bytes it had already inflated to the running total, so many
entries with a wrong CRC (checked only after the whole stream inflates) could
each inflate up to the part limit for free. The bytes are now charged first
and the size check runs before the read result is looked at.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>