* chore: resolve SonarCloud maintainability issues
- Replace lambda closures with method references (S1612)
- Replace wildcard imports with explicit imports (S2208)
- Refactor complex functions with extracted helpers (S3776)
- Replace OR pattern with range pattern in math parser (S9047)
* style: apply rustfmt to warichu helper
* fix(compose): revert slice conversion in tests and remove unused var in warichu
* Group warichu row layout scalars to satisfy S107 parameter limit
* style: satisfy rustfmt in warichu.rs
---------
Co-authored-by: Dylan Taylor <dylan@craftmaker.dev>
Reject a non-null Prev entry that cannot be read as an offset instead of silently treating it as the end of the guarded trailer chain. Keep absent/null end markers and readable numeric offsets unchanged.
Reproduced CodeRabbit discussion_r4366245854 with an independent probe and a failing native regression before the fix. The synthetic regression covers seven unreadable value types plus absent/null and integer/real history controls.
Validation: all 9 cargo xtask ci gates passed (3063 tests passed, 6 skipped; 16 WASM-safe crates). One unrelated GPU test was labelled leaky by nextest in that run and passed cleanly in an isolated rerun. Independent checks: 17 original supported PDF variants still pass qpdf, logical preservation and Poppler pixels; 6 original refusal cases remain refused. All 11 additional Prev controls pass (7 refused, 4 accepted).
Integrate current main (d8d78851) without rewriting the contributor's commits, retaining craft-libs 9f75d3dc.
Accept literal null trailer entries while rejecting actual or ambiguous encryption/hybrid markers. Structurally inspect bounded xref/trailer history instead of adopting trailer-looking comments. Preflight all selected source variants before editing/undo; preserve reset-to-defaults. Suppress stale unrewritable canvas overrides rather than revealing source content. Keep conservative all-content crop fallback without overcropping layerless/null stubs.
Validation: cargo xtask ci, all 9 gates passed; 3062 tests passed, 6 skipped; all 16 WASM-safe crates checked. 17 supported variants passed independent qpdf, logical preservation and Poppler pixel checks; 6 unsupported/repair-only cases refused. Direct-harness Claude Opus 5 reviewed the exact main-relative delta with no blocking findings.
No native-format changes or unrelated layer features. Synthetic regression fixtures are generated in test code.
* Anchored objects: Relative to Spine mirrors custom positions on right-hand pages
* Anchored objects: page side from the frame's spread position; footnote cache keyed by page side
* Links: use resolve_absolute for stored paths
* Links: check for a missing linked file with is_ok_and
* CI: release only with git dependencies from storytold
The winit patch from skylartaylor/winit at its pinned rev is the one other source allowed.
* Links: Relink to Folder only uses a name that stays a file name in the folder
* CI: the git-sources check reads Cargo.lock as TOML, needs it current and refuses config patches
* CI: the git-sources check parses .cargo/config and .cargo/config.toml for source replacement, patches and paths
* Links: rustfmt
* Links: Relink to Folder only takes a file directly in the chosen folder
* Links: Relink to Folder compares folders as the file system resolves them
A path passes when it lies inside one of the roots of its kind. The roots'
common parent, a sibling named like a root, `..` or a link out of them and
the other kind's roots stay refused; a folder given twice counts once.
New packages are created with the ordinary 0666 request; the process umask
decides the effective mode. Annotated with that reason so the analyzer's
world-writable warning does not change the intended sharing behaviour.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>
Autoflow (Shift-click, autoflow on Place, text.thread with autoflow) inserted
new pages right after the story's last frame, in front of the document's own
pages. As in InDesign, it now threads one frame per column on each page that
follows, whatever else the page holds, and only then adds pages at the end of
the document. It fills whole pages, so an estimate that stops inside a page
can't leave its later columns out of the thread. Frames it set past the end
of the text are removed again; the pages they were on stay. Smart Text Reflow
of the primary text frame keeps adding pages after the story's last page.
Shrinks gap 26: flowing into the primary text frame of the following pages
is left.
A paragraph attribute records that the Bulleted or Numbered List button set
the hanging indent; turning the list off takes the indent away only then
(and while it is still as applied), and clears the mark. A paragraph whose
own indents were 18/-18 before the list keeps them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Paste: other apps' text goes into a text caret even after copying objects (#212)
* Paste: text that matches a copied frame still pastes; groups with text frames aren't anchored (#212)
A selection ending right after a return (a whole paragraph, as a triple
click selects it) applies the list to that paragraph only, and the button
shows pressed for it; a whole frame still takes in its last, empty
paragraph. Turning a list off clears only a hanging indent set on the
paragraph itself: one the paragraph style gives stays.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nothing checked type size. `type.char {"size": 0}` set type with no size at
all, a negative size went in the same way, and 1e9 pt asked the composer for a
line a million times the height of the page — all of it silent, and all of it
saved to the document.
The range is the one this engine already applies to a ruby annotation's size,
0.1 to 1296 pt, which is InDesign's: "size 0: 0.1 to 1296 pt".
It sits in `validate_char_fill`'s place in the chain, so every command that
takes character attributes gets it: `type.char`, `find.change`, the four
`style.character` / `style.paragraph` create and edit commands, and the two
table paths that already validated a fill.
The test covers five refused sizes, the three in range including both ends, and
two of the style commands. Without the check, `{"size": 0}` applies again.
Not from a ticket: found probing `type.char`.
`anchor.create` took any name, including one another anchor already had, so a
document could hold two anchors called "Photo 1". Every reference to the name
then means whichever the search finds first: `xref.insert {"anchor": "Photo 1"}`
and `hyperlink.create {"anchor": "Photo 1"}` take the first, and an export
writes both destinations under the one name. A name nobody can resolve is not a
name.
A name already in the document is a usage error now, naming it and where to
look: "a text anchor named `Photo 1` is already in this document (anchor.list
names them)". A name that is empty or only spaces is refused too — a
cross-reference built from `<txtAnchrName />` would show nothing.
Case still distinguishes names, as it does elsewhere in the document, and
nothing about existing anchors changes: the check is on creation.
The test covers the same name in the same story and in another one, the empty
names, that only one anchor exists afterwards, and that another name and a
name differing in case are both fine. Without the check the duplicate is
created again.
Not from a ticket: found probing `anchor.create`.
Found probing `layout.pages.insert` on a two-page document:
{"count": 0} → one page inserted (0 clamped up to 1)
{"count": -1} → one page (`as_u64` fails, the default applies)
{"count": 1.5} → one page (the same way)
{"count": 10000} → 9999 pages, quietly capped
A count that is given is a number of pages, so each of those is a usage error
now: "count 0: 1 to 9999 pages". A count that is absent still means one page.
`after` is left alone. My first attempt also refused a position past the last
page, and `insert_after_u64_max_appends_at_the_end` caught it: appending is
what an out-of-range `after` is meant to do, and the test now says so in the
same place.
The test covers the five counts, that none of them inserted anything, that a
position past the end still appends, and that `{"count": 2, "after": 0}` still
inserts two pages after the first. Without the check, `{"count": 0}` inserts a
page again.
Not from a ticket: I found it probing the command's parameters.
The same selection written two ways gave different files:
file.exportPdf {"pages": "2,2"} → 1 page
file.exportPdf {"pages": [2, 2]} → 2 pages, the second a copy
file.exportPdf {"pages": [2,2,2]} → 3 pages
`parse_page_range` keeps the first of a repeat (`"2-2,2"` is one page too); the
array form appended whatever it was given. A list says which pages to export,
not how many times, so the array now drops a repeat as the string does. Both
`file.exportPdf` and `file.exportPptx` read the same parameter.
The test pins the agreement between the two forms over nine lists, in PDF and
PPTX, and the two facts I measured while checking it: the pages come out in the
document's order whichever order the list names them in, and `[2, 2]` is now
byte for byte the file `"2"` gives. A page that does not exist is still
refused, and an empty list still exports nothing.
Not from a ticket: I found it probing the page-range parameter.
A cell's text rotation (Cell Options › Text › Rotation) was already read from
IDML and written back, but only 0 is composed, so an imported form's narrow
header columns of vertical labels came in horizontal and wrapped mid-word
("Eingela / ngt am") — and nothing anywhere said why. The reporter's file has
30 such cells.
Composing rotated cell text is a larger change than this. What this does is
stop it being silent, and let the value be set from the API:
- the IDML import reports it once per file: "2 table cell(s) have rotated text
(Cell Options › Text › Rotation): the angle is kept and written back, but the
text is composed horizontally";
- `table.setCell` takes `rotation` (0, 90, 180 or 270, any other angle refused,
negatives and multiples normalised), which nothing could set before;
- `table.get` reports each cell's rotation, so a round trip can be checked.
Tests: the angle through an export and back with the warning counted once for
the file (and no warning when no cell is rotated), and the command's accepted
and refused angles. Each fails without its own part.
Found while testing crates/engine/src/cmd/object.rs:2519
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Brandon Thomas