* Formats: port the craft-libs cutover follow-ups from #2734 to main (craft-libs#778)
Brings #2734's changes (made against the release branch) to main:
IDML CDATA on write and table span parsing, DOCX plain-text accepted
changes, XLSX invalid number cells shown as written, shared CMYK JPEG
polarity, exact PPTX opacity, PDF CMYK fallback wording, and the
craft-libs re-pins. Imported DOCX built-in headings keep main's and
v0.21.0's spelling ("Heading 1"), with a test. craft-pdf, craft-prepress,
craft-cms, the krilla/hayro patches and craft-xml-dtd move to the #1322
squash merge on craft-libs main (a5ac3aa0).
Co-authored-by: Dylan M. Taylor <dylan@dylanmtaylor.com>
* Point the PDF dependency check at the craft-libs main pin
Co-authored-by: Dylan M. Taylor <dylan@dylanmtaylor.com>
* Formats: address review on the cutover follow-ups
- PDF: keep main's CMYK fallback warning. The shared decoder reads planar
CMYK TIFFs and makes associated alpha straight, so neither falls back;
a test now checks both keep their inks.
- DOCX plain-text import uses craft-wordmodel's Paragraph::final_text,
which already drops tracked deletions and expands inline objects.
- IDML: a test pins numeric table spans (fractions round down, spans past
the table or past 32 bits reach its edge, anything under 1 or not a
number spans one cell).
- Cargo.lock: craft-rtf takes craft-codepage at bfea7d98 after the merge
with main.
* PDF test: expect the CMYK fallback warning's main wording
---------
Co-authored-by: Dylan M. Taylor <dylan@dylanmtaylor.com>
* cli: report an empty benchmark layout instead of panicking
Return a clear error before benchmark work starts when the document contains no spreads.
* test: report a benchmark error for documents with no spreads
Create a structurally valid round-tripped document with no spreads and require bench to exit 1 with the clear error.
* cli: report failures after keep-going scripts
Keep executing requested steps while retaining the first failure in Report, so failed scripts return nonzero before save/export.
* docs: describe first script failure in Report
Describe the first failing step without assuming all callers stop immediately.
* test: cover keep-going failure reporting and output guards
Exercise failures before and after a successful step, preserve the first failure, and require both output paths to remain absent.
Reject numeric prerelease identifiers with leading zeros and core components larger than u64 before replacing the workspace version. Extend the existing validator and replacement tests.
Option, with or without Shift, on a letter, digit, symbol or Space types a
character on macOS, so the window takes the key before the menu bar sees
it. The bar carried such a user shortcut as its key equivalent and the
window's dispatcher left the key to the bar, so the shortcut ran nowhere.
These shortcuts now stay in the app, like modifier-less ones.
When a command leaves a live caption behind its source, the pass after
the command rewrites the caption's text but kept the document's
revision. The Pages panel thumbnails, the separations preview, the
preflight worker and the Links panel key their caches by (uid,
revision), so they kept drawing the old caption. The pass now bumps the
revision when it changes the document, as Smart Text Reflow and
auto-size already do.
* Stroke: the text Weight field keeps its value between 0 and 800 pt
With text selected, the Stroke panel's Weight field sent the typed value
to type.char unchanged, so 2000 or -5 became the text's stroke weight.
The object Weight field keeps its value between 0 and 1000 pt. The text
field now keeps its value between 0 and 800 pt, the range of the
character style's Weight field.
* Properties: the text stroke Weight field keeps 0 to 800 pt
The Properties panel's text stroke Weight field capped typed values at
100 pt while the Stroke panel's text Weight field and the character
style's Weight field allow up to 800 pt. It now uses the same 0 to 800 pt
range.
* Updates: take craft-update from craft-libs main at 6e03f65bf
craft-update moves from storytold/craft-libs at 1b34eba41 to 6e03f65bf. Cargo.lock follows.
* Updates: take craft-update from craft-libs main at e98299546
craft-libs e98299546 includes #923: the check of whether the copy can install itself now runs on the update-check thread instead of in Updater::tick on the UI thread.
worker::submit answers Conflict when this session's write is cancelled
and another session's write waits to follow it, so a later edit cannot
replace that waiting write. No test covered that answer. The new test
fails when the check is removed.
* Tables: table.setCell records no undo step when the cells already have the values
When every value table.setCell sets already matches the target cells, the
command leaves the document as it is: no undo step and no new revision.
Setting a cell's text to the text it already holds keeps its story, and a
stroke the chosen edges already show as their highest-priority local
stroke is not renumbered. The edited table replaces the old one directly,
so a change copies the table once, as before.
* Tables: table.setCell keeps a cell's character formats when the text is the same
A table.setCell call whose text matched a cell with more than one character format replaced the text anyway: every character took the first character's format and the call recorded an undo step. The cell's story is now left as it is when its text already matches.
* Tables: table.setCell applies a stroke when a neighbor draws the shared edge
When both copies of a shared edge had the table's highest priority and the neighboring cell won the tie, a stroke edit whose result matched the selected cell's own copy was skipped, and the edge kept the neighbor's stroke. The edit is now skipped only when the selected cell's stroke is the one drawn.
* Tables: table.setCell checks the whole drawn side before skipping a stroke
A repeated stroke edit was applied again, with a new undo step, after any other edge in the table had been edited, because the check required the edge to hold the table's highest priority. The check now compares the cell's stroke with every part of the side as drawn, so an edit elsewhere does not count, and a merged cell's side still takes the edit when a neighbor draws a different stroke along part of it.
File › Place now sizes a raster by the resolution stored in the file,
and one point per pixel only when the file stores none. edit.pasteImage
uses the same size, so its params text and comment say so, and a test
pastes a 144 ppi PNG and checks that it comes in at half its pixel size.
* Captions: live captions follow a gesture when it commits
A preview skips the live caption refresh, and committing the gesture ran
none, so a caption kept its old text until the next command. Committing
an interaction now brings live captions up to date in the same undo
step, auto-sizes the caption frames it changed, and gives the document
a new revision.
* Captions: a live caption inside a group follows its source
The refresh after a command and after a committed gesture looked for live
captions only at the top of each spread, so a document whose only live
caption was grouped never refreshed it. Both checks now look inside groups,
as the refresh itself does.
* Captions: keep the caret inside a live caption that refreshes
When a live caption's story is replaced with its source's text, the text
selection kept its old offsets. Typing into a live caption, or a source
change that shortens one, left the caret past the end of the story, and
the IME text context reported nothing. The selection is now clamped to
the refreshed story, after a command and after a committed gesture.
Lines already ignore a baseline grid whose steps can't be counted in an
f64 across the column (a tiny increment, or a first grid line very far
off). Drop caps still sized themselves to that grid's line pitch, and
vertical justification still moved the lines by its steps, so the drop
cap missed line N's baseline and Bottom stopped short of the inset.
Both now use the same countable-grid check on the text area.
Hyperlinks and buttons accepted any scheme and passed it to the PDF exporter, so an imported file could emit file:, javascript: or UNC links that a recipient's viewer may follow. Add a shared http/https/mailto check at both command entry points and use the same check at export.
* MCP: cap one request line at 4 MiB
The stdio server read newline-delimited requests with read_until into a
growing Vec, so a client that never sends a newline could grow it without
bound. One line now stops at 4 MiB, answers with an invalid-request error and
drops the rest of the line, then keeps serving the next request.
* mcp: only discard the remainder of a line that was actually cut
* mcp: serialize the too-long reply as JSON bytes
* mcp: format the too-long-line reply
* mcp: cover the line-cap discard and EOF paths
* PDF: allow only web and mail link schemes
Hyperlink and button URLs were written into PDF link annotations whatever
their scheme, so a document could carry javascript:, data:, file: or a custom
scheme into a reader. Only http, https, mailto and ftp targets are written
now; other strings are left out of the PDF.
* pdf: move the link-scheme test below the items it tests
When an image in view lacked a level for one of its drawings, warm left
it to decode and counted its decode estimate, but also kept its cached
levels in the bytes of other images, so a decode that fit could be left
to the render. The levels the estimate covers (the levels drawn for a
full decode, the levels from the one read from the blocks to the one
drawn for a reduced JPEG) now count only with the image; other cached
levels stay with the other images. A JPEG without a reduced estimate
(CMYK) is counted at its full decode, and so are its levels. The
estimates are worked out only for the images that still need decoding,
after the check that two or more do.
The one-at-a-time test checks the peak against the two values the
thread timing allows: one large reservation, or one large and the small
image's.
A 16-bit CMYKA TIFF counts its five 16-bit samples beside the 8-bit CMYKA copy and the inks and alpha split from it, 20 bytes a pixel. The tiff encoder has no 16-bit CMYKA type, so the test defines one.
File > Save, Save As, Save a Copy and the saves before close and quit,
asked for while the open-time link or migration checks of the document
run, wait for them and then write, so the first save of a document saved
before linked storage already writes links. The UI defers the save and
re-issues it once links.checking() turns false; the status line counts
the files left ("Checking N linked images…"), Esc saves at once.
Recovery snapshots and the control channel's file.save don't wait.
links.progress reports how many files are left.
- AssetData::load_for_output uses linked bytes already in memory only while
the file still holds them (Loader::holds: fingerprint, else hash, else the
kept bytes); a changed file is exported from disk with the warning.
- PDF export keeps output bytes per asset only for linked assets, and tells a
placed PDF by its bytes (a proxy by the asset's type).
- links.list reports a link outside the automation read roots as missing
while the open-time check still runs.