* docx: ignore w:lnNumType without w:countBy (#6209)
* DOCX: record the w:lnNumType rule in the parity docs (#6209)
A row for `w:lnNumType` (§17.6.8), its status line and a revision entry in
`docs/file-format-parity.md`, the progress-log bullet in `ROADMAP.md`, a comment in
the reader saying why a missing `w:countBy` leaves numbering off, and a save check in
the zero-interval test.
* View: editable Markdown source view for .md files (#5826)
Co-authored-by: oldgranola <marken.john206@gmail.com>
* Address review feedback on Markdown source view
- Fix view.markdownSourceText purity and input validation
- Single-pass block line mapping in markdown parser
- Reuse parsed documents and block line ranges on apply and leave
- Bound memory usage with MAX_SOURCE_BYTES and clamp caret byte offsets
- Isolate temp directories in tests and test caret validation
- Preserve dirty state while markdown text edits are pending
- Flush pending markdown text edits in swap_doc
* Address CodeRabbit review feedback on Markdown source view
- Exempt view.markdownSourceText in command refusal
- Preserve original source line mapping when filtering link definitions
- Map between egui char indices and byte-offset caret in history and undo
* Remap block line numbers in a single linear pass
* Refuse over-limit Markdown Source edits to prevent document truncation
* Keep a Markdown file's own text when it is opened from handed-over bytes
`file.open` kept a Markdown file's exact text only for path-backed opens, so a
platform that hands the bytes over (data) lost front matter, comments and the
author's line endings on the first save, and the rewritten flag could not say
so. The kept text now comes from the same bytes `read_document` decodes, still
bounded by MAX_SOURCE_BYTES through FileText::from_bytes; the path read stays
for opens without data.
---------
Co-authored-by: oldgranola <marken.john206@gmail.com>
Co-authored-by: CommandCodeBot <noreply@commandcode.ai>
Giving the dialog up cancels the request. If the portal then fails with an
error other than its cancellation code, the worker opened the rfd dialog even
though nothing was waiting for the answer, leaving an unwanted Save dialog on
screen. The fallback now checks this request's cancellation token first.
system_clipboard::fill recomputed privateClipboard whenever the host could report ownership, overwriting the value the Paste Special dialog captured when it opened. Only compute it when the caller supplied none, so the dialog's format list and the paste agree while ribbon and agent calls keep reading ownership from the clipboard.
An abandoned portal request finishing late cleared the token of a newer
request in the shared slot, which left the newer Save dialog uncancellable.
Each arming now carries an identity and only its owner clears the slot; a
regression test arms two requests and disarms the older one.
Paste selected WordCraft's own fragment whenever the system clipboard text matched, even when the host reported that another program owned the clipboard, so a plain-text copy made elsewhere could paste the earlier bold fragment back. Read privateClipboard in paste_content and in the Formatted Text source, using the session fragment only when ownership is true or the caller omits the flag (agents and the CLI keep working). Carry the flag through the Paste Special request and dialog so its format list matches.
- Recognize the XDG document portal mount inside a Flatpak sandbox
(`/run/flatpak/doc/`), so a changed extension reopens the dialog for the
overwrite confirmation like the host mount does.
- Give the portal SaveFile request a real cancellation path: the CancelToken is
armed before the waiting thread starts and cancelled when the system dialog is
given up (File > Open > Browse in WordCraft...). The request closes and the
answer reads as a cancel, so a portal that stops answering cannot keep Save As
stuck; a late cancel after the answer is ignored.
- Map the cancellation error to a cancelled dialog instead of the rfd fallback.
- Regression tests for the Flatpak mount, the CancelHandle arm/cancel/disarm,
and the UI calling the host's cancel hook when the dialog is given up.
Co-authored-by: CommandCodeBot <noreply@commandcode.ai>
release.yml also runs nightly on main with craft-common-actions' dry-run: every release build
except macOS, with nothing signed, pushed or drafted. Breakage that only the release build hits
(riscv64, Flatpak, FreeBSD, packaging) then shows up the next morning instead of on release
day, as it did on 11 Nov 2026.
Each third-party action that StepSecurity publishes a secure drop-in
replacement for now uses that replacement, pinned to a commit with its
version as a comment. Versions follow what we used: the newest
StepSecurity release in the same major, or the closest major they
publish. dtolnay/rust-toolchain refs that named a toolchain (stable,
nightly, 1.76) become the replacement's `toolchain` input.
Adds step-security/harden-runner (audit, pinned to v2.22.4 like
craft-common-actions) as the first step of each job that did not have it,
so every workflow in the org's public repos reports its egress to
StepSecurity. Audit, not block: block-mode allow lists come from each
job's observed traffic.
The deb and rpm only recommend the X11 libraries, and the release tarball
doesn't install them. When the display-server choice (#2184) or
WORDCRAFT_WAYLAND=0 asked for XWayland without them, the app failed to
start: winit panicked in xkbcommon-dl without libxkbcommon-x11 (seen on
Fedora 44 and Ubuntu 26.04).
uses_x11 now also takes a check of the libraries winit's X11 backend opens
(libX11, libXcursor, libX11-xcb, libXi, libxkbcommon-x11), asked only when
X11 would be forced. If any is missing, X11 isn't forced and the log line
names them. The packages now also recommend libX11-xcb and
libxkbcommon-x11, and the README lists the libraries XWayland needs.
Tests: missing_x11_libraries_leave_the_app_on_wayland (every way of asking
for X11, and that asking for Wayland doesn't run the check).
Co-authored-by: bmh129 <11478682+bmh129@users.noreply.github.com>
The Print job runs its own apt-get, so it failed while ubuntu.com did not resolve on the runners
(11 Nov 2026). configure-apt-mirrors (craft-common-actions) points apt at mirrors outside
ubuntu.com when those don't answer, as the shared jobs already do.
* feat(engine): format Brazilian date fields and insert localized dates
* feat(ui): persist preferred date and time format
* feat(ui): implement Word-style Insert > Date & Time modal
* feat(ribbon): open Date & Time dialog from Insert tab
* test: Brazilian Word TIME fields respect document defaults after DOCX round-trip
* fix(docx): refresh Word date/time fields using their run language
* ci: add temporary date/time regression validation workflow
* style: apply rustfmt output to insert.rs
* style: apply rustfmt output to date_fields.rs
* style: apply rustfmt output to dialogs_insert.rs
* i18n: reuse translated date-format and language labels
* i18n(pt-BR): translate Update automatically in date/time dialog
* fix(i18n): translate date dialog options and audit generated date previews
* chore: remove temporary date/time regression workflow after focused validation
* fix(date): resolve inherited locale and preserve it in fixed text
* fix(ui): display date insertion errors without closing the dialog and document state
* test: cover inherited language, static date text, and escaped date pictures
* test(ui): exercise date language, picture, checkbox, OK, and invalid-value feedback via egui
* style: apply rustfmt to date insertion and regression tests
* fix(engine): borrow caret position while resolving inherited date language
* fix(date): honor selected locale when the date picture is omitted
* test(ui): target date dialog language combo instead of ribbon menu
* test(date): use raw field instruction literal in locale regression
* test(date): match actual single-backslash DATE field instruction
* test(ui): select actual date dialog language option by accessible label
* fix(i18n): add missing automatic date update translation (Romanian)
* fix(i18n): add missing automatic date update translation (Slovenian)
* fix(ui): open date_time dialog from keytip, normalize pt tags, and document items
* fix(ui): default date_time_update preference to false
---------
Co-authored-by: felipegavilanes-source <felipegavilanes@gmail.com>