Commit Graph
145 Commits
Author SHA1 Message Date
FredApps ae257ba784 Merge upstream font tests and window scrolling 2026-11-08 17:26:55 +01:00
FredApps 0495fbb789 Merge pull request #2783 from vangel76/test/fixed-layout-fonts-2729
#2729: Layout tests use pinned bundled fonts
2026-11-08 17:24:43 +01:00
batistalucasv c4d5f70c1d DOCX: resolve LibreOffice font fallback lists to first installed family (#5843) (#5851)
* #5843: resolve LibreOffice semicolon-separated font fallback lists to first installed family

* #5843: address review comments on font fallback list resolution

* #5843: verify fallback list text on cache hit and accept Times New Roman in tests

* #5843: bound the fallback lists read from a DOCX and reuse their classification

A document could name thousands of fallback families in `w:ascii`/`w:hAnsi`.
Nothing bounded that value, and the name-based rules (roman classification,
Thai family, named substitutes) then scanned every candidate for each shaped
piece, so opening the document took seconds.

`w:ascii` and `w:hAnsi` now keep the leading families within MAX_FONT_NAME
bytes, cut at a `;` so no name is split, as `w:cs`/`w:eastAsia` already bound
their names. The family a real list would pick is kept: a LibreOffice list
names a few families, well under the bound.

The Thai-family and named-substitute answers are worked out once per list and
kept in the fallback-list cache, keyed by the list text, so shaping no longer
rescans candidates for every piece. Resolution still reads every candidate.

A 500-paragraph document whose font names 20,000 candidates laid out in 10.8 s;
bounded on read it takes 17 ms.

Tests:
- a hostile fallback list keeps its leading families, cut on a `;`
- a real LibreOffice list is read whole
- a later Thai or named fallback still classifies, and lists are cached
2026-11-08 05:55:24 +01:00
vangel76andClaude Opus 5.5 9655b2da50 Merge upstream/main into test/fixed-layout-fonts-2729
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-08 04:21:38 +01:00
vangel76andClaude Opus 5.5 66e169c343 Merge upstream/main into ui-text-hinting
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-08 04:21:35 +01:00
050b6b0da3 Arabic letters join across a change of formatting (#4093) (#5691)
* Arabic letters join across a change of formatting (#4093)

Each run of a paragraph was shaped on its own, so a letter at a run's edge
took the form of a word's start or end. A coloured letter stood apart from
its word, and a word edited with Track Changes on read with a space in it
(ال ذكاء).

Runs are now shaped with the text shown on either side of them, through the
shaper's pre- and post-context (fonts::ShapeContext). Runs left out of the
layout (hidden text, deletions in a final view) are skipped, so letters join
what is on the page. Only paragraphs with a joining script do the work.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2QwHfTULQtWYgVEXk

* #4093 review: no letter joins a drop cap formatted on its own

A drop cap letter in a run of its own (coloured, say) was still passed to
the next run as the text before it, so that run's first letter joined the
drop cap, which is drawn apart from the word. The shown text now has a floor
past the drop cap, and no run takes context from before it. The drop cap
test also covers the letter in a run of its own.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LvG6E2QwHfTULQtWYgVEXk

---------

Co-authored-by: Filip Strajnar <filip.strajnar@proton.me>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Martin Ball <martin.ball@ucscaregroup.co.uk>
2026-11-08 03:44:01 +01:00
vangel76andClaude Sonnet 5.5 ddb1d37de3 Merge upstream/main into test/fixed-layout-fonts-2729
Keep upstream's split of layout_para and install the pinned fonts at its
start. Refit the main tests added since (#4250 widow and footnote length,
#4720 chart fallback font) to the pinned test fonts. Merge the two
wordcraft-layout dev-dependency tables.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-07 18:22:44 +01:00
vangel76andClaude Sonnet 5.5 2d6da7a326 Merge upstream/main into ui-text-hinting
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-07 11:38:09 +01:00
Martin Ballandpick2510 9a9246915c Layout: Times and Courier take Times New Roman's and Courier New's line spacing (#3241)
Lands #3241.

Co-Authored-By: pick2510 <4068086+pick2510@users.noreply.github.com>
2026-11-06 02:03:34 +01:00
vangel76andClaude Sonnet 5.5 67c9b8ac31 Merge upstream/main into ui-text-hinting
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-06 01:06:04 +01:00
db8d944c5c Fonts: reuse shared MATH construction records (#4660)
Co-Authored-By: OPUZZLES <109076338+OPUZZLES@users.noreply.github.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-05 18:50:25 +01:00
vangel76andClaude Sonnet 5.5 75a905802c Merge upstream/main into ui-text-hinting
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-05 03:36:06 +01:00
Aleksander HelgakerandClaude Opus 5.5 ba58e70cc8 Flag document fonts that aren't installed (#4770)
* Gap 4: flag document fonts that aren't installed

The font box (ribbon, mini toolbar, both Font dialog boxes) shows a
warning mark when the current font isn't installed; its hover note
names the font drawn instead. The status bar says "N fonts not
installed" and lists each with the faces drawn instead, plus Add Font….
Agents get `font.missing` and `format.state.fontShownAs`.

Layout records, where it picks a face per text segment, the fonts the
shown text asks for that aren't installed and the faces drawn instead
(ParaLayout::substituted, DocLayout::substituted_fonts), so hidden
text, unused styles and orphaned stories never count; list labels,
field results and East Asian and complex-script text do, and so does a
missing roman font drawn in Cambria. Metric-compatible substitutes and
symbol fonts drawn as Unicode aren't flagged. Bounded for hostile
files: 64 pairs a paragraph, 1024 a document, names kept to 256
characters, 20 rows in the popup.

Ported from alehel/wordcraft#1 onto current main; English strings only
(translations belong in storytold/craft-i18n).

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

* Gap 4: missing-font review fixes

- Judge a padded font name by the face layout drew for it, and skip blank names.
- Name the face CJK text is drawn in for a missing East Asian font, not Cambria.
- Keep the missing-fonts list per window, so several windows don't rebuild it every frame.
- Expose the font box's warning to screen readers as its description.
- Count missing fonts with craft-i18n plurals (`i18n::tn`).
- One table of metric twins (`fonts::word::METRIC_TWINS`): a missing font is drawn in an
  installed twin first, and the warning reads the same table.
- Find the status-bar item by part of its label in the click test.

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

* Gap 4: warn for a font picked at the caret; drop fonts of hidden list items

- At a collapsed caret the font box warns for a missing font typing will use, also at the end
  of text in another font (it waited for the first letter).
- A list item that shows nothing (#3982) reports no missing font: its number was shaped, and
  its font recorded, before layout dropped it.

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

* Gap 4: say where Add Font… is, and that it needs the font file

The font box's note said "Add Font… installs it": Add Font… opens a file picker at the end of
the list, so it only helps with a font file at hand.

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

* Gap 4: docs say the popup lists 20, and when the font box warns ahead

CodeRabbit review: the status-bar popup lists the first 20 missing fonts, then a count; the font
box also warns at a collapsed caret for a font no shown text uses yet; `fontShownAs` is the
font's substitute, fallback faces aside.

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

* Gap 4: the Font dialog warns for a missing font over a selection

CodeRabbit review: the dialog's boxes show the fonts OK applies, which no shown text may use
yet (an Asian text font over Latin text, a font typed into the dialog), so they name the
substitute whatever is selected. The ribbon's box over a selection still reads the list.

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

* Gap 4: the Asian font box names the face CJK text gets

CodeRabbit review: the status bar's list is by font name, so for a font Latin text also uses
it named the Latin text's substitute. The Asian text box now always looks up the face layout
draws Chinese, Japanese and Korean text in.

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

* Gap 4: gaps.md status on the current line, missing-font scope and twin wording

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

* Gap 4: gaps.md says which missing fonts are flagged and what font.missing lists

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

* Gap 4: font.missing lists what the current view lays out

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

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-04 11:48:48 +01:00
1ff68d64e3 i18n: Arabic interface with mirrored RTL shell (+ parity docs) (#5370)
* Arabic interface: mirrored right-to-left window and Arabic UI fonts

Adds the Arabic (`ar`) interface language's code: its catalogs are craft-i18n's
`locales/ar`, so this changes no catalog.

- `ar` is registered in `i18n::LANGUAGES` and `Lang::rtl()` says it is written
  right to left; its translations are used as written (egui's text layout
  orders each line's bidi text itself). A ribbon location reads from the right
  with a leftwards mark.
- `rtl.rs`: the window is mirrored. Containers open in `rtl::root`; columns,
  grids, checkboxes, radio buttons, menus, rows, panels, the dialog tab strip
  and the ribbon's scrolling row follow the language's direction. Title bar,
  status bar, ribbon tabs and groups, File pages (including Print, with the
  settings on the right and a solid scrollbar gutter), dialogs, menus and the
  task panes (their docking edge swaps) are laid out from the right. egui's
  `Grid` panics in a right-aligned layout, so grids go through `rtl::grid`.
- Arabic UI font fallback: the craft-fonts Arabic face, else an installed
  Arabic font, loaded when the interface is Arabic.
- The "Auto (200%)" size label keeps its percent sign inside the isolated
  number.
- Tests: Arabic locale, saved preference and lookup order on fixture catalogs;
  an interface audit (skipped without the craft-i18n Arabic catalog) that lays
  the app out right to left at several widths, checks tab, group, pane, Print
  and Go To positions and that nothing panics.

The Arabic work builds on Abdelhaq El Amraoui's mirrored-interface PR (#737).

Co-Authored-By: Abdelhaq El Amraoui <abdelhaqelamraoui@gmail.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Arabic interface: pin craft-i18n at a commit with the Arabic catalog; review fixes

- craft-i18n.rev and the Flatpak manifest pin 19358491 (craft-i18n main), the
  first commit line that has locales/ar/wordcraft.tsv. The old pin had no
  Arabic catalog, so `ar` was registered but not embedded and two i18n tests
  (the_language_setting_is_a_ui_command_and_persists,
  the_languages_offered_are_the_embedded_ones_of_the_list) failed in CI.
- The Arabic font fallback is requested for any right-to-left interface
  (`Lang::rtl()`), not only after the language picker sets the flag, so an
  automatic or saved Arabic gets the installed Arabic face.
- The file browser, the update window and Reading Mode's toolbar run their
  bodies in `rtl::root`, so their far-side buttons and rows are mirrored.
- A group in a right-to-left ribbon row is measured from the right edge: the
  cursor's left edge is negative infinity there, so its width was NaN and the
  ribbon planner dropped every group, never compacting a narrow Arabic window.
  A narrow window now folds groups into drop-downs.
- The mirrored checkbox and radio marks are labelled by their text, so screen
  readers name them.
- Tests: group width in a right-to-left row; the Arabic interface audit checks
  Reading Mode's toolbar.

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

* Arabic interface: ask for the installed Arabic font separately from the CJK one

A previous commit made every right-to-left interface request the installed CJK
fallback font (`system_cjk`) so that the installed Arabic font would load. That
also made every Arabic launch read a large system CJK font file when no
craft-fonts input is embedded (the Flatpak build has none).

`install_ui_fonts` now passes a separate `system_arabic` flag (`Lang::rtl()`):
the installed Arabic font is read for it, and `system_cjk` is again set only
by a CJK interface or by CJK text on screen. `WordApp::fonts_system_arabic`
remembers what was installed, so a language change reinstalls the fonts.

Tests: the_installed_arabic_font_is_asked_for_by_arabic_alone (the Arabic flag
adds the installed font only when none is embedded, and the CJK request alone
adds no Arabic font). Checked a headless screenshot of the Arabic interface
without craft-fonts: the installed Arabic font draws the labels.

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

* Fix the AutoSave switch knob: on is at the right, mirrored in Arabic

The title bar's AutoSave knob was placed with `lead_x(rtl == on, ..)`, which
reverses main's mapping in both directions: left to right it drew the knob at
the left when the switch was on, and right to left at the right. main draws it
at the right when on and at the left when off, so the mirrored interface must
draw it at the left when on: `rtl != on`.

The position is now `autosave_knob_x(rtl, on, track)`, pinned by a test for all
four cases. A headless screenshot of the off switch shows the knob at the left
in English and at the right in Arabic.

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

* Arabic interface: file names and paths in right-to-left lines; kerning row order

Tester feedback on the Arabic interface:

- File > Open's Recent list, the built-in file browser and the title bar drew
  file names and paths raw, from the left edge. In an Arabic line the marks of
  a name ("(نهائي) 2.docx") and the folders of a path with Arabic parts were
  scrambled. Rows now start at the side reading starts from (`rtl::lead_x`),
  names go through `i18n::isolate` (a first-strong isolate, with a left-to-right
  mark before the extension so `2.docx` stays `2.docx`) and paths through
  `i18n::isolate_path` (a left-to-right isolate with a mark after every
  separator, so neighbouring Arabic folders do not swap). Other languages get
  the text untouched; saved and spoken text never carries the marks.
- Font dialog, Kerning row: the dialog's grid is drawn left to right inside a
  right-to-left window, so "Points and above" came after its number; in an
  Arabic interface it now comes first, so the line reads "14 نقطة فأكثر" from the
  right. (The grid's columns themselves still keep their left-to-right order.)

Tests: names_and_paths_are_isolated_in_a_right_to_left_interface.

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

* Arabic interface: mirror through craft_ui::direction and Style::right_to_left

The suite's craft-ui now has the direction helpers (`craft_ui::direction`) and
its egui fork a `Style::right_to_left` that mirrors egui's own chrome. Use them
instead of our parallel wrappers:

- `rtl::sync(ctx)` applies the interface direction every frame the language's
  changes (`direction::apply`): it sets the shared direction state and
  `Style::right_to_left` on every theme, which draws check boxes and radio
  buttons with the mark at the reading start, mirrors submenus, scroll bars,
  title bars and the combo boxes' arrows (the Print page's choice lists, the
  font list) and opens combo menus from the right.
- `rtl::checkbox`, `radio_value`, `radio` and `indeterminate_checkbox` are
  gone: about 200 call sites use plain `ui.checkbox`, `ui.radio_value`, ... and
  egui does the mirroring (and names the control by its text, as a plain widget).
- `lead_x`, `lead_center`, `lead_top`, `start_panel` and `end_panel` delegate to
  `craft_ui::direction`. `root`, `flows`, `island`, `grid`, `menu`, `vertical`,
  `far_side` and the other layout helpers stay: the shared crate has no
  equivalent that fits them.

Visual changes: the combo boxes' drop-down arrows and lists are mirrored in
Arabic; nothing changes in left-to-right languages (checked headless
screenshots of the English main window, Font dialog and Print page).

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

* Recent files and the file browser: full path on hover, long names keep their start

Review feedback:

- Each row of File > Open's Recent list shows the document's full path as its
  tooltip (the path line under the name can be longer than the row).
- A long name in the built-in file browser is laid out in one row, with an
  ellipsis at its end, from the side reading starts from, instead of being
  clipped at the row's edge: in a right-to-left interface the old text anchored
  at the right edge cut the end of an Arabic name and the start of a Latin one.
  The whole name is the row's tooltip when it was shortened.

Checked headless screenshots of the file browser with a long English and a long
Arabic name.

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

* Font combos in a right-to-left row: the arrow joins its field on the left

The ribbon's font name and size boxes are drawn from a field and a separate
arrow box that share one outline: the field is square and the arrow's box is
rounded on the side where the arrow goes, painted a border width under the
field. That assumed the arrow is right of the field. In a right-to-left row
(the Arabic ribbon) the arrow comes after the field on its left, so the field
was square on the wrong side and the arrow box rounded on the side facing the
field: two boxes with a gap-like seam, the arrow looking outside the frame.

`arrow_box_shape(rtl, ..)` mirrors the radii and the underlap, and the field's
square corners are chosen by the row's direction (the nw/sw pair in right to
left). Left-to-right rows are drawn exactly as before (English before/after
crops are identical).

Test: the_arrow_box_is_rounded_away_from_its_field.

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

* Pin craft-i18n 4612850 (merged Arabic WordCraft rows, craft-i18n#76)

The pin and the Flatpak manifest commit move from 1935849 to 4612850, which is on
craft-i18n main and adds the 70 Arabic rows this branch relies on.

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

* Pin craft-i18n cac2cf0 (Arabic number agreement and status-bar rows)

Moves the pin and the Flatpak manifest commit from 4612850 to the current craft-i18n
main, which includes craft-i18n#95 (48bed03) and #81 (ef58ff3).

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

* Arabic: isolate paths shown beside labels; mirror the Print page zoom

File > Info shows Location as its own left-to-right path label (cut short with the full path on hover, not wrapped), and the open-failure, Exported and dictionary-file messages isolate their paths. The Print preview zoom puts minus at the slider's left end and plus at its right in Arabic, as the status bar does.

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

* Arabic: mirror the label/field rows of the Font, Info, Paragraph, Page Setup, Insert Table and Options forms

rtl::Table lays a form out as an egui Grid in a left-to-right interface (same id, columns and spacing, so the output is unchanged) and, where the interface is mirrored, with each row's cells running from the right, every column as wide as its widest cell. File > Info puts Properties at the right and Statistics after it, and the Font dialog's kerning row reads 14 points and above.

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

* docs: record Arabic mirrored interface landing (#1195)

Update the language counts, Arabic parity row, gap #11, and UI/ROADMAP
status so they match the craft-i18n catalog and mirrored shell already
on this branch.

* Fix Organizer Copy arrow in RTL and refresh roadmap status.

Point the Copy glyph at the Document column when the shell is mirrored,
and bump docs/roadmap.md status/revision history for Arabic UI (#1195).

* fix(ui): keep pane dock left/right physical in RTL

ui.paneDock and view.stylesPaneDock now store and report physical window sides; only unset defaults still open on the reading-start edge. Also wrap the language-packs grid for RTL, document the contract, and load a system Arabic face for Arabic names in LTR UIs.

* docs: clarify mirrored default pane edges in RTL

Start-default panes open on the physical right in Arabic; end-default
panes open on the physical left — not every never-docked pane on the right.

* refactor(ui): one PaneState for a saved pane place

The four saved-dock arms built the same PaneState; only the place and
whether it was set explicitly differ.

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

* style(ui): rustfmt the panes re-exports

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

---------

Co-authored-by: The Legend Of MTS Dev <tlo.mts.tsh@gmail.com>
Co-authored-by: Abdelhaq El Amraoui <abdelhaqelamraoui@gmail.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: email <email@email.com>
Co-authored-by: FredApps <153173854+FredApps@users.noreply.github.com>
2026-11-03 15:00:45 +01:00
Dylan M. Taylor e00ff04d4a Fonts: refuse WOFF2 collections whose rebuilt directories are huge or list a table twice (#5332)
The WOFF2 decoder rebuilds every collection font's table directory (16 bytes
per listed table) before checking its output size, while the file spends as
little as one byte per listed table. A small collection listing the same
tables over and over made it allocate hundreds of megabytes twice.

Preflight the collection directory before decoding: the rebuilt directories
must fit 16 MiB and no font may list a table twice.
2026-11-02 22:24:55 -05:00
Dylan M. Taylor 220994decf Fonts: cut names read from a font's name table to 128 characters (#5333)
A name record can hold 64K characters, and every face copies its family into
each localized name and every variable-font named instance (with their
localized names again), so a small font could claim hundreds of megabytes of
name strings before any later check.

name() now keeps at most MAX_NAME_CHARS characters, the limit localized names
already used, so every copy stays small.
2026-11-02 22:24:52 -05:00
vangel76 95669b9eb9 Merge upstream/main into ui-text-hinting 2026-10-31 00:08:15 +01:00
acccf913ab Traditional Chinese interface: draw hanzi to the Taiwan standard with an installed TC font (#3380)
* i18n: Traditional Chinese interface draws hanzi with an installed TC font

The zh-hant interface shared the Simplified Chinese font order, so its first
CJK face was the embedded Noto Sans CJK SC and traditional characters were
drawn with mainland glyph shapes (說, 資, 這, 為, 戶).

On the desktop, a Traditional Chinese interface now puts an installed
Traditional Chinese font before every CJK face: PingFang TC on macOS,
Microsoft JhengHei on Windows, Noto Sans CJK TC on Linux. Noto Sans CJK SC
stays after it as the fallback. Without an installed TC font, and on wasm,
the order is unchanged.

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

* Credits: add TobyWu666 to contributors/people.toml

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

* i18n: find Traditional Chinese fonts at the root of a font folder

Linux user font folders (~/.fonts, ~/.local/share/fonts) often hold font
files directly; the Traditional Chinese candidates only named files in
distribution subfolders, so those fonts were missed (CodeRabbit review).

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

* i18n: skip the Traditional Chinese font test without embedded catalogs

Since the catalogs come from craft-i18n (#4072), a build without
CRAFT_I18N_DIR is English only, so the test now skips like the other
catalog tests. Also rustfmt hant_font_chars.

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

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
2026-10-29 22:03:12 -04:00
vangel76andClaude Sonnet 5.5 4524bc6ba9 Merge upstream/main into ui-text-hinting
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-29 22:41:52 +01:00
Dylan M. Taylor 0436139c57 Refactor fonts for strict Clippy checks 2026-10-28 01:06:48 +00:00
vangel76andClaude Sonnet 5.5 4bfd84514c Merge upstream/main into test/fixed-layout-fonts-2729
Main split fontdb.rs, math.rs and the layout test files into modules; the pinned-font
changes are reapplied to the new locations.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-27 19:47:34 +01:00
aiongg 4b8b06fe03 Merge pull request #4602 from Silvertree2010/perf/export-speed
Speed up PDF export by about 24 %
2026-10-27 13:41:24 -04:00
NetterSilverandClaude Opus 5.5 de3bfaf049 Keep harfrust shape plans per face
Each shaping call let harfrust compile a new shape plan (the face's
lookups for the run's script, direction, language and features), which
took about as long as shaping a short run. Each thread now keeps up to 64
plans (the oldest dropped to make room), found by face and harfrust's
ShapePlanKey::matches, and passes the plan to the shaper. A kept plan is
only used for the face it was made for: the face id (from one process-wide
counter, so no two faces share it) together with the font bytes' address,
length and face index. The plan is the one the shaper would have made for
the buffer, so the glyphs are the same; tests shape with kept and fresh
plans and compare, and check the per-thread bound.

PDF export of the export_bench document (259 pages), interleaved with main
(A/A within 1 %): median 421 ms -> 319 ms (-24.2 %), measured on a loaded
machine; on a quieter run earlier 294 ms -> 224 ms (-23.8 %). PNG export of
page 1 (which lays out the whole document): about -34 %. The PDF and PNG
bytes are identical to main's.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 17:39:37 +01:00
NetterSilverandClaude Opus 5.5 ffa127c615 Skip per-character table lookups for plain text in layout and PDF export
Shortcuts that give the same answers as the lookups they skip:

- is_rtl: nothing before U+0590 is right to left, so Latin, Greek and
  Cyrillic text skips the bidi class search.
- strong_rtl and is_mark: ASCII letters are the only strong ASCII
  characters, and no ASCII character is a mark.
- cluster_starts: in ASCII text every character starts a grapheme except LF
  after CR, and no ASCII character joins a cluster.
- display lines: a glyph run's vectors and text are sized up front, and the
  symbol-font check is made once per run instead of once per cluster.
- PDF glyph runs: the baseline (the y most glyphs share) is only worked
  out for italic runs, the only ones that use it, and a run on one baseline
  returns it without counting.

Tests check each shortcut against the full lookup (every code point for the
bidi classes, every ASCII pair for the clusters).

PDF export of the export_bench document, interleaved with main: median
293 ms -> 241 ms (-18 % with the previous commit). The PDF bytes are
identical.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 13:39:09 +01:00
Martin BallandClaude Opus 5.5 04ebc4ed42 #3995: Thai one-face fallback tests skip unless a Thai font with Latin is installed
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) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-27 09:52:32 +00:00
aiongg a2a0d69193 refactor(fonts): split fontdb.rs into fontdb/ 2026-10-26 21:16:49 +00:00
aiongg d3d7152b45 refactor(fonts): move tests_sysfonts.rs to fontdb/tests/sysfonts.rs 2026-10-26 20:54:02 +00:00
aiongg c5b1efb97d refactor(fonts): move fontdb tests to fontdb/tests.rs 2026-10-26 20:54:02 +00:00
aiongg ffb21794c5 refactor(fonts): move fontdb/user tests to fontdb/user/tests.rs 2026-10-26 20:54:02 +00:00
Dylan M. Taylor d859c673c1 Sonar: reduce cognitive complexity in highlight, review and outline commands 2026-10-26 16:01:36 -04:00
vangel76andClaude Sonnet 5.5 1a950a3631 Fix FlipPen initialisers after the hinting-strike change on main
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-26 19:18:45 +01:00
vangel76 f4ff1f842d Merge upstream/main
# Conflicts:
#	ROADMAP.md
#	docs/ui-parity.md
2026-10-26 19:14:40 +01:00
vangel76 d8899ed345 Merge upstream/main 2026-10-26 19:13:12 +01:00
Martin BallandClaude Opus 5.5 328a58dd03 #3869: review — hint_strike is native-only like HINT_STRIKE (wasm build); composite and 0x7FFF contour-count tests
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HUodYU9r4mPHed4pPsjHC2
2026-10-26 02:43:10 +00:00
Martin BallandClaude Opus 5.5 ddda3c4670 #3869: review — the draw that turns hinting off is redrawn unhinted, so the outline cache never keeps a hinted glyph from a tripped breaker
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HUodYU9r4mPHed4pPsjHC2
2026-10-26 02:09:31 +00:00
Martin BallandClaude Opus 5.5 b094b29958 #3869: review — WOFF2 point pre-check reads glyf nContours as u16 like the decoder (only 0xFFFF is composite)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HUodYU9r4mPHed4pPsjHC2
2026-10-26 02:09:24 +00:00
vangel76 878e5e3457 Merge upstream/main into test/fixed-layout-fonts-2729 2026-10-26 02:19:00 +01:00
vangel76andClaude Sonnet 5.5 71aa12c076 Fix hinting tests and imports after merging upstream/main
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-26 02:16:31 +01:00
vangel76 cfa99b45ce Merge upstream/main into ui-text-hinting 2026-10-26 02:15:38 +01:00
Martin BallandClaude Opus 5.5 2d9bcb6b1b #3692: review — fonts timing test gets a margin a busy CI machine can't break
named_instances_and_collection_members_are_capped: 2 s -> 10 s. The face counts in the test
(one face for a million identical members, MAX_FILE_FACES from add_font) are what catch the old
code, which built a face per member; the time (under 0.1 s on a loaded machine, so 10 s is over
100x) only backs up "read once".

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:02:21 -04:00
Martin BallandClaude Opus 5.5 b059da2d4e #3692: review — docs: why the hinting breaker is timed, and what brotli-decompressor brings
The hinting breaker stays wall-clock: skrifa's interpreter budget is fixed,
counts instructions rather than the points each one moves, and reports no
work count, so there is nothing deterministic to bound. The comment now says
the result can depend on the machine by design, and that only faces needing
the interpreter for their contours are hinted at all. Cargo.toml notes that
brotli-decompressor is the crate and version wuff already uses: pure Rust,
forbid(unsafe_code) by default, wasm-ready, BSD-3-Clause/MIT.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:02:21 -04:00
Martin BallandClaude Opus 5.5 57e9e0f244 #3703: review — font store adds run one at a time, so racing adds never share a temporary or a file
CodeRabbit on #3869: two concurrent install_font calls could pick the same
destination and temporary, and the leftover-unlink retry could remove the
other call's open temporary. write_unique now holds a process-wide lock.
In-process only (noted in a ponytail comment).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:02:21 -04:00
Martin BallandClaude Opus 5.5 68d45318bb #3692: review — a colour glyph's clip outline is drawn within the payload left, refused before it is built whole
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:02:11 -04:00
Martin BallandClaude Opus 5.5 d69c386f8f #3692: review — a font's temporary file is created afresh, so a link planted at its name is not written through
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CX1hMKX28764Lrx5moxs1o
2026-10-25 21:02:11 -04:00
Martin BallandClaude Opus 5.5 e541a6f62f #3692: hinting turns off only when slow draws are a real share of a face's draws
Ultrareview finding 5: the slow-draw count only grew, so three draws slowed
by a busy machine at any point in a session switched hinting off until
restart, breaking hint-reliant faces (MingLiU, DFKai-SB). Hinting now stops
only when at least three draws were slow and they are at least one in 16 of
the face's hinted draws: a face slow throughout still stops at its third
glyph, and slow glyphs spread among fast ones are still bounded. Counting
interpreter work instead of wall-clock stays with #3827.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:11 -04:00
Martin BallandClaude Opus 5.5 bbe1dad1e7 #3692: hinting turns off after three slow glyphs, not one
CodeRabbit review of finding 62's fix: one hinted draw slowed by a busy
machine (scheduler pause, page fault) switched hinting off for the face for
the rest of the process, leaving hint-reliant faces with timing-dependent
missing strokes. A strike counter now drops hinting only at the third slow
glyph.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:11 -04:00
Martin BallandClaude Opus 5.5 c7148d6ba9 #3692: hint-reliant fonts with huge zones are drawn unhinted, and a slow hinted glyph turns hinting off
Finding 62 (partial): skrifa 0.47 caps a hinting run at 1M instructions
but not the points each moves (SHZ, IUP... walk a whole zone), and a font
opts into the interpreter by its family name. Faces asking for a
twilight zone over 1,024 points or a stack over 4,096 values get no
interpreter; a glyph whose hinting buffer passes 256 KiB is drawn
unhinted; and (native) one hinted glyph over 50 ms turns hinting off for
its face. Residual: a face within the limits can still spend about
1M x zone size point moves on one glyph (and its prep program) before
the strike lands; the full fix is a point-work budget in skrifa or an
isolated worker with a deadline.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:01 -04:00
Martin BallandClaude Opus 5.5 b246ea166a #3692: COLR gradient stops and clip paths count against a per-glyph budget
Finding 61: Recorder copied every gradient's stop list (and every clip
glyph's outline) per op, capped only by op count, so a compact COLR graph
reusing one long stop list could expand to billions of stops. Stops and
path elements now share MAX_PAYLOAD (65,536) per glyph, charged before
copying; past it the glyph falls back to its outline and no further clip
outlines are built. The render cache charges each glyph's real ops,
stops and path bytes (ops_bytes) instead of a flat 64 bytes per op.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:01 -04:00
Martin BallandClaude Opus 5.5 ce7e4ea28e #3716: a font file's faces are capped before they are built
Finding 59: enumerate_faces took a collection's member count straight from
its header (a TTC can list one table directory millions of times) and
font_faces read every fvar named instance (up to 65535, with a linear
style dedupe and a name-table lookup each). Every entry became a FontFace
kept for the session, and add_fonts then checked each new face against
every face in the database.

Now a file yields at most 256 faces (MAX_FILE_FACES; user fonts ask for
one more so they can still be refused as "more than 256 faces"), named
instances are read to the same cap, a face with the family and style of
an earlier one in the file is skipped (repeated collection offsets), and
add_fonts dedupes against an indexed set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:01 -04:00
Martin BallandClaude Opus 5.5 6c6326486f #3692: font names are debug-quoted in log warnings
Finding 46: install_font logged the caller-supplied file name raw when the
font couldn't be stored, and the desktop log writes line breaks as they
come, so a name could forge extra log records. The name is now logged
with {:?} (escaped and quoted), and so are the font-derived family names in
the two glyph-outline warnings, which come from a document's own fonts.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-25 21:02:01 -04:00