On a desktop with a Ryzen integrated GPU without a monitor and an NVIDIA
card driving both monitors, Automatic (power saving on Windows) picked
the integrated GPU; Windows presented every frame across adapters and
the integrated GPU's driver reset, losing the graphics device or taking
every monitor down for minutes (same bug as pdfcraft#378).
Automatic now asks the system which GPU drives which display (Windows:
EnumDisplayDevices through winsafe; Linux: /sys/class/drm) and ranks the
GPU driving a display, and among those the primary display's, ahead of
the power preference. Where the system can't say, or no adapter's PCI
ids match (OpenGL, Metal), the order is the power preference's as
before, so nothing gets worse. Power Saving and High Performance chosen
by the user, and WGPU_POWER_PREF, keep ordering by kind alone.
The log now lists the display GPUs, which adapters drive a display, and
the adapter drawn on with the reason, the first question in every
black-window report. Tests cover the device-id parser, the rankings
(desktop with all monitors on the discrete GPU, hybrid laptop, a monitor
on each GPU, no match), the sysfs reader and a live EnumDisplayDevices
call.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017ZwZjADUZP4VzvrFGpiZch
An ellipse kept pie angles in its model, but nothing drew or set them. Now
the Transform and Properties panels show Ellipse Properties with Pie Start
Angle, Pie End Angle and Invert Pie, as in Illustrator, and
object.setLiveShape takes {pieStart, pieEnd, invertPie}. The pie runs
counterclockwise on screen from the start angle (0 at 3 o'clock) to the
end angle and closes through the centre; 0 to 360 is the whole ellipse,
an end of 0 means 360, and Invert Pie swaps the angles to show the other
part. One undo step each; a bad angle is an error. The labels are
translated in every complete language. The on-canvas pie widget is still
to come.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The toolbar's All Tools drawer listed the tools' names as plain buttons.
It now shows each tool with its icon and name in a grid, under the
toolbar's group headings (Select, Shapes, Draw, Modify, Type, Navigate,
Color) and headings for the tools only the Advanced toolbar shows
(Perspective, Symbols, Graph, Artboards, Slices). The active tool is
highlighted, a tool's tooltip gives its name and shortcut, and picking one
selects it and closes the drawer. The headings use labels translated
already.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The toolbar and the dock's icon column scroll when the window is too
short for them. Their scroll bar was egui's floating one, which widens to
10 pt while scrolled or dragged and covered the group labels and the
icons' right edge. It now stays at most 4 pt wide (2 pt at rest), at the
column's edge.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A font library can hold two versions of one family and style (a new
Montserrat-Bold.ttf beside an old montserrat.ttf). The font database kept
only the first one loaded, so type set in the other couldn't be matched
when a PDF or .ai was opened: it came in as outlines. Now every version
loads (the first stays the one family and style resolve to), PDF import
tries the other versions when the glyphs differ, and type set in a version
the family and style alone wouldn't pick names it in CharStyle::font_version
(left out when None, so documents without it are unchanged). Layout,
shaping, the substituted-glyph highlight, Find Font, Package and Document
Info follow it; changing the font clears it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Japanese now translates the status-bar and error messages (`@msg`
templates) like Spanish, French, Italian, Russian and Ukrainian do:
556 rows at the end of ja.tsv, the 498 message literals the test scan
finds plus the 58 that the other catalogs also carry (web start-up,
relinking, export notes, PostScript read errors).
`ja` joins COMPLETE_MESSAGES, so complete_languages_translate_every_message
keeps it complete. The registry comment, docs/development.md and the
ROADMAP line now list Japanese with the other five. The messages
themselves stay English for agents and MCP.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
- Artboard Options listed its fields by key (Height, Name, Width); it now
lists Name, then Width and Height.
- One-key tool and command shortcuts (V, P, X, D…) matched the character
typed, so with a layout without Latin letters (Persian, Arabic, Russian,
Greek…) they did nothing. When the text typed holds no ASCII character,
the key's position stands in: the V key selects the Selection tool
whatever it types. Latin layouts (AZERTY…) keep the letter typed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
key_equivalents_run_through_the_normal_shortcut_path_once failed about one
full ui-egui run in four ("the menus weren't read again", 5 vs 4). Each
app mirrors its shortcut overrides into a process-wide store every frame,
bumping shortcut_editor::GENERATION, and the plug-in registry's revision is
process-wide too. Tests running alongside with other shortcuts flip them,
which rightly makes the menus read again. The test now checks two idle
frames during which those generations held still (trying up to 50 times),
so it still fails if idle frames re-read the menus on their own.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
As in Illustrator, the Pen shows its continue cursor over an end of any
open path, selected or not, and a click there selects the path and draws
on from that end (reversing it from its first end). While drawing, it
shows a join cursor over an end of another open path, and a click joins
the two into one path, in one undo step, which finishes the path.
path.join takes {ids: [a, b], ends: ["first"|"last", "first"|"last"]}
to join the ends asked for (the Pen's join); without them it joins the
selected paths' nearest ends as before. Joining a path with itself, or
anything that isn't a path, is an error.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Object › Crop Image and the image's Crop Image buttons (Control bar,
Properties) now show a crop box on the selected image, as Illustrator
does, instead of cropping straight to the artboard (which did nothing for
an image inside its artboard). The box starts on the image's part over its
artboard; its handles and its inside drag it, always within the image, and
the image outside it is dimmed. The Control bar shows its X/Y/W/H with
Apply and Cancel; Enter applies and Escape cancels, both back to the
Selection tool. Apply runs object.cropImage {rect}, one undo step.
- tools: the Crop Image tool (cropimage.rs), with crop_box shared by the
tool and the Control bar.
- ui: ui.cropImage (enabled with an image selected), the Control bar's
crop fields and buttons, "Apply" translated (es, fr, it, ru, uk, ja,
zh-hans, zh-hant).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
craft-fonts main holds only Noto Sans Arabic so far (the other eleven are in
storytold/craft-fonts#11), and releases pin an older craft-fonts commit, so
say which builds list them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#763 landed beside #771 with five chunks_exact(4) calls in
crates/affinity/tests/raster.rs, which clippy 1.99's chunks_exact_to_as_chunks
rejects, so cargo xtask ci failed at clippy on main.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
File › Package and Document Info › Font Details looked fonts up with
FontDb::face and compared the face's family with the name the document
stores. A font a document named by another of its names (a PostScript
or localized name, as SVG and PDF files do) was reported as not
available and left out of the package, although the canvas drew it and
exports embedded it.
- Package and Font Details resolve fonts with FontDb::resolve, as
text.fonts and the renderer do. A style the family lacks is still
copied as its closest style; the report now names the style it
copies, and Font Details lists the font as substituted, with that
style's file and permission.
- FontFace::embeddable reads fsType as exports do (Embedding), so fonts
that allow bitmap embedding only (bit 9), or that set the restricted
bit together with reserved bit 0 (0x0003), are no longer copied.
Document Info's permission labels follow the same rule: when a font
sets more than one usage bit, the label names the most permissive
one.
- The Package dialog and its status line name the fonts it didn't copy.
The source audit linked commercial template listings and other pages it
didn't use, and described the environment it ran in. Keep the two sources
the corpus uses, say in a sentence what kinds of sources were left out and
why, and state plainly that reopening our output in Affinity is still to do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The font scan read only the macOS font folders and the Additional
Fonts Folder, so fonts that apps and font managers register with
CoreText from their own folders were missing from the font menus, and
exports set such type in the fallback font. CoreText's font manager
lists them with the installed fonts.
- `mac_fonts.rs` lists the files of the fonts the font manager has
available (`CTFontManagerCopyAvailableFontURLs`, through
objc2-core-text), keeping only file URLs as `open_documents.rs` does
(#433). The desktop app and `vectorcraft-cli` install it as they
install the DirectWrite lister on Windows, and the scan reads the
files outside the font folders in place.
- The scan leaves out faces with no outlines the font engine draws (no
glyf, CFF, CFF2 or VARC table), on every platform. CoreText lists
Apple's PingFang UI interface font (macOS 27), whose outlines are in
an hvgl table; if the scan cataloged it, PingFang SC would count as
installed, and Chinese characters in fallback text, Japanese kanji
included, would draw blank.
- Tests: the list holds the installed fonts by full path, a font
registered from another folder is listed while it is registered, and
a face without outlines the engine draws isn't cataloged.
The Save Swatch Library dialog lists Swatch Exchange (.ase) as a
format, and swatch.library.save takes format: "ase" or a path ending
in .ase. Each solid color keeps its model (RGB, CMYK, Lab or Gray) and
its type (spot, global or process), and color groups stay groups.
Gradients and patterns are left out, and a tint swatch is written as
the color it shows. Without a path or user: true, the command returns
the file as dataBase64.
A swatch exchange library opens under the name of its file. When Save
Swatch Library or Save Graphic Style Library saves to a file, the save
panel suggests the Name from the dialog as the file name, as a save to
User Defined does, and the web app downloads the file under that name.
A name made only of dots and spaces, such as "...", gives the file name
Library (Library.vcswatches) in place of a hidden file.
palette_io::write returns the file's bytes. A property test writes
generated libraries and reads them back, and an import fuzz test loads
mutated copies of a file the writer wrote.
Clippy 1.99.0 added `chunks_exact_to_as_chunks`, which fails `cargo xtask ci`
at its clippy step. Eight uses of `chunks_exact(_mut)` with a constant size now
use `as_chunks(_mut)` (stable since 1.88, within the 1.95 minimum): four in
crates/affinity, three in crates/effects and crates/engine tests, and two in
xtask's sha256. docs/development.md says the gate targets the current stable.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
References can share data, so a damaged fork could make the scan copy one
large resource up to 256 times: cap the total copied like the fork itself.
Off macOS no file has a fork, so the scan doesn't look at each font file twice.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With View › Snap to Grid on, a selection moved with the Selection, Group
Selection or Direct Selection tool lands on the grid: on each axis,
whichever of its two edges and its centre is nearest a gridline (every
subdivision, 9 pt by default) moves onto it. Snap to Grid takes over from
Smart Guides and Snap to Point while it is on, as it does for drawn points.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The test took the first installed family that isn't bundled and expected its
PDF text to stay live. The PDF doesn't embed the font, so the page draws a
stand-in and text stays live only where the installed font's glyphs match it.
With suitcase fonts now found, the first family on a Mac is a condensed
display face whose glyphs don't, so the importer rightly kept outlines and the
test failed. Which family comes first depends on the machine's fonts: the test
now tries up to 200 families and checks the first that stays live.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Replace the Trunk post_build hook packaging/web/copy-fonts.sh with `cargo xtask web-fonts` (xtask/src/web_fonts.rs), so the web build only needs trunk, the wasm32 target and cargo, with no POSIX shell, and works the same on Windows.
The new task reuses crates/text/build/web_fonts.rs for list parsing and URLs, so build.rs and the copier can't disagree on fonts/<sha16>/<file>. It still does nothing without CRAFT_FONTS_DIR, checks each file's SHA-256 against the manifest, and copies the licence next to each face. A listed face that an older craft-fonts checkout lacks is now skipped with a warning, as build.rs does, instead of failing the build. A relative CRAFT_FONTS_DIR is resolved from the workspace root.
Update the references in Trunk.toml, fonts.rs, build.rs, craft_fonts.rs and web-fonts.txt, and document the web build's requirements in docs/development.md. Add tests for the copy, a hash mismatch and the no-op case.
Signed-off-by: Ahmed Moussa <moussa.ahmed95@gmail.com>
macOS lists fonts such as the ones Microsoft Office installs in
/Library/Fonts/Microsoft (Bookman Old Style, Impact, Georgia, Tahoma, ...) but
their files have nothing in the data fork and no font extension: the fonts are
`sfnt` resources in the resource fork. The font scan only read data forks of
.ttf/.otf/.ttc/.otc files, so those families were never cataloged and a file
using one (an EPS naming BookmanOldStyle-Bold) showed the font as missing.
The scan now catalogs a file with an empty data fork and a resource fork by
the fonts in that fork, and loading reads them the same way. The fork reader
(crates/text/src/suitcase.rs) follows Apple's published resource file format,
caps what a damaged fork can make it read and never panics. macOS only.
Tests: the fork reader on a synthetic fork built around a bundled font,
including truncated and garbage forks; a macOS scan of a suitcase file.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Reinhard Schuerer
BFlat
Juan Manuel Verges Solanas
Ikki Nishihara
Dmytro Matviichuk
DWAA1660
dotCooCoo
Matthew Farrow
Ahmed Moussa
Riccardo Balloch