* Use chrono for UTC/calendar conversion instead of hand-rolled copies
Replaces the hand-rolled Howard Hinnant civil/days-from-civil copies and the
leap-year/days-in-month tables with chrono as a workspace dependency:
- apps/vectorcraft/src/logging.rs: timestamp (log line time)
- crates/doc/src/metadata.rs: civil (Unix seconds -> [y,m,d,h,mi,s]) and iso8601
- crates/pdf/src/file_info.rs: unix (date -> seconds) and in_month, used by the
PDF/XMP date parsers (pdf_date, xmp_date)
Output formats and parsing semantics are unchanged and pinned by the existing
tests (PDF D:... and XMP dates, offsets, leap-year rejection, year clamp 0-9999,
pre-epoch clamp). chrono uses only std, so wasm stays safe. AGENTS.md/CLAUDE.md
now say time goes through chrono.
Follow-up to storytold/cadcraft#1785
* Never panic on a clock outside chrono's range; keep 5-digit years unprefixed
chrono's From<SystemTime> unwraps, so a clock past year 262143 panicked
inside the logger. timestamp now uses the fallible DateTime::from_timestamp
and reads such a clock (or a pre-epoch one) as the epoch instead. Years past
9999 are written with {:04} like before, because %Y would add a '+'. Tests
cover the out-of-range and 5-digit-year cases.
* Don't build an unrepresentable SystemTime in the far-future test
UNIX_EPOCH + Duration::from_secs(u64::MAX / 2) overflows Windows'
SystemTime and panicked before timestamp() ran, failing the Windows
test job. Build it with checked_add and only assert where it exists.
* time: preserve old out-of-range date conversion (no epoch fallback)
`DateTime::from_timestamp(...).unwrap_or_default()` silently reset any instant
past chrono's year range to 1970-01-01. Restore the old conversion domain with
a cycle-based helper that routes the calendar through chrono (the calendar
repeats every 400 years = 146_097 days, so month/day/time come from chrono and
only the year is carried in i128, keeping every i64 exact).
- apps/vectorcraft/src/logging.rs: `timestamp` via `utc_parts`; keep the
pre-1970 epoch clamp; u64 secs past i64::MAX saturate, not epoch.
- crates/doc/src/metadata.rs: `civil` cycle-based, keeping its clamp EXACTLY
as the old code had it: the input is clamped to 0000-01-01..9999-12-31 first
and the year is clamped to 0..=9999 afterwards. (With the input clamp the old
`unwrap_or_default()` was already unreachable; this drops it and makes the
conversion explicit, with identical output.)
Pinned old outputs in tests:
logging u64::MAX/2 s -> 292277026596-12-04T15:30:07.000Z (was epoch),
10_000_000_000_000 s -> 318857-05-20T17:46:40.000Z
metadata civil(i64::MIN)[0]=0, civil(i64::MAX)[0]=9999;
iso8601(i64::MIN)=0000-01-01T00:00:00Z,
iso8601(i64::MAX)=9999-12-31T23:59:59Z (year clamp, as before)
* Don't build a year-318857 SystemTime unchecked in the logging test
Windows' SystemTime only reaches about year 30828, so
UNIX_EPOCH + Duration::from_secs(10_000_000_000_000) panicked before
timestamp() ran and failed the Windows test job. Build it with
checked_add and assert where the platform can hold it. The year-10000
case fits everywhere and stays unconditional.
* Name the 400-year range extension as the one exception to the time rule
The rule said never to hand-roll calendar math, but the range fix
(shifting by whole 400-year Gregorian cycles so dates past chrono's year
range keep their old output) is cycle arithmetic, so the rule and the code
disagreed. The rule now names that exception: the library still computes
month, day and time, and only whole cycles are carried in the year.
* Say the library computes month and day; time-of-day is plain arithmetic
Narrows the range-extension wording: the date library computes the
calendar fields, and splitting seconds into hours, minutes and seconds
is ordinary arithmetic, which is what the code does.
A workspace remembers the changes made in it: switching away keeps them
(UiState::workspace_changes, kept with the preferences), switching back
restores them, and Reset <name> returns to the saved layout. Saving a
workspace starts it with no changes; renaming keeps them, deleting drops
them, and deleting the current workspace shows Essentials as it was
left. Saved preferences recover a damaged remembered layout instead of
losing the whole file.
Split from #2965 as a reviewer asked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KcHisW1vt6pkVPYbzL6Cpg
#2027 gave vectorcraft-text and vectorcraft-engine a `bundled-fonts`
feature for an embedder that supplies every document font at run
time, but vectorcraft-ui-egui (and vectorcraft-mcp) named the engine
with its default features, so an embedder that hosts the panels got
the fonts whatever it asked for. Each now has a `bundled-fonts`
feature of its own, on by default, that turns the engine's on, so a
downstream user keeps the fonts unless it asks otherwise. For a
member to leave the engine's default features out, the workspace has
to declare the engine without them, as it already declares
vectorcraft-text; the crates that used the engine's default now ask
for the fonts: the three apps and vectorcraft-testkit. Off,
vectorcraft-text's copy of the fonts (3.4 MB) is left out; theme.rs's
own copies of five of them, the panels' interface faces, stay either
way.
Nothing the workspace builds changes: for every workspace member and
for the whole workspace, with and without dev-dependencies, for all
targets, wasm32-unknown-unknown, aarch64-apple-darwin and
x86_64-pc-windows-msvc, `cargo tree -e normal[,dev] -f '{p} [{f}]'`
resolves the same features for every package as on main, apart from
the new feature names (the engine's `default` no longer listed;
ui-egui's and mcp's `bundled-fonts` and `default`).
A test of the panels can't see their configuration without the fonts
(their dev-dependencies turn them on), so `cargo xtask ci`'s
`no-bundled-fonts` step now also reads that tree for
vectorcraft-ui-egui and vectorcraft-mcp: with their default features
it must show the fonts, and without them, for any target (a wasm-only
dependency too), it must not. Documented in docs/development.md
(Fonts › Leaving out the bundled fonts).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(window): adopt shared craft-window for Linux decorations
Decide the window frame on Linux/BSD via the shared craft-window crate
(storytold/craft-libs) instead of VectorCraft's unconditional
`CUSTOM_TITLEBAR = !macos`, so the app matches the rest of the Craft
suite per desktop.
- Probe `craft_window::Choice::probe(Prefix::new("VECTORCRAFT"))` once at
startup; custom titlebar is on when the frame is `Frame::App(_)` and the
user has not opted into the system title bar.
- Preserve the `systemTitleBar` preference: opting in forces the native
(desktop) frame back on, as before.
- Keep macOS and Windows behavior exactly as today (decided by
`CUSTOM_TITLEBAR`, not the probe): macOS integrated title strip,
Windows app-drawn caption. craft-window returns `Frame::Desktop` for
both; VectorCraft's overdraw is its own choice.
- The only behavioral change: KDE/Wayland now keeps KWin's native
decorations (Frame::Desktop) instead of always drawing a custom
titlebar; GNOME still draws its own (Frame::App).
- `VECTORCRAFT_TITLEBAR` override and the `systemTitleBar` pref both still
work (the env var is craft-window's override for the `VECTORCRAFT` prefix).
NOT MERGEABLE YET: craft-window is open PR storytold/craft-libs#217; the
dep points at the PR branch and must be repointed to a merged `rev`
before this can merge.
* fix(window): native frame when X11 is forced; pin craft-window to merged rev
- When Preferences > Linux Display Server > X11 forces the window onto
XWayland, treat the frame as the desktop's (native X11 decorations)
rather than letting a GNOME/Wayland probe pick Frame::App and draw our
own title bar on what is really an X11 window. An explicit
VECTORCRAFT_TITLEBAR override still wins (checked via Choice::overridden).
- Repoint craft-window from the (now-merged) feature branch to the squash
commit on storytold/craft-libs (craft-libs#217); regenerate Cargo.lock.
- The WAYLAND_SOCKET concern raised in review is handled in the merged
craft-window crate itself (the probe skips when WAYLAND_SOCKET is set).
Verified: cargo build + test -p vectorcraft (131 tests) pass against the
merged dep; windows7 (--no-default-features) still compiles; rustfmt clean.
* fix(window): only force native frame when X11 libs are actually present
Follow-up to review: the XWayland title-bar override now also checks
xwayland::x11_libraries_found(), matching the launch path's own condition.
If X11 is requested but the libraries are missing, startup stays on
Wayland, so the frame decision must keep the probe's app-drawn result
rather than flipping to native. Verified: tests pass, rustfmt clean.
- ui-egui: the font-install poll (#2576) only runs on native builds, but its
timestamp field, its initializer and FONT_INSTALL_POLL were compiled for
wasm too, where nothing read them. They're gated like the code that reads
them.
- engine: `Session::store_profile` is only called by the native Load
Profile (cmd/colormgmt.rs); gated the same way.
- app (macOS): `App::logic_frame` checked `!file_dialog_open()` twice, one
`if` nested in an identical one, with the comment repeated. One check.
`cargo xtask wasm` runs `cargo check` without `-D warnings`, so these
warnings didn't fail CI; after this change it prints none.
On Linux the save dialog is the XDG desktop portal's, shown through
rfd. rfd sends no `current_filter` and reads only the file name from
the portal's answer, and the GTK file chooser keeps the typed name
when another file type is picked. With "PDF-compatible .ai" picked,
`poster` typed for a new document was saved as a native
`poster.vectorcraft`, and so was the suggested
`Untitled-1.vectorcraft` left as it was.
When a save offers two or more file types (Save As, Save a Copy, Save
as Template and the first Save of a new document), the app now calls
the portal's SaveFile itself over D-Bus. The dialog starts on the
suggested file's type, and its answer names the type picked. A name
typed without an extension, or the suggested name left as it was,
gets that type's extension. A name typed with another save format's
extension, other than the suggested file's (for Save as Template, a
template format's), is saved in that format, and any other extension
gets the picked type's added. The dialog keeps the window as its
parent on X11 and Wayland, and rfd shows the dialog when the portal
can't. The same code runs on the BSDs; Windows, macOS and the web
build keep their dialogs.
Docs: docs/target-app-parity.md.
On a Wayland session a drawing tablet's pen could not draw or click:
winit 0.30 does not bind the Wayland tablet protocol, and neither KWin
(since Plasma 6.3) nor GNOME's Mutter turns tablet input into pointer
input. Linux Display Server's X11 (XWayland) choice let the pen work as
a mouse, but it opened the window through XWayland on every start, also
without a tablet.
Linux Display Server has a third choice, X11 (XWayland) When a Pen Is
Attached (x11WithPen). It opens the window through XWayland on a
Wayland session only when /proc/bus/input/devices lists a pen as the
app starts: a device with the BTN_TOOL_PEN key and the ABS_X and ABS_Y
axes, the check PhotoCraft's tablet.rs makes. Automatic stays the
default. On Linux the Performance page shows a note for tablet pens
that don't draw.
README.md named KDE alone; the KWin commit's message says Mutter does
not emulate tablet events either, so README.md and docs/development.md
now name GNOME. The Pen pressure row of docs/hardware-parity.md said
Linux has pressure through XWayland; winit 0.30 reports none on X11
either.
Docs: README.md, docs/development.md, docs/hardware-parity.md,
docs/target-app-parity.md.
On a machine busy with other builds, five tests on main failed now and
then and passed when run alone. The canvas test for dragging an object
with raster effects gave the render worker 10 s for its frame, which
took 0.10 s alone and 5.8 to 9.4 s with four ui-egui test runs at
once. The three tests_sysclip_probe tests gave the clipboard probe's
thread 2 or 3 s. The control server's persistent-requests test sent
each request with writeln!, which sends the JSON to the socket in many
small writes, and a pause of more than the test's 120 ms line deadline
between two of them closed the connection.
The canvas test now waits up to 60 s for the worker, as another test in
canvas.rs does. The probe tests wait up to 60 s too. The control server
test sends each request line in one write and keeps its 120 ms line
deadline. Only test code changes.
Each swatch or graphic style library now opens in a panel of its own,
`library:<kind>:<library id>`, in the shared docking system: several can be open
at once, and they float, group as tabs, dock and close like any panel.
Opening a library that is open brings its panel forward; Previous / Next
Library show the neighbouring library in the same panel.
The library panel menu gains Persistent. The panels' view, Show Find Field
and Persistent are kept in the UI preferences (`library_panels`); at startup
the persistent panels reopen where the saved layout had them, a library
loaded from a file read again from its path (a missing file is skipped with
a status-bar note). `ui.libraryPanel {panel, library?, view?, showFind?,
persistent?}` reaches all of it; docs/mcp.md documents the panels.
Install Plug-in (plugin.install) now keeps the plug-in's module in a
Plug-ins folder next to the preferences, as Illustrator keeps plug-ins in
its Plug-ins folder, and the desktop app and vectorcraft-cli install the
ones there at startup (a module that doesn't load is skipped with a log
record, at most 512 MB are read). Remove Plug-in deletes it from the
folder. The Additional Plug-ins Folder preference is still loaded from
where it is. Confined MCP servers and VECTORCRAFT_NO_PREFS keep nothing,
as with the PDF preset and colour profile stores.
Docs: docs/mcp.md, docs/plugins.md, docs/target-app-parity.md.
When pressing Ctrl+P to open the Print dialog and clicking Setup (or Settings), print.printerSetup previously discarded the selected printer name and unconditionally launched explorer ms-settings:printers. On Windows 10, this opened the Windows Settings Printers and scanners page, displaying the list of installed printers instead of the printer driver printing preferences.
Launch Windows native Printing Preferences dialog for the target printer using rundll32.exe printui.dll,PrintUIEntry /e /n printer, falling back to the system default printer or Settings if no printer is available or detected. Add unit tests covering command construction and fallback behavior.
Fixes#2569.
After the desktop save panel closes, the app adds the file type's
extension to a name typed without it (or with one the save does not
write). The panel asked only about the name as typed, so on Linux
typing `card` with `card.vectorcraft` in the folder wrote over
`card.vectorcraft` without a question. On Linux the save panel is the
XDG desktop portal, and its GTK and GNOME file choosers return the
name as typed. Windows and macOS add the extension themselves only
when the panel lists file types, and the panels of Document Info's
Save, the presets dialogs' Export…, the CSS Properties export, the
swatch and graphic style library saves and the Libraries panel's
export list none.
When the added extension names a file that exists, Save, Save As, Save
a Copy, Save as Template, the exports and the other writes through a
save panel now ask "“card.vectorcraft” already exists. Replace it?"
first (dialog `replaceFile`). OK runs the write again with the same
answer and replaces the file. Cancel writes nothing and shows the
dialog it was asked over again. The host reports whether a file exists
through a new `Services::exists`.
Docs: docs/control-protocol.md, docs/target-app-parity.md.
The EPS reader read PostScript by any name, but the Open and Place
dialogs hid .ps and .epsf files, and a .ps or .epsf file whose header
is only %! could not be opened or placed ("not a format VectorCraft
reads"). The reference app lists .eps, .epsf and .ps under Open and
Place.
- A read-only PostScript format (ps) joins the format table. A .ps
file opens and places through the EPS reader and reports the
ps format. document.export does not write .ps; file.print does.
- .epsf is an EPS extension.
- OPEN_EXTS and PLACE_EXTS list epsf and ps, so the dialogs, the CLI
help and the MCP open_file description list them.
UNASSOCIATED_EXTS lists them too: the macOS bundle declares neither.
- A .ps or .epsf file whose header is only %! now pastes as EPS when
copied in a file manager, as one with a %!PS header already did, and
clipboard.importEps reads PostScript whose header is only %!, which
it rejected before.
Docs: docs/mcp.md, docs/file-format-parity.md (.ps row),
docs/target-app-parity.md.
PDF presets are kept one file each in a PDF Presets folder, ICC profiles as copies in a Color Profiles folder, next to the preferences; the desktop app and vectorcraft-cli use them. Adds color.removeProfile.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On macOS, the save sheets of Save As, Save a Copy and the first Save of
a new document have a Format menu again, and the Save as Template sheet
now has one too. The panel is an NSSavePanel built in
apps/vectorcraft/src/mac_save_format.rs, and the menu is its accessory
view, set before beginSheetModalForWindow:completionHandler: shows the
sheet. The menu from #1235 was added after rfd had begun the sheet, and
recent macOS versions dismissed the sheet at once (#1800, #1831,
#1518); #1855 removed it.
Choosing a format changes the extension in the name field and puts
that format's extension first in the allowed file types, which AppKit
adds to a name typed without one. A name typed with another listed
extension is saved in that format. Save panels with one file type and
the open and folder panels are still shown by rfd.
Docs: docs/target-app-parity.md, docs/ui-parity.md.
A folder or app bundle, a link to another kind of file, or a file with
any other extension is refused with a message instead of being handed to
the system's default app. On Windows, a path containing a quotation mark
is refused before it reaches Explorer.
The install folder is where the user sees the app: next to VectorCraft.app on
macOS (nothing is written inside the signed bundle) and next to the .AppImage
file on Linux ($APPIMAGE; the mounted executable's folder is read-only).
The first time VectorCraftData is created, the settings in the user's profile
are copied into it, so a portable zip that gains portable.txt keeps its users'
settings; the originals stay, a second start copies nothing, a copy that fails
halfway is logged and left as it is.
VECTORCRAFT_NO_PREFS never looks for the marker or creates the folder; a
portable copy skips the one-time read of DrawCraft's old settings; the
unwritable-folder warning goes to the log only (eprintln! panics on a closed
standard error).
When a graphics driver crashed the process while the window started,
as Intel's Direct3D 12 driver (igd12umd64.dll) did on an Intel UHD
Graphics 605 (#2243), the next start picked the same adapter again. The
restart without a failing adapter and the first-frame watch run inside
the process, so the crash ended them along with it.
Each start now writes graphics-start.json in the log folder before the
graphics start, adds the adapter once one is picked, and clears both
when the first frame reaches the screen or the app ends without
crashing. A start that finds an earlier start still recorded there
leaves that adapter out, as a restart does, logs it and shows a note in
the status bar. When the crash came while the adapters were being
listed (wgpu's Direct3D 12 backend creates a device on each GPU to list
it), the next starts list OpenGL's adapters alone, except on macOS. The
adapters stay left out until the app version or the Graphics Processor
preference changes, a start shows its first frame on one, or no adapter
is left. WGPU_BACKEND and WGPU_ADAPTER_NAME choose regardless of the
record.
Docs: docs/development.md, docs/target-app-parity.md.
--help and -h used to be taken for a file to open: from a terminal, SSH or an
agent's shell the window opened, or the start hung or failed without a display.
They now print the desktop app's options and point to vectorcraft-cli for
headless work, exiting before logging, preferences or the window, as
--version does.
The Color Picker gets an eyedropper button beside the New and Original
chips, and `ui.sampleScreenColor {stroke?, picker?}` shows the system's
color sampler. The command answers at once. When the person clicks a
color anywhere on the screen, the color becomes the fill or the stroke
through paint.setFill / paint.setStroke as one undo step, or with
`picker` it becomes the open Color Picker's color, which previews on the
selection as any pick does. Escape cancels the sampler. The sampled sRGB
color goes through Cms::srgb_to_rgb, so in a Display P3 working space a
color inside sRGB shows as it did on the screen. `ui.inspect` reports
`screenSampler` while the sampler is open.
- macOS uses NSColorSampler (apps/vectorcraft/src/mac_screen_color.rs).
- The web app uses the browser's EyeDropper API where the browser has
one.
- Without a sampler (Windows, Linux, other browsers) the button is gray
and the command returns an error. Keyboard Shortcuts lists the
command, so it can be given a key.
Docs: docs/control-protocol.md, docs/target-app-parity.md.
The worker answered "timeout" after 60 s and left the request on the queue, so
the app ran it whenever it next drained the queue: an edit nobody was waiting
for, and a second one if the caller had retried. A request now leaves the
queue once: taken by the app, or given up by its caller, whichever comes
first. The timeout reply says which ("will not run" / "still running").
The preferences file was written only when the app quit normally, so settings
confirmed with OK were lost when the session ended any other way (the system
shutting down, the process being ended, a crash). OK now asks the host to
write the file at once.
The Format menu added to the save sheet after rfd began it (#866) made recent
macOS versions (26.6.2 and 27.0.1 reported) dismiss the sheet at once: Save As,
Save a Copy and Save of an untitled document flashed and answered as cancelled,
while panels without it (Open, Export As) worked. The accessory is withdrawn
until it can be set before the sheet is presented; the format still follows
the extension typed in the name field, as before #866.
Fix#1726 with idle, absolute line and reply-write deadlines and connection-slot guards. Add real-loopback regressions and document reconnect behavior.
Validated with cargo xtask ci: all seven gates passed; 6156 tests passed, zero failed, seven existing ignored.
winit 0.30 gives an undecorated window WS_CAPTION | WS_BORDER; on some Windows
systems the compositor still reserves that frame, so the content is drawn
offset while the pointer isn't, and the cursor no longer points at what is under
its tip. Strip the two styles once and again each frame (winit restores them on
maximize/restore), as PhotoCraft does (storytold/photocraft#2401).
The control server wrote each reply with writeln!(out, "{r}") on the
bare TcpStream. serde_json formats a Value one piece at a time, and
each piece (a brace, a quotation mark, a key, a colon, a number) was
sent as a write of its own: the 17 MB document.inspect reply of a
document with 50,000 objects on one layer was sent in 7,857,200 writes.
On Windows 11, sending that reply to a client over loopback took 34 s,
and sending the 28 MB document.json reply took 57 s.
vectorcraft-cli mcp reads every reply from a running app over this
connection.
write_line now writes each reply, each refusal and the "too many
control connections" line through a 64 KiB BufWriter, then flushes
it. The 17 MB reply takes 262 writes. Sending it now takes 0.09 s,
and sending the 28 MB reply takes 0.2 s. The bytes are the same.
Docs: docs/target-app-parity.md.
Paste read PDF only from the Windows clipboard and EPS nowhere, so on macOS a copied
equation or drawing pasted nothing new (the last thing copied in the app came back).
- macOS: the pasteboard's PDF, EPS and SVG are read, ahead of text and bitmaps.
- Windows: EPS is read beside PDF.
- Everywhere: an .eps file copied in a file manager pastes.
- clipboard.importEps loads an EPS into the clipboard as document.open reads one.
--control took the next argument as its port and dropped it when it wasn't a number, so the app
started without a control server and `--control --automation-read-root DIR` opened DIR as a
document. automation_args::ControlArg now reads --control <port> or --control=<port> once and
requires a port from 1 to 65535; VECTORCRAFT_CONTROL_PORT must be one too, checked only when
--control isn't given so --control and --version win over a bad value. Errors go through the
same exit as the root flags (#1157).
Tests: unit tests in automation_args.rs; tests/automation_args.rs runs the binary for each bad
form and checks the exact message and a prompt failing exit.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Save serializes JSON straight into the compressor and writes images from their Arc into the
file; open reads each image into its own buffer, and NodeKind reads its tag first instead of
buffering every node. Already-compressed images are stored rather than deflated again in the
same gzip stream. Limits apply per section, so any file VectorCraft writes opens again. Data
Recovery writes each image once, content-addressed, instead of every interval. Native files open
on a background thread. Files stay readable by main both ways; JSON-only saves are identical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
Localize native/web file filters, diagnostic labels, numbered presets
and contributor tooltips with English fallback and Czech plural forms.
Fix gradient adjective agreement and disambiguate Japanese printer marks
in PDF/Print choices and summaries. Keep file extensions and names intact.
Add Czech catalog rows and regression coverage; update localization evidence.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On a hybrid laptop whose Wayland compositor runs on the discrete GPU
(Hyprland with AQ_DRM_DEVICES naming the NVIDIA card first), the built-in
panel still hangs off the integrated GPU, so the display rule put it first,
and the window stayed empty: the compositor took the frames and showed
nothing, with no error, so neither the restart on a failed adapter (#502)
nor the first-frame watch (#964) noticed.
Which GPU drives a display stands in for which GPU the compositor draws
with. Under Wayland the compositor says so itself: the main_device of its
linux-dmabuf feedback (version 4), the device it imports client buffers on,
which Mesa's Vulkan device-select layer reads for the same purpose.
compositor_gpu.rs asks for it on a short-lived connection, maps the dev_t
to PCI ids through /sys/dev/char, and under Automatic adapter_order ranks
that adapter ahead of the display GPUs. No answer (X11, an older compositor,
a GPU no adapter matches) leaves the existing rules in charge; a user's own
choice, WGPU_ADAPTER_NAME and the restart skip list still come first.
wayland-client and wayland-protocols are already in the tree through winit;
they become direct Linux-only dependencies.
Dylan M. Taylor