Moves both pins from 94a3fb5 (the PR head at the time) to 0d76ea0, the
merge. That brings in the last two review fixes: an app briefly in use is
stopped when agents go off, and an apps.json that can't be read counts as
agents off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gateway now holds only an app's connection during a call (listing,
searching and describing answer at once) and asks the switch again once a
queued call has its app, so it never starts after agents are turned off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The launcher writes, reads and removes ArtCraft/mcp/apps.json through
`craft-ai-agents` (storytold/craft-ai-integrations#197), the same crate the
gateway reads it with and every Craft app checks the switch with, instead
of its own copy of the format.
- Every app is started as plain `<app>-cli mcp`: by the suite's rule that
command works in the open window when the app runs and accepts agents, and
headless otherwise. The per-app argument table (`--live` for PhotoCraft,
`--headless` for VectorCraft) and reading each CLI's `--help` to see
whether it takes them are gone.
- Gateway and craft-ai-agents pinned to the head of #197 (35e0352).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gateway now asks the launcher's app list before every call: with the
switch off it refuses every tool and stops the app servers it started, and
it no longer falls back to CLIs on PATH.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
apps/craft-ai-mcp builds the gateway from storytold/craft-ai-integrations
(crates/ai-mcp, craft_ai_mcp::cli) as craft-ai-mcp, and every package puts it
beside the launcher, where Settings › AI agents finds it:
- Windows: built with the launcher, PE-checked as a console program, signed,
in the MSI (the launcher's component) and the portable zip; listed in the
install policy's payloads
- macOS: lipo'd and placed in Contents/MacOS, signed with the hardened runtime
before the main executable
- Linux: /usr/bin in the deb and rpm (nfpm), the tarball and the AppImage;
/app/bin in both Flatpak manifests
- FreeBSD: bin/, checked like the launcher
The gateway needs only serde, serde_json and dirs. It is pinned to the head
of storytold/craft-ai-integrations#197 until that merges.
* Load and verify the signed website catalog with bundled fallback
* Preserve signed fixture bytes across platforms and fix web lint
* Document the opt-in live catalog integration check
Each installed app's menu gains "Launch options…": flags typed there are
saved in Settings.launch_flags (app.ron, keyed by catalog id) and added to
the app's command line on Open. core::args::split turns the text into
arguments (whitespace, "…"/'…' quoting, backslashes kept for Windows
paths) and refuses an unclosed quote, both in the dialog and at launch.
Flatpak, launcher-unpacked, plain binary and macOS .app installs all pass
the flags through; strings are translated in every complete catalog.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The UI's language is set per frame from the saved preference (`LauncherUi::logic`
calls `i18n::set_current`), and a start that never got a frame has set nothing:
`i18n::current()` is still English, so the dialog answered a German user in
English.
The failure path now reads the saved settings itself (`saved_settings`, the same
read a tray start does) and uses the preference as saved. It skips the UI's
CJK-font gate on purpose: this dialog is a native one, drawn with the system's
own faces, so a language egui could not have drawn is still readable.
Verified end to end: with `language:"de"` saved, forcing the Direct3D 12 failure
on the 32-bit build shows "ArtCraft Launcher konnte sein Fenster auf
Direct3D 12 nicht starten: …".
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The test called `wgpu_options` directly, so it never touched the line that
passes the attempt to `native_options`. Dropping that argument would have left
the test green while every start retried the backend that had just failed.
Reading `native_options(false, attempt).wgpu_options` covers the whole path.
Also records why the macOS startup label names Metal alone. The default mask
asks for GL as well, but wgpu's GL backend loads `libEGL.dylib`, which macOS
doesn't ship, so a start there finds no OpenGL adapter to name.
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Every test the two commits before this one add now says what it holds the
launcher to, instead of only being named for it. The docstring coverage check
on the PR asked for it, and the reason is worth writing down anyway: the
retry-once rule, why a non-renderer failure is not retried, and why the
fallback is OpenGL are all decisions a later change could undo by accident.
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A launcher whose window couldn't be created exited with code 1 and said
nothing. The reason only reached the log file, which someone whose launcher
never opens does not know to read, so a startup failure looked like a launcher
that was simply broken.
The failure path now shows a native error dialog naming the backend that was
tried and the error wgpu reported, in the user's language. The text is a `@msg`
template, so it is translated where it is shown and the backend and error text
are carried over into the translation; the backend names (Direct3D 12, OpenGL,
Vulkan, Metal) stay in English, as app names do.
Fixes#1081
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The launcher built its window once, on wgpu's first-choice backend, and exited
with code 1 when that backend couldn't create `egui_pipeline`. Nothing appeared
on screen, so it looked like the launcher had never started at all: on
Direct3D 12 some drivers refuse that pipeline (#111, #1081), and a 32-bit build
refuses it on every Direct3D 12 adapter even where the 64-bit build of the same
commit draws fine.
`StartupBackends` now tries the fallback backend after a start that couldn't
create its renderer. The fallback is OpenGL everywhere it exists (Metal is
macOS's only backend, so there is nothing to fall back to there). `WGPU_BACKEND`
still names the one backend a start uses, and a window that failed for another
reason — no event loop, no display — is not retried, so its error isn't buried.
Reproduced with the shipped `windows-x86` 0.14.0 binary and with a local i686
build: `egui_pipeline` fails in `ID3D12Device2::CreatePipelineState` with
E_INVALIDARG (0x80070057) on the RTX 4080 SUPER and on the AMD integrated GPU,
in both debug and release, while the x86_64 build of the same commit starts on
Direct3D 12. With this change the 32-bit build logs the failure and opens its
window on OpenGL. The 32-bit Direct3D 12 failure itself is still open: it is
below this repository, in wgpu's Direct3D 12 pipeline creation.
Part of #1081, not all of it: the issue also asks for a readable error naming
the backend when every backend fails. That needs a new user-facing string and
its row in every `complete_menus` catalog, so it stays open rather than closing
here.
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* macOS: the Dock icon brings the window back after closing it (#456)
Closing the window to the menu bar ordered it out, and clicking the Dock
icon afterward brought nothing back: the Dock sends a "reopen" Apple
Event, which neither winit 0.30 nor eframe passes on, and handling it
directly needs unsafe code.
On macOS, closing to the tray now hides the app, as Cmd-H does. macOS
unhides a hidden app by itself, window and all, when its Dock icon is
clicked, it is opened again from Finder or Spotlight, or it is picked
with Cmd-Tab. "Show ArtCraft Launcher" unhides it before showing the
window. A start in the tray keeps ordering the window out, as the app
is still being activated then, which would unhide it.
* macOS: say why Shell.visible isn't resynced after the Dock unhides the app
A left click on the tray icon opens its menu on macOS, so the toggle that
reads this flag never arrives there.
* fix(logging): drop HTTP wire dumps whatever RUST_LOG says
ureq_proto traces every byte it sends as hex-and-ASCII rows, the
Authorization header with the optional ARTCRAFT_LAUNCHER_GITHUB_TOKEN
included. The builder capped ureq_proto at debug, but env_logger lets the
longest matching module win, so RUST_LOG=ureq_proto::util=trace still
wrote the token to standard error and the log file, split across rows
where the Bearer redaction can't see it.
Wrap the env_logger Logger in a gate that drops ureq_proto trace records
before they are formatted, whatever the filter says.
* test(logging): pin wire-dump test filters against inherited RUST_LOG
* fix(installer): install new Windows apps for everyone, in Program Files\ArtCraft\<App>
A fresh MSI install now asks Windows Installer for the machine-wide context
(ALLUSERS=1) instead of taking the package's own default, which for a
dual-scope package is per-user (%LOCALAPPDATA%). Updates are unchanged: they
keep the context of the copy they replace. README row restored to Program Files.
* fix(ui): show the real default install folder, not just "System default"
The install-folder fields (the per-app Install dialog and Settings'
Default install folder) now show the folder a fresh install really
uses: Program Files\ArtCraft for a Windows MSI (machine-wide, from
the ALLUSERS=1 scope), the launcher's own apps folder for a portable
ZIP, /Applications on macOS. The host answers through the new
Host::default_install_folder, computed by native install::default_folder;
where no single default exists (web build, Linux system packages,
FreeBSD) the field keeps saying "System default". The same value is
the start folder of Browse, so it no longer starts at a literal
"System default" string.
* fix(review): the default folder follows the package and the installer
CodeRabbit review of #972: the Windows default now follows the selected
package's architecture (an x86 MSI picked because a release ships no
x64 one installs under Program Files (x86)), and answers nothing when
the architecture is unknown, so the field says "System default"
rather than guessing. The macOS default comes from the installer's own
applications_dir, so a machine where /Applications is not writable
shows ~/Applications. Browse now starts at the nearest folder that
exists, so a default created only by the first install
(Program Files\ArtCraft) no longer falls back to the picker's
unrelated start location. The desktop host memoises its answer because
the macOS path probes the disk and the fields redraw often.
* fix(review): keep the Linux and macOS builds of default_folder warning-free
The ubuntu Clippy gate caught what the dev machine cannot check (ring
needs a Linux C compiler): the Arch import is Windows-only, the macOS
arm consumes its unused parameters, and the Linux test calls the
three-argument signature.
* fix(ci): import Arch on every target, default_folder's signature uses it everywhere
0f5bd68 gated the import to Windows but the signature names Arch on all
targets, so the Linux clippy job failed with E0425.
* fix(review): the folder hint matches the real install and touches nothing
CodeRabbit review of #972.
- The install dialog's default follows the package the backend will really
install: the preferred formats first, then every other format this
platform installs, as choose_kinds does. A Portable ZIP preference with a
release that ships only an MSI now shows Program Files, not the apps
folder. The choice lives in the new install_target module.
- Settings covers every app, so it shows a folder only when every app that
can be installed shares one; a 32-bit-only app beside a 64-bit one leaves
it at "System default" instead of using the launcher's architecture.
- The macOS hint no longer borrows the installer's probe, which wrote a
fixed-name file in /Applications (following a planted link) and created
~/Applications when it fell back. Both now use write_probe (fresh name,
create_new), and only the installer creates ~/Applications.
- The desktop host caches each question, since Settings now asks per app.
* Migrate Launcher to egui 0.37 and preserve current-main behavior
* egui037: record final published pins and local qualification
* fix(ui): gate CJK language drawing until fonts are installed
* Document migrated launcher UI contracts and regression invariants
- Redact Bearer tokens with multiple spaces and add regression test.
- Explicitly disable proxies on direct and direct_download ureq agents.
- Add direct_download agent with download timeout profile for loopback downloads.
- Preserve proxy username, password, and NO_PROXY exclusions in proxy::pick().
with_no_window_it_sleeps_until_woken read Instant::now() after spawning
the thread whose 200 ms pause it measures, so a runner that held the
test thread back between the two failed started.elapsed() >= pause
(seen on main's ubuntu CI). Read the clock first.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The close button parks the launcher in the tray by design, but to
someone who doesn't know that, a window that vanishes with nothing to
show for it looks hung, and the reports end with killing the process
(#45, #215).
The first time the window closes to the tray, a system notification
says the launcher is still running, that it's in the tray, and that
Settings can turn this off. `LauncherUi::hidden_to_tray` sends it
through the host and records it in `Settings::tray_hint_shown`, so it
happens once per install. The desktop shell calls it from the
close-request path only, for both ways of going to the tray; a window
hidden from the tray icon, or started hidden at login, gets no hint.
Two strings, with rows in all ten complete catalogs. A UI test checks
the hint is sent once and recorded.
Co-authored-by: YangXZ2008 <141027290+YangXZ2008@users.noreply.github.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
egui requests wgpu's default limits on every backend but GL, and device
creation fails when the adapter falls short of any one of them. The
Raspberry Pi 5's V3D driver (Mesa V3DV) allows 15 inter-stage shader
variables against the default 16, so the launcher exited before the first
frame with "Limit 'max_inter_stage_shader_variables' value 16 is better
than allowed 15".
Wrap egui's device descriptor so each requested limit is lowered to the
adapter's where it is below the request. GPUs that meet the defaults get
the same request as before.
Includes regression coverage with the V3D limit.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
`std::env::args` panics on an argument that is not valid UTF-8, and it panics
while collecting them, before the launcher looks at what it was given. Any
start-up with such an argument ended at
internal error: panicked at library/std/src/env.rs:878:51:
called `Result::unwrap()` on an `Err` value: "/tmp/caf\xE9.png"
with exit 101 and the launcher's own panic report, even for `--version`, which
the packaging smoke test runs on the staged binary.
The launcher looks for two flags and ignores everything else a desktop start-up
brings along (`-psn_0_…` from Finder, a file dropped on the icon, an old login
entry's leftovers), so an argument it cannot read is one it would have ignored
anyway: `args_os`, and a warning in the log naming it.
tests/arguments.rs covers it, and `--version` with the arguments a desktop
adds. Without the change the first test fails with exit 101 and that report.
wgpu 30 turns on InstanceFlags::VALIDATION_INDIRECT_CALL by default and
compiles its indirect-validation compute shaders in request_device. On
Intel Iris and Iris Pro Macs, Metal's compiler fails on them with an
internal error, the device is lost and the launcher exits at startup.
wgpu_options() on macOS now removes the flag when the build is x86_64,
then applies with_env() so WGPU_VALIDATION_INDIRECT_CALL=1 turns it back
on. The launcher issues no indirect draws or dispatches. Apple silicon,
Windows and Linux keep wgpu's default. This follows PhotoCraft's fix in
storytold/photocraft#3171 (apps/photocraft/src/gpu_startup.rs).
Fixes#344
`?lang=de#apps` gave the tag `de#apps`, so the page fell back to the browser's
languages. The parsing is a small tested function now.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Every real quit sent ViewportCommand::Close, and eframe dropped the
window inside AppKit's run loop. With the window focused, AppKit threw
in -[_NSTouchBarFinderObservation invalidate]. craft-quit's
on_close_request cancels that close and quits through
-[NSApplication terminate:] instead. It runs last in logic, so a close
put away in the tray (CancelClose) is left alone.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
On Windows the launcher finished its own update with a detached
powershell.exe started with -ExecutionPolicy Bypass -WindowStyle Hidden.
The script waited for the launcher to exit, ran the new MSI or unpacked
the portable zip with Expand-Archive, deleted the download and started the
new exe. Antivirus heuristics score that sequence, and the command line
sits in the signed exe as a string.
The update helper is now a copy of the launcher itself, put in
<cache>\launcher-update-helper and started with --finish-update. It waits
for the old process with OpenProcess and WaitForSingleObject (winsafe),
runs msiexec /i /passive /norestart and checks its exit code, or unpacks
the portable zip with the checks portable app installs already use, and
then starts the launcher again. main() hands --finish-update to the helper
before logging, the single-instance claim and the window.
The helper finishes only the launcher's own update:
- the exe is this launcher's MSI or portable installation, and the
helper, the download and the staging folder are in that installation's
cache;
- the MSI, every program in the portable zip and the exe it starts carry
a valid Authenticode signature from the helper's own publisher
(verifysign, through WinTrust).
The launcher checks its own signature, the installed exe and the new MSI
before it quits, so an unsigned build reports that it doesn't install
updates instead of closing. A failed update clears the self-update marker,
so the next check offers it again, and the next start logs the error.
After unpacking a portable app, the launcher checks the exe's
ProductVersion against the release. It started a hidden powershell.exe
-NoLogo -NoProfile -NonInteractive -Command to read
FileVersionInfo.ProductVersion, which costs a PowerShell start on every
portable install and is one of the command lines antivirus heuristics
score.
executable_version now reads the ProductVersion string from the file's
VERSIONINFO with GetFileVersionInfo and VerQueryValue (winsafe), in the
first language the file lists, and no process is started.
The expected path was written out as E:\Suite\ArtCraftLauncherData, which is
one file name on Linux, so the test would have failed on the ubuntu-latest job
the first time it ran. Joining the components gives the same check on both
platforms.
portable.txt puts ArtCraftLauncherData next to the exe, but the settings still went
to eframe's storage folder in the user's profile, so a suite carried on a stick lost
its settings on another computer.
path() now returns the suite's own data folder when a portable suite is detected, and
main.rs passes it as persistence_path so eframe writes the window state to the same
file. Nothing changes for a normal install.
* feat(install): choose where Craft apps are installed
Settings gets a default install folder, and a fresh install asks for the
app's folder (prefilled with the default). Updates stay beside the copy
they replace. Works for MSI (INSTALLFOLDER), DMG bundles, and AppImages;
on Linux a custom folder means AppImage. Flatpak and the web build keep
the system location and skip the prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(windows): integrate portable installs with custom locations
* fix(install): ask before writing over something already in the install folder
Staging used a predictable `.<name>.partial` path beside the destination, so a
file the user kept under that name could be overwritten (AppImage) or deleted
(macOS). Stage in a fresh private folder instead.
A new install whose destination already exists now stops before downloading
and asks "Replace what's there?". Replace installs over it: the AppImage or
.app is swapped in, a portable install writes over same-named files only, and
the MSI installs into the existing folder. Cancel leaves everything as it was.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(install): a planted .partial symlink can't redirect AppImage staging
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Skyfuller1303 <skyfullerdesign@gmail.com>
Present Windows DX12 through DirectComposition while retaining the presentation-system override and existing graphics selection.
Use whole-pixel glyph placement for the bundled fonts and inset the card hover rule within its hairlines.
Refs #339.
* Add a hand-written details page for each Craft App
core::details holds, for every app in the suite, a headline, a few
paragraphs, a list of features, where it runs and the list of its
pictures. The copy is the launcher's own: written from each app's page on
getartcraft.com and, for the five newer apps, from its README. It is
compiled in, so it works offline and in the web build, and it leaves out
figures that go stale (effect counts, timings).
The pictures are not in this repository. A page names them (a key and a
caption each) and AppDetails::screenshot_url turns a key into the address
of a 1200 px JPEG in storytold/artcraft-launcher-assets, pinned to a tag
(SCREENSHOTS_REV) so a saved copy never goes stale. It only answers for
keys the page lists, so no address is ever built from anyone else's text.
MAX_SCREENSHOT_BYTES caps what is accepted, and looks_like_jpeg refuses a
login page or error page that arrives as a 200.
PdfCraft lists none yet: its screenshots still show its old name
(PrintCraft). ATTRIBUTION.md points to the pictures repository, which
carries each picture's own provenance.
Tests check that every app has exactly one page, that each is filled in,
and that addresses come only from a page's own pictures.
* Open an app's details in a drawer from its card
Clicking a card (or an app's name in the table) slides a drawer in from
the right over the dimmed grid. It shows the app's headline, version and
main button, links to its website, GitHub and releases, a gallery of
screenshots, its features, the facts and notes of its latest release, and
the same trademark notice as the grid. Esc, the close button or a click
on the grid closes it. The page's text is the hand-written copy in
core::details; the version, date, download size and notes come from the
release the launcher already fetches.
The pictures are fetched when a page is opened, not bundled. The UI sends
Command::FetchScreenshots and gets one Event::Screenshot per picture, the
lead one first. The desktop backend downloads them through the launcher's
Http (its User-Agent, a 2 MB cap, redirects followed) and keeps them under
<cache>/screenshots/<revision>/<app>/<key>.jpg, so later opens, and any
time offline, read from disk. Only a JPEG is kept: a login page or error
page that arrives as a 200 is refused. A missing picture doesn't hold the
others back, but once the network has failed the rest fail at once instead
of each waiting for its own timeout. The web build does the same with
fetch and leaves the caching to the browser.
Until a picture is here, a frame of the same size says "Loading pictures…"
or, when it failed, that it couldn't be loaded and that reopening the page
tries again; nothing moves when the picture arrives. The lead picture
opens full-window, the arrows over the pictures and the arrow keys flip
through them, wrapping around, and a scroll bar stays visible (egui's
floating one only shows under the pointer).
The drawer's labels are tl! strings, with 13 rows in each of ru, zh-CN,
de, fr and pt-BR; details.rs joins the sources the catalog coverage test
reads. Every size taken from the window is kept non-negative: the new
tests resize it from 40 x 30 to 4000 x 2400 with the drawer open.
actions() becomes actions_in(), which can leave the links out of the
... menu where the page already shows them as buttons. The image crate's
jpeg feature is enabled to decode the pictures (zune-jpeg and zune-core
join Cargo.lock). The dim and the enlarged picture's backdrop are colours
in theme.rs.
* Bring the details drawer up to date with main
- Website is always a button now (#198); drop the "Like {name}" line and the trademark notice,
which main removed with the product comparisons (93caf01).
- Bring back widgets::eyebrow for the drawer's section headings (#187 dropped its last caller).
- The drawer's ⋯ menu leaves out only the links, so Hide/Unhide (#185) stays.
- No slide when the OS asks for reduced motion (#201).
- Translate the screen readers' "{name} details" label, and add the drawer's strings to the
Spanish, Czech, Polish, Hungarian and Japanese catalogs.
- AGENTS.md: name the pictures' download origin.
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* Linux: draw on the desktop's GPU, not always the discrete one
egui asks wgpu for the high-performance adapter. On a hybrid laptop in
PRIME on-demand mode (Intel + NVIDIA, X11, GNOME) that is the NVIDIA GPU
while the display server runs on the Intel one, so every frame is copied
between the GPUs. During a resize the window then shows stale or garbled
frames and can stop drawing until it is resized again.
On Linux and the BSDs, ask for no power preference instead: wgpu then
takes the first adapter Vulkan lists, and Mesa's device-select layer lists
the display server's GPU first (or the one DRI_PRIME picks). Measured on
an RTX 4060 + Intel RPL-S laptop: in on-demand mode the list starts with
the Intel GPU, in `prime-select nvidia` mode with the NVIDIA one, so the
launcher draws on whichever GPU drives the screens. WGPU_POWER_PREF still
overrides it. Windows and macOS keep their current choice.
Refs #69
* Linux GPU: keep wgpu's adapter order, but a software rasterizer last
Without Mesa's device-select layer the Vulkan order follows the ICD files, and lavapipe can come
before the real driver.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Add a Polish interface
- pl.tsv: the full catalog (287 rows), written clean-room from the English
labels; plurals use the same one|few|many rule as Russian.
- Registered as the seventh language: "Polski" in the picker, locales pl,
pl-PL and pl_PL.UTF-8, complete_menus = true so the completeness tests
cover it.
- tests/polish.rs: the Polish UI test, modeled on the French one.
- README and AGENTS updated for the new language.
* Add Hungarian localization and fix responsive settings layout
* i18n: add Czech (cs) interface language
Czech catalog with every key of the most complete catalog (287 rows), the
cs language entry with locale detection (cs-CZ, cs_CZ.UTF-8, ?lang=cs),
a Czech plural rule (1 / 2-4 / other) with a unit test, and the README
language list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n: add Japanese (ja) interface language
ja.tsv translates the whole interface, the tray menu and tooltip, each app's
tagline and the backend's install/update messages into Japanese, and
Settings > Language offers 日本語. Japanese locales (ja, ja-JP, ja_JP.UTF-8…)
select it; Japanese has one plural form.
Japanese is a CJK language, so cjk_font.rs now looks for a Japanese system
font (Hiragino Sans on macOS; Yu Gothic, Meiryo or MS Gothic on Windows;
Noto Sans CJK JP, Noto Sans JP, IPAGothic, Takao or VL Gothic on Linux and
FreeBSD), falling back to a Chinese one, which draws kana and most kanji,
where the system has none. The font is loaded per language, so switching
between Japanese and Chinese reloads the right face.
A fallback face is nudged onto the body font's baseline: egui puts a fallback
glyph's baseline at that face's own ascent and only centres the two line
boxes, which left a face whose metrics differ from Inter's riding high. On
Windows, Yu Gothic's line gap is half an em, so its kana and kanji sat about
0.29 em above the Latin text beside them (Microsoft YaHei, used for Chinese,
happens to match). The nudge, applied through a FontTweak, is the difference
in ascent and line height, and helps the Chinese faces too.
The README and AGENTS list the new language. Tests: the complete-catalog
checks cover ja.tsv (placeholders, plural forms, every tl! literal, taglines,
package formats), plus a locale/plural test, a Japanese glyph-coverage test,
a baseline test and tests/japanese.rs (grid, table, dates, a backend message).
* i18n: make the CJK font choice language-neutral
The CJK comments, the missing-font note and the plural-rule docs named the
languages, so each new one meant editing them all; they now say "CJK". Each
language's writing system (its `script`) now lives in the registry, and a
language is drawn only with a face for its own script — never another's,
which would leave characters as tofu — so a language the system has no font
for stays in English, like any other. Adding Korean or Traditional Chinese is
a registry row and a candidate list, not a special case.
* Hide Craft apps the user is not interested in.
* Read the Linux desktop appearance portal so system theme matching works on Wayland.
* UI: follow the OS's animation setting
With Windows' "Animation effects" switched off, the launcher still glided each
notch of the mouse wheel over several frames and faded its dropdowns in. egui
does both by default, and nothing asked the OS.
The shell now says what the OS wants (Host::reduce_motion) and the UI asks it
about once a second while it draws, so a change shows without a restart. With
fewer animations asked for:
- a mouse wheel scrolls at once. egui has no switch for its wheel smoothing but
applies a trackpad's small steps as they come, so each wheel event is
rewritten as a run of those in the shell's raw_input_hook;
- dropdowns, tooltips and dialogs appear without a fade, and scrolling to a
place jumps there (style.animation_time, style.scroll_animation);
- the rule along the top of a hovered card appears instead of wiping in.
Progress bars and spinners keep moving: they say that work is going on.
Where the setting comes from:
- Windows: SPI_GETCLIENTAREAANIMATION, read as bit 33 of UserPreferencesMask in
the registry, as SystemParametersInfoW needs unsafe;
- macOS: NSWorkspace's accessibilityDisplayShouldReduceMotion;
- Linux and the BSDs: the XDG settings portal over D-Bus (its reduced-motion,
GNOME's enable-animations, KDE's AnimationDurationFactor), followed on a
thread so the UI never waits for D-Bus;
- web: prefers-reduced-motion.
Animations stay on wherever the setting can't be read. winreg and zbus become
direct dependencies of the desktop app; both were already in Cargo.lock.
The Windows path was run on a machine with the setting off. The macOS and Linux
paths were type-checked for those targets but never run, and the web one only
compiled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* UI tests: hovering a card follows the OS's animation setting
With animations off, moving the pointer onto a card leaves the window still
within 3 frames: its rule and its icon's rise show at once. With them on, the
0.3 s wipe takes 19 frames at 60 frames a second. Each test fails when the
card's timing is made to ignore the setting.
The test window no longer has a download in progress, so it is still unless
something animates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n(pl): use Polish plural rule, so 21 is "21 aplikacji"
plural_east_slavic treats 21, 31 and 101 as "one", which is right for
Russian but not Polish, where only 1 itself takes the singular.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* system theme: back off to every 5 minutes when there's no appearance portal
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Merge #189 and main into #201: one raw_input_hook, and test anchors on taglines
Main no longer shows the "Like …" product lines, so the tests find cards by their taglines.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n(cs): translate "Installed · latest {latest}" (added by #195)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n(pl): translate "Installed · latest {latest}" (added by #195)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n(hu): translate "Installed · latest {latest}" (added by #195)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n(ja): translate "Installed · latest {latest}" (added by #195)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* i18n: hide/unhide strings for the Spanish, Czech, Polish, Hungarian and Japanese catalogs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* hungarian test: count only the settings body's links, not the cards' Website links behind the modal
#198 gave every card a Website link, and Hungarian labels both Webhely.
* hungarian test: card links are 14px text, settings icons 24px; filter on that
---------
Co-authored-by: DzikStar <82441193+DzikStar@users.noreply.github.com>
Co-authored-by: ferivoq <ferikegameplay@gmail.com>
Co-authored-by: Losminek <301412321+Losminek@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: leonaci <leonacism@gmail.com>
Co-authored-by: satchmo91 <eherrmann1991@gmail.com>
Co-authored-by: Vikram Bhaskaran <vikram.bhaskaran@outlook.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* Read the Linux desktop appearance portal so system theme matching works on Wayland.
* system theme: back off to every 5 minutes when there's no appearance portal
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
wgpu's default DX12 compiler (Auto) loads whichever dxcompiler.dll the DLL
search path finds. The launcher ships none, and without a matching dxil.dll
the unsigned-DXIL warning fails wgpu's indirect-validation pipeline, the
device is lost and the launcher exits before the first frame. Pin the
compiler to FXC, which is part of Windows; WGPU_DX12_COMPILER still
overrides it. GL stays in the Windows backends: it is the deliberate
fallback for machines whose D3D12 adapters can't present to the window.
Fixes#126
Co-authored-by: Fabio Lussi <198260595+itrycodinganditsbad@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A launcher started from the Dock, the Start menu or at login has no
terminal, and the Windows build has no console, so its log records and
panic reports were lost.
env_logger now writes each record to standard error and to
logs/artcraft-launcher.log in the launcher's data folder (portable mode
included). Each start keeps the previous two runs' logs; the file is
capped at 16 MiB. The file is opened once this is the only launcher
running, so a second start doesn't rotate the first one's log away.
Panics are reported through the logger. The default filter is now info
for the launcher's crates and warn for the rest; RUST_LOG still wins.
Ported from storytold/pdfcraft#239.
Co-authored-by: Fabio Lussi <198260595+itrycodinganditsbad@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
tray-icon's default id is "<pid>-<n>" and the title is empty, so
StatusNotifierItem hosts showed numbers like "61965-1". On Linux and the
BSDs, give the icon the id "artcraft-launcher" and the title
"ArtCraft Launcher". Windows and macOS are unchanged.
Fixes#136
Co-authored-by: Fabio Lussi <198260595+itrycodinganditsbad@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>