Repository navigation
Backport the remaining quattro fixes, docs and translations to v4-5-0 - #17529
Merged
Merged
Conversation
…15382) * Keep the idle spawn test's exec logger off the vfork child's stack Qt 6.12 starts a QProcess child with clone(CLONE_VM | CLONE_VFORK) on a stack carved out of the parent's own, about 8KB above the parent's frame in clone(). The logger's execve wrapper built its line in an 8KB stack buffer, ran over the return address the parent resumes on, and the test shell segfaulted before it became ready. Build the line in pages mapped for the call instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Size the vfork child stack in the logger's comment by SIGSTKSZ Qt 6.12's forkfd gives the child `char childStack[SIGSTKSZ]`, and under _GNU_SOURCE glibc computes SIGSTKSZ at run time: 4 * the kernel's minimum signal stack, never below 8192. On a CPU with AVX-512 it is 13504, so the old 8KB buffer only overran the child stack on some machines; saying "about 8KB" hides why the test passes on one machine and crashes on another. Co-Authored-By: Codex Medium <noreply@openai.com> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit baa0670)
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit af21bdc)
) * Time the lock's first appearance once the scratch shell is quiet The lock service answers while the shell is still drawing its first frames of the wallpaper, bar, dock and notifications. A GPU draws them in milliseconds. Mesa's software renderer, which a machine without a GPU uses, takes a second or two, mostly in the buffer swap, and a lock preview timed meanwhile waits for those frames: in a GPU-less VM on GitHub's runners the fastest of three first appearances measured 587 to 1440 ms against budgets of about 1200 ms, and failed 10 of 18 runs. Wait until the shell uses at most one clock tick of CPU in 200 ms, for at most 10 s, before timing. The lock's own work starts only with the preview request, after the wait. In the same VMs the fastest first appearance is then 42 to 67 ms, on EPYC 7763, EPYC 9V74 and Xeon 8573C hosts, and 0 of 18 runs fail; the laptop the budget came from measured 74 ms. * Fail the lock test by name when the scratch shell dies while settling A shell that exited during the quiet wait read as quiet: its stat could not be read, every sample came back 0, and the test went on to time a preview that could not appear, failing one step later under the wrong name. A sample that cannot be read now prints the scratch log and fails as the shell having stopped. (cherry picked from commit 068a122)
The lock service answers once the shell has built its plugin services, QML work on the shell's main thread. Its budget scaled with process_speed_percent, the cost of 100 fork/execs, which does not track that work: in KVM guests on GitHub's runners, Xeon 8370C and 8573C hosts start processes 25 to 40% faster than EPYC hosts but build the shell no faster, so their budget came out at about 2100ms against a build of 2100 to 2300ms, and the check failed about half the time there. A bash loop and quickshell --version tracked the build no better. The time until the shell first answers, once Quickshell has loaded its config and before the plugin services, is the same kind of work on the same thread. The view build took 2.6 to 3.2 times it on EPYC 7763, 9V45 and 9V74 and Xeon 8370C, 8573C and 6973P-C hosts, so the budget is now four times the fastest of those answers. Across 47 runs on 12 runners the build used 66 to 78% of that budget on every CPU, and a lock service that spends an extra second building failed it 48 times out of 48. A slower start before the first answer moves both and is left to bar-startup-budget-test.sh. (cherry picked from commit fc91b6a)
The leak checks counted /proc/$$/fd with find inside a command substitution. Bash makes the substitution's pipe, forks, and closes the pipe's write end in the parent only afterwards (subst.c:7208, 7251 and 7425 in bash 5.3), so a find that runs before that close counts the write end as well. In full-suite runs the test then failed one of its "leaked FD" checks now and then while the same five descriptors were open before and after: 2 of 9 runs in VMs, and one of three in CI. Count with a glob of /proc/$$/fd in the test's own shell instead. In a VM, 20000 counts the old way read 6 instead of 5 eleven times, and two more under load; the glob read 5 in all 40000. The test passed 20 of 20 runs and still failed on a descriptor deliberately left open. (cherry picked from commit dff02d3)
Under pipefail, "printf ... | grep -q PATTERN" can fail with the pattern present: grep -q exits at its first match, printf's next write gets EPIPE, and pipefail turns that into a failed check. In the VM test run of #16835, privileged-heredoc failed "a shell special parameter cannot hide a later baked $HOME path" that way, with the expected text printed in its own failure message after "printf: Broken pipe". Check against a here-string instead, which bash writes before grep starts, in privileged-heredoc and preinstalls; luks-rekey-journal captures cryptsetup's output first and still returns 1 when it fails. On one shared core, 11 of 20000 piped checks failed with the text present and 0 of 40000 here-string ones did. Under load in a VM, privileged-heredoc and preinstalls passed 10 of 10 each, and luks-rekey-journal passed 6 of 6 runs. (cherry picked from commit d2b6669)
(cherry picked from commit 4872b40)
* docs(hotkeys): identify physical resize keys * test(hotkeys): pin physical resize bindings * docs(hotkeys): clarify physical resize keys on JIS and Spanish * test(hotkeys): remove brittle assertion from documentation-only change (cherry picked from commit 62161e0)
The record of which migrations have run lives under the home directory, which a restore leaves alone, and a migration runs only when its marker is missing. After a restore the migrations of the rolled-back update stay marked as done while their changes to the root filesystem are gone, and the next update skips them. Say so in the System snapshots chapter, next to the note about ~/.config. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 45ca581)
(cherry picked from commit 2861f26)
Co-authored-by: Kbujar <krzysztof.bujar@marcel.pl> (cherry picked from commit 1d6b470)
* Document ASUS Fn+Super Windows-key lock in troubleshooting Easy-to-miss firmware chord disables Super with no UI feedback, so Omarchy binds look broken. Point users at Fn+Super before digging into Hypr config. * Match the manual's key notation and scope the Super lock entry to what is known The manual writes chords as inline code, so Fn + Super does too. The lock was reported on a TUF (#5078) as well as documented for ROG models, so the entry names both. Fn + Esc is Fn-lock only on some ASUS models, so the entry describes Fn-lock instead of naming a chord that may be wrong for the reader's laptop. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> * Tell readers to undo Fn + Super when it was not the lock The chord toggles, so pressing it on a laptop whose Super binds broke for another reason turns the lock on and leaves Super dead after the real problem is fixed. Raised by Greptile in review. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit 633d87f)
…ted fill (#15818) textPreformat paired cyan with the bare editor background, so inline code in the Markdown editor, hovers and chat was cyan on nothing; on a light palette that falls to 2.6:1. Every theme installed from git gets this generated theme, since it cannot ship vscode.json. Foreground on muted at 25% is the template's own fill idiom (keybindingLabel already pairs them) and the shape VS Code's default themes use, and it clears 6:1 on every shipped palette. Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit a4b086b)
Replace the roadster starlight background in ethereal (cherry picked from commit 573b518)
* Keep the still on screen until OWE takes a video background A switch from a still to a video cleared the shell's still at once. The background layer turned transparent before OWE had a frame, so the desktop showed an old frame of OWE's parked renderer or an empty background. OWE then drew the still again and faded it into the video, so the switch flickered before the fade. The layer now keeps the current still when the next background is a video, until OWE suspends the layer after the video's first frame. OWE fades from the same still, so the switch is one fade. A switch with a suspended layer, an instant switch, and a switch from a video work as before. test/shell.d/background-video-keeps-still-test.sh checks both cases. * Let a reveal finish when a video is chosen during it A video chosen during a reveal between 2 stills switched the base image to the incoming still and cleared the reveal's images at once. The base image loads on its own, so the desktop could show a blank frame or an old frame of OWE's renderer. The video selection now leaves a running or pending reveal alone. The reveal finishes on its incoming still, and the reveal's images cover the desktop until the base image is ready. That still is the one OWE fades from. At the end of the reveal, a video as the current background no longer replaces the still. The test moves into background-pending-theme-test.sh and reuses its setup and Quickshell launch. It chooses a video while a reveal is pending and checks that the stills stay on screen and that the reveal ends on its still. (cherry picked from commit 432ae66)
The dropdown test walks a long list with Down presses and asserts that the selection reaches the last row. Each delegate sets currentIndex on a pointer move. The walk scrolls rows under a stationary pointer, so a row that scrolls under the pointer takes the selection back mid-walk. Two CI runs stopped at 5 and 9 of 11. The test moved the pointer to (500, 600) only in cleanup. That coordinate misses the popup by accident, and it does not run before the long-list walk. Park the pointer at (0, 0) before the walk instead. That point is always inside the window and off the popup. Assert currentIndex after every press instead of a fixed wait(50). A stalled walk then reports the step that stalls. With the pointer left over the list, the old test failed all 9 long-list cases on 12 of 12 runs. The fixed test passed 40 of 40 runs with the pointer over the list, and 80 of 80 runs through test/shell.d/dropdown-test.sh. (cherry picked from commit bda3aaf)
lspci names Wildcat Lake's GPU only "[Intel Graphics]", which the video-acceleration check did not match, so intel-media-driver was never installed and video decoded on the CPU. Match that generic name, and install the driver on existing machines it missed. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit cec1d26)
* Document recovery from broken session environment files * Check UWSM environment syntax with sh * Drop the claim that a .bak suffix does not disable an env.d file UWSM 0.27.0 (upstream commit 4cfc107b, fixing Vladimir-csp/uwsm#226) skips *.bak, *.disabled, *.broken and similar names in env.d, and checks each file with sh -n before sourcing it, skipping one with a syntax error. Errors that only happen at run time, like ${PATH:}, still pass sh -n and still stop the desktop from starting, so the recovery steps stand; moving the file out of env.d works on every version. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Read the preloader's journal by identifier, and soften the fatal-error claim journalctl --grep matches only the message, and UWSM's "Env output mark ... not found" line contains neither "uwsm" nor "preloader", so the old filter hid it; -t uwsm_env-preloader shows every line the preloader wrote. Not every run-time error aborts sourcing, since UWSM does not set -e, so only a fatal one is said to block login. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> --------- Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit 5ef4527)
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit dfdf0b0)
* Let the Lua assertions in shell tests fail Handing a heredoc to lua with no script argument runs it as a REPL. An error prints a traceback and the interpreter still exits 0, so eight assertions across two files could not fail their tests: five checking that the Hyprland bootstrap clears cached modules, three checking that saved workspace layouts load. They have been passing on their own merits, which is luck rather than coverage. Read the heredocs as scripts with an explicit -, and check the exit status where the Lua assertions are the only assertions. base-test.sh does not set -e, so a nonzero exit alone would still have left the pass below to run. The four remaining call sites hand their output back for the shell to grep, so their failures already bit. Switched anyway, so the next assertion written in one is not born dead. Added a style test for it, since nothing about the working form advertises why it matters. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Read the Lua heredocs quattro gained since as scripts too hw-nvidia-test, screensaver-workspace-test and shell-shortcuts-test each landed on quattro with a bare `lua <<'LUA'`, which the new style test refuses. All three hand their output back to the shell, so their failures already bit; switching them keeps the guard green and the next assertion written inside one armed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Fail the binding checks that only assert absence when Lua errors These three read the bindings Lua prints and pass when a binding is missing, so a Lua error that prints nothing passed them too. Now that the heredoc runs with lua -, its exit status says whether the config loaded. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Say why a bare Lua heredoc cannot fail, and catch it without a space lua with no script argument is not a REPL when stdin is not a terminal: it runs stdin with dofile and discards the error status, so it exits 0. The guard also missed lua<<'LUA', which Bash accepts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> * Run quattro's new Lua assertion blocks as scripts hyprland-startup-cursor-test.sh and dictation-test.sh arrived on quattro with three bare lua heredocs made only of assertions, so none of them could fail, and shell-test-style-test.sh fails on them. Each now runs as lua - and fails its test on Lua's exit status. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Arm the Lua heredocs quattro added in the hold-binding and JIS tests bind-hold-test.sh is made only of Lua assertions, so as a bare heredoc none of them could fail; it now runs as lua - and fails on Lua's exit status. hyprland-jis-bindings-test.sh hands its output back for comparison, so it switches to lua - to keep the next assertion written there from being born dead, and to satisfy shell-test-style-test.sh. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Name the JIS case whose Lua failed Now that fire() exits nonzero on a Lua error, set -e would end the test at the command substitution with no not-ok line; failing there names the binding that broke. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Fail the uninstalled-backend and keyboard layout checks when their Lua errors The uninstalled dictation backend check only asserts that no dictation binding is printed, so a Lua error printing nothing passed it, as the absence checks armed earlier did. The keyboard layout resolver's captures discarded its status too, so an error after the expected output still compared equal; both now fail the way the JIS captures do. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> * Move the Lua heredoc guard into the test runner's test Omarchy's AGENTS.md asks for a new check to extend the area's existing test file rather than add a *-test.sh, and test-runner-test.sh is where the suite's own health is tested. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Omabot <david@hey.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit 8422598)
* Document quiet Limine menu for dual boot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Say how to reach Limine when Direct Boot is on With Setup > Direct Boot the firmware launches the Omarchy UKI without stopping at Limine, so pressing a key never shows the menu. Point to the firmware boot menu or turning Direct Boot off, as the snapshots guide does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit ec7d472)
(cherry picked from commit 8337e23)
* Say what the security manual's defaults actually are The Security chapter called full-disk encryption mandatory, while Getting started documents opting out at the formatting confirmation, and said all incoming traffic is blocked except LocalSend, while the firewall also admits DNS from Docker networks and UFW's built-in multicast DNS rule answers for Avahi. State the opt-out and the three exceptions. Translations are left to the catalog sync. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Don't call the firewall's exceptions a complete list UFW's stock before.rules also accept ping, DHCP replies and UPnP discovery ahead of Omarchy's own rules, so "three exceptions" told a reader everything else was blocked. The Docker DNS rule also admits all of 172.16.0.0/12 and 192.168.0.0/16 rather than only the host's containers, so it is described by the address it answers on and the ranges it accepts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit 7a0e3ea)
(cherry picked from commit 6a3815b)
* Fix additional assorted typos in navigation * Call it the file manager, as the rest of the manual does The app is Files, but the hotkeys table and the GUIs page both call what Super + Shift + F opens the file manager. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Put the blank line back before the four-way tiling image Resolving the conflict with quattro moved the blank line above the "Now try" paragraph and left the image with none before it, so Markdown folds it into that paragraph instead of giving it a block of its own like every other image on the page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit cdca799)
omarchy-channel-set runs omarchy-update -y to skip the start confirmation, but -y also set OMARCHY_UPDATE_UNATTENDED=1, so the required reboot after switching channels printed "Run omarchy-system reboot when ready." instead of offering a gum confirm. -y now respects an explicit OMARCHY_UPDATE_UNATTENDED, and the channel switch passes 0 since the user is still at the terminal. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 5fa15f2)
* Open OpenCode 2 with the Omarchy theme OpenCode 2 reads its TUI theme from cli.json and ignores the theme key in opencode.json, so fresh installs started on the stock theme. Ship cli.json with the system theme and add it for existing users who haven't picked one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the OpenCode 1 settings from opencode.json OpenCode 2 ignores the theme key there, and spells the update policy update: "disable". Keep only that, since 2.x still reads it from opencode.json and would otherwise nag about updates mise owns. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 73ae9cd)
* Let WirePlumber find the cameras at login * Clean up the camera test's background fixture on exit * Wait for the camera at login only on Snapdragon * Time the camera wait from before the grant is scheduled The background chmod's one-second timer started before wait_for_camera took its start time, so on a loaded runner a slow command substitution shortened the measured wait below the 900 ms floor with correct code. Taking the start before forking the grant makes the floor hold however late the helper starts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> Co-Authored-By: muse-wonder via pi, thinking high <noreply@meta.com> --------- Co-authored-by: Omarchybot <317366263+omarchybot@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Codex Medium <noreply@openai.com> Co-authored-by: muse-wonder via pi, thinking high <noreply@meta.com> (cherry picked from commit 6f0e6fb)
(cherry picked from commit 09fa23e)
* Reload Hyprland resiliently from outside the session Running omarchy update from a remotely started herdr left the Superwhisper migration failing silently: the shell had no HYPRLAND_INSTANCE_SIGNATURE, so `hyprctl reload >/dev/null` exited 1 with its error discarded on stdout, and the update stopped with nothing to show for why. omarchy-restart-hyprctl now finds the newest live Hyprland instance when the signature is missing or stale, does nothing when no Hyprland is running, and reports a real reload failure on stderr. The dictation installers and removers reload through it and no longer fail once their config is saved. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Simplify omarchy-restart-hyprctl to hyprctl --instance 0 outside the session hyprctl already finds the running instance itself, with no signature and no XDG_RUNTIME_DIR, so the instance discovery and stale-signature handling go. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Rename to omarchy-restart-hyprland and fix theme-set budget test The budget test runs theme-set without a Hyprland signature, so the helper took the --instance 0 path the test didn't expect. Drop the repeated caller comments. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Test that a failed Hyprland reload doesn't fail Superwhisper setup Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Reload Hyprland through omarchy-restart-hyprctl everywhere Go back to the original name, route every bin/ reload through it, and let the helper absorb a missing Hyprland so callers need no || true. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 5c0d91a)
#7460's runner check rejects a bare lua heredoc, whose errors exit 0. quattro dropped this test with the native screensaver, so only 4.5 still has it.
Run outside a session, the helper reaches Hyprland with --instance 0, which these tests' hyprctl stubs didn't expect. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 857afe1)
Co-authored-by: Ahmad <8452482+ahmadtv@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> (cherry picked from commit 265ce5d)
Translation syncs only run on quattro, so 4.5's catalogs missed strings its backports added (Amp's menu labels, the download toast's Cut) and kept stale ones (the dictation hint still named F9). This copies 686 translations, 392 of them manual chapters, only where 4.5's English is identical to quattro's and quattro's manifest records the translation as current.
Contributor
[High impact] The reviewed changes appear safe to merge. SummaryBackports the remaining fixes, documentation, tests, and translations to
Known failures acknowledged by
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Backports the quattro fixes #17386 left out, plus the docs, test fixes and translations 4.5 never got. Every commit is a
cherry-pick -xof its quattro commit, in the order it landed there, except two noted below.Fixes
HYPRLAND_INSTANCE_SIGNATURE,hyprctl reloadfailed and the update stopped with no reason shown.latest(2.2.5 today), which reads its theme fromcli.json, so 4.5 users get the stock theme.Background.qmlcode it guards is on 4.5 as is.Docs and backgrounds
Test fixes
#15382, #16896, #16952, #16953, #16959, #7460, #17117, #17455
#7460 adds a runner check that fails any shell test feeding Lua a bare heredoc, since
luawithout-exits 0 on errors. quattro deletedscreensaver-workspace-test.shalong with the old screensaver, so 4.5 still had one there. The commit after it switches that test tolua -.Translations
Translation syncs only run on quattro, so 4.5's catalogs missed what its backports added (Amp's menu labels, the download toast's Cut) and kept stale text: German's dictation hint still says Right Alt or F9 and Super + Ctrl + X, from before #17188.
The last commit copies 686 translations from quattro (196 CLI strings, 98 menu labels, 392 manual chapters). It only takes a unit when 4.5's English is identical to quattro's and quattro's manifest records the translation as current, and it records the same hash. Pending units drop from 931 to 245; the rest have English that differs from quattro's, so they wait for a real sync.
Adapted picks
omarchy-hyprland-monitor-internal-mirrorit keeps 4.5's single-display mirror and swaps itshyprctl reloadforomarchy-restart-hyprctl.-m 1.Testing
./test/allon this branch: CLI passes, and 412 of 415 shell test files pass. The three that fail (background-intro-lifecycle-test.sh,launch-about-test.sh,runtime-smoke-test.sh) fail on plainv4-5-0with the same assertions.I18N_STRUCTURAL=1 bash test/shell.d/i18n-test.shandscripts/i18n-sync.py extract-scan --check, the two checksi18n-checkruns, both pass.test-runner-test.shfails with Let the Lua assertions in shell tests fail #7460 alone and passes once the screensaver test useslua -.Scope and risk
Two migrations come along: the Wildcat Lake driver install (#17150) and OpenCode's
cli.json(#17293). Each keeps its quattro filename, so a machine that later moves to quattro won't run it again.Still held for later: rtkit (#16824), prebuilt NVIDIA modules (#17409), the Hyprland 0.57 fixes (#17103, #17152), the Zenbook A16 and Surface Pro 11 support, and the feature work.