Repository navigation
Merge upstream quattro (notifications, Appearance, aarch64 stable and rc) - #89
Merged
Merged
Conversation
The shortcut repair in 1786643346 walked every Chromium-family profile root except Brave Origin's, which Omarchy has installed since before the extension id was pinned. Those profiles keep Alt+Shift+L bound to a ghost id, so Copy URL works from the toolbar but not the shortcut. Add the Brave Origin roots to the repair and rerun it through a new migration, following the same pattern as the T2 defaults rerun. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
- 🐛 re-key this system's limine entry in place instead of rewriting limine.conf from the template, which dropped Windows and other Linux entries on a shared ESP - 🐛 remove only the replaced identity's machine-id directory, not every one limine.conf mentions - 🔒 check hashes only on the rebuilt entry, since other systems' entries now survive and are not ours to judge Fixes omacom#13689 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019inPHb7J2FGA6MZc8kqedz Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
- ✅ prove the reset keeps other systems' entries and directories, keeps this entry first, and collapses duplicates - ✅ prove hash verification ignores other systems' entries but still rejects a stale or missing rebuilt UKI Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019inPHb7J2FGA6MZc8kqedz
- 🐛 stop treating other systems' machine-ids as stale, which made first boot rewrite limine.conf from the template and wipe the dual-boot entries the reset had just kept - 🐛 rebuild only this machine's own entry in place on the LUKS re-key and repair paths, keeping every other entry and machine-id directory Fixes omacom#13689 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019inPHb7J2FGA6MZc8kqedz
- ✅ prove other systems' machine-ids no longer trigger a first-boot rebuild, and a machine without its own entry still gets one - ✅ prove a first-boot rebuild keeps other systems' entries and directories and keeps this entry first Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019inPHb7J2FGA6MZc8kqedz
1786643346 keeps the backup it writes beside every profile it repairs, and reads any backup as a repair still to verify. Rerun across every root, the follow-up asked everyone whose Chromium, Chrome or Brave profile the first pass had already fixed to close that browser again, and failed the migration queue when they declined. The rerun now sets profile_roots to the Brave Origin roots before sourcing, and the original keeps its full list when nothing set one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com>
Users with T3 Code had the app opened in the middle of omarchy update. The migration that activates the Omarchy theme for existing installs re-ran the whole installer, and the installer finishes by launching the app and announcing that T3 Code has been installed. Run only the installer's theme steps in the migration: render the palette when the current theme predates the template, publish it, and select it with the T3 CLI, honouring T3CODE_HOME the same way. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012cEGKnoy56LPPiFKucTexU
- 🐛 the staged `limine-update` overwrites the running system's UKI and the re-key empties its entry, so a failure before the hashes verify left that system unbootable - 🐛 a retry then treated the half-staged entry as another system's and appended a new one after it, moving `default_entry` onto another system - 🐛 the previous identity's snapshot UKIs were deleted before anything verified; they now go only once the rebuilt entry does - ✅ cover a failed staging, the retry after it, and a verified commit Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SdXqTwm63rRMwotZ1xqLC8 Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
- 🐛 a failed first-boot rebuild leaves this machine's bare entry header, which counted as an entry, so later boots never repaired it - ✅ cover the bare-header case in the first-boot staleness check Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SdXqTwm63rRMwotZ1xqLC8
- 🐛 until the factory root replaces `@`, the running root is still what boots; a failed ESP unmount or rename left it with the factory UKI, which may not match its kernel - ♻️ mount the ESP outside the staged root and bind it into the chroot, so the cleanup trap can still restore it after the chroot bind is gone - ✅ check that the commit runs only after `swap_done=1`, and stub it in the accounts harness Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
- ✅ awk's `END` ran after the early `exit` and replaced its failure, so a commit before `swap_done=1` still passed Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU
- 🐛 errexit still applies in the EXIT trap, so a failing log message stopped the cleanup before it unmounted and deleted the staged root - 🐛 after the root switch the reset is staged; a failed ESP unmount or old-directory removal no longer reports it as failed and skips the reboot prompt - ✅ run the failed-staging case with a failing log Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU
- 🐛 if `@` was renamed away but the factory root could not take its place, cleanup deleted the staged root and left the machine with no `@` to boot - 🐛 the running root now goes back to `@`; if that fails too, cleanup keeps the staged root so both systems survive for recovery - ✅ run the accounts harness under the production cleanup trap and cover both failures Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
- 🐛 SIGINT or SIGTERM between the root switch and `commit_next_boot` ran the cleanup with the restore still armed, putting the old kernel back under the factory root - 🐛 cleanup now restores only while the staged root still stands apart, which holds up to a failed or rolled-back switch and stops holding the moment the rename lands - ✅ cover a SIGTERM right after the switch through the production trap Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU
- 🐛 a failed staging left UKIs and loaders the rebuild added, such as one for a kernel only the factory snapshot has; on a nearly full ESP they could stop the originals from being copied back - ♻️ share the backed-up path list between backup and restore - ✅ the failed-staging check now sees an added UKI and `EFI/BOOT` fallback removed Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018LCZwBtEFSXmRbSYsbxVbU
Each frame of the marquee redraws the bar, and it kept running while the track was paused and while the bar was hidden. Stopped, the title reads from its start again. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
omarchy-remove-service-1password shipped with only the install row (d7adaee), and Remove > Services was built from Dropbox and Tailscale alone (ee97329). A test now requires a Remove > Services row for every installable service that ships a removal script. Fixes omacom#15485. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011K1peyrh5MdZ6ttHkVq8Gh
…croll speed Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
The fix is one menu row; a standalone test for a data omission adds more code than it guards.
1788604821 dates from September and sorts ahead of 88 migrations quattro has since gained, so a late updater who declines its browser prompt would hold all of them pending. Numbered last, a decline holds nothing else. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The rerun runs on every machine after 1786643346, so listing Brave Origin in the shipped migration too only repaired it twice for late updaters, and a Brave Origin window opened between the two would be asked to close a second time. The shipped migration keeps just the hook that lets the rerun narrow its roots. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1aa1423 retired the full test once 1786643346 shipped and froze, and the cases restored here re-tested that frozen behaviour. What this change adds is the rerun, so the test keeps the two things it must do: reach a Brave Origin profile, and not ask again about a profile the first repair already fixed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e Maps back Super + Shift + S (omacom#16048) Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
) * Move Appearance and Style to the top of Setup Trigger > Appearance becomes Setup > Appearance and the root Style menu becomes Setup > Style right below it. Translation catalog keys and the manifest are renamed in place so no translation goes stale, and `omarchy menu summon style` keeps working through an alias. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Gather the window looks under Setup > Appearance > Windows Window Gaps, Window Corners, Window Opacity and 1-Window Ratio move into a Windows submenu right below Dock. Translation keys and the manifest are renamed in place so existing translations stay current. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Make Appearance a top-level menu holding the Style options Appearance moves back to the root, right before Trigger, and takes in Theme, Background, Unlock, Font, Hyprland, Screensaver and About from Style, which goes away. `omarchy menu summon style` now routes to Appearance. Translation keys and the manifest are renamed in place. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop Appearance > Hyprland, gather Screensaver and About under Branding Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Use the Omarchy logo for Appearance > Branding Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Only show Appearance > Dock on a floating workspace The dock only appears on floating workspaces, so its settings are hidden everywhere else. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Reorder Appearance: Font and Menu Bar up, Unlock down Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the CJK glyph picker from Appearance > Font Language and input-method setup already pick the forms, and `omarchy font cjk` stays for choosing them by hand. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Reorder Appearance and drop Herdr Theme Sync from it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Name each menu toggle by the state it is in Every toggle becomes a pair of rows, one shown while the feature is on and one while it is off, both flipping it: Animations On / Animations Off, Notifications On / Notifications Silenced, Hybrid GPU Mode / Integrated GPU Mode, and so on. The three readers now shared by a pair join GUARD_READERS so the batch runs each once. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Label menu toggles by what picking them does The row shown while animations are on now reads Animations Off, and so on for every toggle; switches that read better as verbs say so (Show Battery Percentage, Silence Notifications, Switch to Integrated GPU, Disable Touchpad). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
SUPER + SHIFT is for apps and SUPER + CTRL for panels and menus, so the calculator moves from SUPER + CTRL + Q to SUPER + SHIFT + Q and btop from SUPER + CTRL + T to SUPER + SHIFT + T. Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…o N (omacom#16072) * Add Super + Ctrl + T for themes and W for backgrounds, move Network to N Nightlight loses its binding and stays in the toggle menu, which frees N for Network and W for the background switcher. The existing Space binds for background and theme stay. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Update the manual for the new theme, background, network and app hotkeys Also catches up on Calculator and Activity, which moved to Super + Shift + Q and T in omacom#16071 without the manual following. Translations are left for the next translation refresh, as with earlier hotkey changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the unbound nightlight description from the English keybindings catalog Regenerated with scripts/i18n-sync.py extract-apply, which keeps the catalog equal to the descriptions the Hyprland bindings declare. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…15522) * Let install media preset the language, keyboard and timezone A China ISO should install in Chinese with the Pinyin keyboard row and the China timezone chosen, while every choice stays the user's to change. The setup form, shared by the ISO configurator and first-boot owner setup, now preselects OMARCHY_FORM_KEYBOARD_DEFAULT and OMARCHY_FORM_TIMEZONE_DEFAULT when the media sets them, and English (US) and the network's guess otherwise. The language travels as data: the installer writes a registry code to /etc/omarchy/language, system finalization generates its locale as root, and user finalization makes it the user's desktop language through `omarchy-language seed`, which writes the language and session LANG without elevating, notifying, or seeding an input method over the keyboard row the user picked. The system locale stays English, since the console cannot draw every script. An installer language picker can write the same marker. Provisioning also names the Downloads, Pictures, Videos, Documents and Music folders in English up front, so a first login in another language does not create translated folders beside the ones Omarchy made and bookmarked; the Omarchy 4 upgrade does the same where the user has not named them. Fixes omacom#15509. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep a user's own language and folders, and carry install media's choices to first-boot setup Seeding a preset language now does nothing when the user already has one, so re-running provisioning after a change under Setup > Language keeps that choice, and it writes the session file without touching a session bus. Folder names are pinned only where the user has not named the folder, in any language, the same rule the Omarchy 4 upgrade follows. Install media for one market can leave its keyboard and timezone choices in /etc/omarchy/form-defaults, read as data, so a first-boot owner setup preselects them as the live installer does; the live installer's own environment still wins. Co-Authored-By: Codex Medium <noreply@openai.com> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Settle a deferred install's region against the owner's timezone, and create the folders it names A deferred install from one market's media takes the media's region before anyone has chosen a timezone. First-boot owner setup now settles it once the owner picks one, when the media left that region as REGION in /etc/omarchy/form-defaults: a timezone a region profile lists picks that region, Hong Kong, Macau and Taiwan are always World, any other timezone keeps the media's region, and a change re-renders pacman for the installed channel. The English folder names are created as well as recorded, since xdg-user-dirs resets a named folder that does not exist back to the home directory. Co-Authored-By: Codex Medium <noreply@openai.com> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Escape commas in the keyboard preselection gum splits --selected on commas, so a default like "Chinese (Simplified, Pinyin)" or "English (US, Dvorak)" never matched and the list stayed on English (US). --------- 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>
* Offer stable and rc on aarch64 Every aarch64 install ended up on edge, whatever ISO it came from: default/pacman/aarch64/ held edge templates only, and omarchy_pacman_default_channel returned edge on every ARM platform, so install finalization fell back to edge for an rc or stable install. 4.5 is the first release whose omarchy and omarchy-settings run on aarch64, and pkgs.omarchy.org/rc/aarch64 serves 4.5.0rc1. The rc and stable templates are the edge ones with the channel's Omarchy repository and Arch Linux ARM mirror: arm-rc-mirror and arm-stable-mirror.omarchy.org, the snapshots promoted from arm-mirror.omarchy.org, as x86_64's rc-mirror and stable-mirror are. Every aarch64 platform but Apple Silicon now defaults to stable; Apple Silicon keeps edge templates alone, since its packages and Asahi ALARM are published for edge only. omarchy-version-channel recognised only the x86_64 mirror hosts, so an ARM machine already reported "unknown / edge", and on rc or stable the menu's checkmark and a refresh with no channel named would not have known the channel. It now accepts the arm- hosts too, and the templates test checks that it names every platform's channel. * Stop listing Arch Linux ARM's [aur] on aarch64 Arch Linux ARM's [aur] holds 12 packages, none of which an Omarchy install or package list names, and on Oct 3 we decided not to carry it. The aarch64 edge, rc and stable templates now list [omarchy], [core], [extra] and [alarm] only, so new installs and channel switches stop reading it. Machines installed before keep [aur] in their /etc/pacman.conf, since an update never rewrites that file, so Omarchy's mirror keeps serving it. Apple Silicon's template keeps [aur]: the Mac image builder in omarchy-mac-installer requires it in the installed system.
…ev (omacom#15903) On edge, omarchy-channel-set stable first runs omarchy-refresh-pacman stable, whose omarchy-update-pacman -Syyuu downgrades to whatever the stable repositories list, and only then installs omarchy and omarchy-settings. Until 2026-10-18 23:06 UTC stable's omarchy.db still listed omarchy-dev and omarchy-settings-dev 4.0.0.r1832 from August, so -uu moved an edge machine from r7052 to r1832. That build predates command-scoped sudo: the wrapper check after the refresh stopped the switch before the swap and printed "Packages are switched. Run 'omarchy update'". They were not. The machine was left on omarchy-dev r1832 with stable repositories, omarchy update found nothing to do, and r1832's shell has no theme colours on Qt 6.12, so the bar is invisible (omacom#15881, omacom#15879). Stable and rc publish omarchy and omarchy-settings; the -dev pair is edge's. The refresh now passes --ignore omarchy-dev,omarchy-settings-dev for stable and rc, so the swap is the only transaction that touches the pair. pacman warns that it ignores the downgrade, keeps r7052 with its wrapper, and the swap replaces it with the release. The refresh then treats the -dev pair exactly as it does now that the stale build is gone from the repository (pacman still checks the pair's dependencies), so the ignore adds no failure of its own. Doing the swap first instead would put a tree without the wrapper in place before the downgrade-capable refresh, which could then no longer run and would leave edge's newer Arch packages under stable repositories. Edge refreshes are unchanged. The stop now asks pacman which pair is installed (omarchy-dev provides omarchy, so pacman -Q omarchy omarchy-settings answers for either) and says so: with the destination's packages in, it names them and says to run omarchy update; when a transaction before the swap replaced the tree, it names what is installed and says to run the channel switch again, which an r1832 tree completes on its own. In a VM (a 4.0.4 install moved to edge r7052, LUKS, Limine), against the live stable repository while it still listed r1832: unchanged, channel set stable left omarchy-dev r1832 (channel-current edge, version-channel stable), omarchy update did nothing, and after a reboot the shell logged 221 "of undefined" warnings with an empty bar. With this change the same switch installed omarchy 4.0.5, omarchy update finished, and after a reboot the bar was back with no failed units, coredumps or pending migrations. Edge to rc landed on 4.0.6rc1, edge to edge and rc or stable back to edge still work, and leaving dev against a copy of the stale database no longer installs r1832 on the way.
The shared-ESP import's intermediate UKI-only check had replaced the check that already covers kernel and initramfs entries and skips limine_history. USB and Thunderbolt trust are still scrubbed from the factory baseline. Co-authored-by: Mark Groves <mark-groves@users.noreply.github.com>
|
Important Review skippedBot user detected. To trigger a single review, invoke the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
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.
Upstream
quattromoved from81fd2d23(already on the fork via #88) toce39c5e9(2026-10-20 21:45 UTC). 115 first-parent commits, 100 of them non-merges, cherry-picked onto forkquattro. The 15 merge commits were not replayed; their changes are in the commits they merged. Cursor stays on the official tarball, not a mise tool. Face, Grok Bot, and Cursor menu entries stay. The power-profile menu provider stays out. Factory reset still scrubs inherited USBGuard and Thunderbolt trust.Notifications are a new engine (omacom#15932)
The shell notification daemon is replaced in place. Cards stack by source (per site for a browser notification, per app otherwise), a hover opens the stack, and one clock moves every card. Do Not Disturb, snooze, and a history panel open from the bar indicator. History stays in the shell. A content update restarts the countdown. The previous engine's Do Not Disturb is read until the new quiet state is saved, so there is no migration.
Site icons are never fetched. The helper sandbox, Pillow, inline reply, and the KDE Connect helper are gone. Icons resolve in the shell, sender images load directly, and KDE Connect is an ordinary sender. The deck is one fixed layer per output and stays mapped while idle so Hyprland does not re-evaluate focus; it unmaps over a fullscreen window when nothing is showing.
Elsewhen can name real places and the people in them (omacom#16027)
A city can be a real place, "here" can be set, and people in a city can be named. The clock moves when here is on another zone. Place, weather, and profile reads are bounded, and a person added while another fetch is finishing is still fetched. The globe jump goes to the town that was picked.
Node is a lazy mise tool, and upgrade pruning is package-owned (omacom#15876)
Setup no longer installs Node. It is declared in
etc/mise/conf.d/omarchy-tools.tomland built on first use. Migration1792343099retires the setup-time install.upgrade.auto_prune = falseships inetc/mise/conf.d/omarchy.toml, so provisioning no longer writes that setting into the user's mise config. Migration1792343225drops the setting from a config Omarchy wrote. Thecursor-agenttool line stays out of the tools list. The default-agent check expects 11auto_updatetools and 21 lazy tools.Appearance is a top-level menu (omacom#15973)
Appearance sits at the root, immediately before Trigger, and holds what Style used to: theme, background, unlock, font, Hyprland, screensaver, and About.
omarchy menu summon styleroutes to Appearance. Window gaps, corners, opacity, and the one-window ratio sit under Appearance > Windows. Calculator moves to Super + Shift + Q and Activity (btop) to Super + Shift + T (omacom#16071). Super + Ctrl + T opens themes and Super + Ctrl + W opens backgrounds; Network takes N, and Nightlight keeps its menu row without a binding (omacom#16072). The keyboard screenshot is Super + Ctrl + Alt + C, and Google Maps has Super + Shift + S back (omacom#16048).Install media can preset language, keyboard, and timezone (omacom#15522)
The setup form preselects
OMARCHY_FORM_KEYBOARD_DEFAULTandOMARCHY_FORM_TIMEZONE_DEFAULTwhen the media sets them. The installer writes a language code to/etc/omarchy/language; system finalization generates that locale, andomarchy-language seedmakes it the user's desktop language without elevating or seeding an input method over the keyboard row they picked. The system locale stays English. Downloads, Pictures, Videos, Documents, and Music are named in English up front so a first login in another language does not create a second set of folders. A deferred install settles its region from the owner's timezone once they pick one; Hong Kong, Macau, and Taiwan stay World.aarch64 can install stable and rc (omacom#16065, omacom#15903)
Every aarch64 install was ending on edge, because the ARM pacman templates and the default channel were edge only. 4.5 is the first release whose
omarchyandomarchy-settingsrun on aarch64. rc and stable templates now exist for aarch64, and every ARM platform except Apple Silicon defaults to stable. Apple Silicon stays on edge templates. A switch from edge to stable or rc ignoresomarchy-devandomarchy-settings-devduring the repository refresh, so pacman cannot downgrade to an old-devbuild that is still listed and then stop before the release swap. That downgrade had left machines on an August build with stable repositories and an invisible bar.Factory reset keeps other systems' boot entries
On a shared ESP, the reset re-keys this installation's Limine entry in place instead of rewriting
limine.conffrom the template, which had dropped Windows and other Linux entries. Only the replaced machine-id directory is removed. Hashes are checked on this installation's rebuilt files (the UKI, or the kernel and initramfs where UKIs are off) and snapshot copies underlimine_historyare left alone. A stage that fails after touching the ESP restores this system's boot files while the staged root is still standing apart. Cleanup no longer aborts the unmount, a root switch that fails halfway puts the running root back, and a signal between the rename and the boot commit does not put the old kernel under the factory root.OMARCHY_PATHis exported so the confirmation prompt still has its strings under sudo (omacom#16032). The fork's USBGuard disable and Thunderbolt trust wipe stay beside that.T3 Code, Brave Origin, and smaller fixes
An update installs the T3 Code AppImage with the channel already chosen and
--no-launch, and migration1792345511themes it. Migration1789091250never shipped in a tag and is removed, so an update no longer stops to ask for a channel and open the app. A customT3CODE_HOMEcounts as having run T3 Code. The package is dropped only once the app is themed.The Brave Origin Copy URL repair runs again, on Brave Origin profiles only. It is migration
1792474908, numbered after the migrations already on quattro, so declining its browser prompt does not hold those back.Also in this range: calculator bindings follow the preinstall gate (omacom#7716); 1Password is under Remove > Services (omacom#15526); the Neovim transparency migration skips a config that is not omarchy-nvim (omacom#15141); the media title scrolls only while the track is playing and the bar is showing it (omacom#15220); a TUI launched under its own app-id gets its terminal's touchpad scroll speed (omacom#15972); the input-method test can compose on a UK install (omacom#16031).
Fork delta against
ce39c5e9is 231 paths, the same set as quattro against81fd2d23.shell/Ui/qmldirkeeps FaceChromeCanvas immediately before FilterInput.