Repository navigation
Backport nine quattro fixes to 4.5 for RC3 - #17220
Conversation
* Let Omasnap's scroll capture see the screen behind its overlay * no-mistakes(review): Drop source-grep assertion for removed no_screen_share rule (cherry picked from commit e71dc3f)
t3_cli starts the AppImage in the background with its output sent to a file and reads that file in a loop until the mount point appears. The background job creates the file only once it runs, so a loop that gets there first fails "head: cannot open .../mount" and, under set -e, ends the install while it themes T3 Code. Create the file empty first; the loop then reads it empty until the mount point is written. The new CI caught it on quattro's first run. In a VM under load, t3code-install-test.sh failed 27 of 60 runs before this change, 22 of them on that missing file, and none of 60 on it after. With the mounter held back 50 ms before its redirection, the shipped loop died in 200 of 200 runs and this one in none. (cherry picked from commit 8478247)
Apple Silicon had edge templates alone, so an rc Mac image had no rc pacman configuration to install and the Mac couldn't follow rc. Add aarch64-apple pacman-rc.conf and mirrorlist-rc: the edge ones with Omarchy's rc repository and the rc Arch Linux ARM mirror. Stable stays closed until the first Mac stable release, and Apple Silicon still defaults to edge. (cherry picked from commit 1676904)
) * Let the installer's timezone picker always be searched by typing The timezone prompt used gum choose whenever it had a guess, and gum choose takes no typed text. Since the setup form gained a timezone fallback (#16049) and omarchy-iso passes the language's capital or UTC, there is always a guess, so no installer user can type a city. With Europe/Amsterdam as the guess, typing "tokyo" moves the cursor up on the "k", and Enter then sets Etc/Zulu. The prompt now always opens gum filter, with the guess as the search text. Enter takes the guess, and erasing it searches all zones. In gum 2.0.2 every one of the 598 zones of tzdata 2026e ranks first for its own exact name, so the guess is the highlighted row. --selected is left out on purpose. At limit 1, gum filter prints the --selected entry even after the user searched for another zone and pressed Enter on it. * List the timezone guess first instead of typing it into the search With the guess as the search text, the user had to erase it before searching. Most people start typing straight away. Typing "london" after the guess Europe/Amsterdam searched "Europe/Amsterdamlondon", which matches no zone, so Enter set UTC. The guess now leads the list and the search starts empty. Enter takes the guess, and the first key typed starts a new search of every zone. A guess that is not a zone timedatectl lists is left out, as gum choose --selected ignored one before. The test now checks the guess is the first entry, listed once, and that the call has neither --value nor --selected. * Reject multiline timezone guesses before reordering the list The whole-line membership pattern also matches adjacent lines. A malformed guess with two real zones therefore promotes both, and a guess containing the whole list makes grep return 1 under set -e. Reject embedded newlines before testing membership. Add checks for adjacent zones and the whole list. The adjacent-zone check fails on the previous commit. All 23 setup-form test groups pass in the booted rc2 ISO, and typing Lon then Enter returns Europe/London. (cherry picked from commit 7f4b1dd)
Hyprland's cm_auto_hdr defaults to switching an HDR-capable monitor into HDR whenever a window goes fullscreen, which left SDR content looking washed out on the Pro Display XDR. Turn it off; no HDR for now. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit f766bfa)
omarchy-settings skipped its etc-overrides on Apple Silicon, so Mac accounts were created from Arch's stock /etc/skel/.bashrc, which never sources default/bash/rc. omarchy-pkgs#1857 fixes the skel file for new accounts; this migration replaces an existing user's ~/.bashrc with default/bashrc, saving the old one as ~/.bashrc.bak. A bashrc that already loads Omarchy's, or a symlink into a dotfiles repo, is left alone. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 0b10569)
We already hide the update-news popup; the donation nag still appears after upgrades. Disable it alongside. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit ee8ad06)
… runs The migration saved the user's ~/.bashrc as ~/.bashrc.bak, overwriting a backup already there. Save it under a name nothing holds, as other migrations do with mktemp, and print that name. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 16bf8af)
…#17000) * Let one sudo password last ten minutes in a terminal Omarchy commands revoked sudo's timestamp on entry and exit and authenticated through a sudo -N wrapper, so every privileged step asked again: three prompts at "Create system snapshot", six for a dev link session. Ship timestamp_timeout=10, drop the revocation and the wrapper, and keep omarchy update's single authorization and keepalive. Channel switching keeps its stop for a destination older than this tree, now keyed on what the next steps need: omarchy-update-pacman, and an updater that restores the caller's PATH. Releases 4.0.5 and 4.0.6 have neither. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> * Make update, refresh and channel switching plain scripts again The privileged-Bash startup check, environment scrubbing, OMARCHY_PATH source-root check and pinned PATH in omarchy-update, omarchy-refresh-pacman and omarchy-channel-set only guarded against code already running as the user, the same threat the password boundary was dropped for. They return to plain Bash as in 4.0.6, with every feature since kept; hooks and mise inherit the caller's PATH directly. The channel swap calls pacman itself again, so a refresh that installs an older release without omarchy-update-pacman cannot strand it, and the older-destination stop is no longer needed. Leaving dev drops the unlinked checkout's bin from PATH so the packaged update runs the packaged helpers. omarchy-update-stay-awake and omarchy-sudo-passwordless keep the security library: OM-SEC-12 and OM-SEC-01 guard against other accounts, not the caller. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> --------- Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Codex Medium <noreply@openai.com> (cherry picked from commit 56a910b)
[Critical impact] The PR changes how sudo authorization persists across Omarchy and normal system commands. The PR appears safe to merge, with a non-blocking recommendation to protect the channel swap from desktop shutdown. Findings
SummaryBackports nine fixes for RC3:
The channel swap should retain its protection from desktop shutdown when calling Acknowledged decisions:
|
| # The refresh can have installed an older release without this tree's helpers, | ||
| # so the swap calls pacman itself. --ask 4 accepts omarchy <-> omarchy-dev | ||
| # replacement prompts without file overwrites. | ||
| sudo env OMARCHY_UPDATE_PACMAN=1 pacman -S --needed --noconfirm --ask 4 "${packages[@]}" |
There was a problem hiding this comment.
Shutdown can interrupt the swap
The direct pacman call loses the protection from desktop shutdown that omarchy-update-pacman provided. If the desktop session ends during the swap, it can kill pacman and leave packages partly installed.
Calling pacman directly avoids relying on a helper that an older release may remove. Keep that approach, but copy the helper's use of systemd-run --scope when systemd is running.
| sudo env OMARCHY_UPDATE_PACMAN=1 pacman -S --needed --noconfirm --ask 4 "${packages[@]}" | |
| scope=() | |
| if [[ -d /run/systemd/system && ! -L /run/systemd/system ]]; then | |
| scope=(systemd-run --scope --quiet --collect) | |
| fi | |
| sudo env OMARCHY_UPDATE_PACMAN=1 "${scope[@]}" pacman -S --needed --noconfirm --ask 4 "${packages[@]}" |
Knowledge Base Used: Update orchestration and system maintenance
Backports nine quattro fixes to 4.5 for RC3. Every commit is a
git cherry-pick -xof the squash commit that landed on quattro, applied in quattro order, and none of them conflicted..bashrcand never loaded Omarchy's~/.bashrc.bakThis supersedes #16894, which backports the same Omasnap change. Its diff is identical to #16715, so #16894 can close once this merges.
Worth a close look
Defaults timestamp_timeout=10inetc/sudoers.d/omarchy-sudo-timeout, and deletes the tests that guarded the old boundary.channel-switch-test.shpasses on this branch.1793315076.sh(Load Omarchy's shell setup on Apple Silicon #17041, amended by Keep an earlier ~/.bashrc.bak when the Apple Silicon bashrc migration runs #17158). It sorts afterv4-5-0's newest,1793069744.Verified
./test/allon this branch: 398 of 403 shell files pass.agent-account-exec,manifest-entrypoints,runtime-smokeandbackground-intro-lifecyclefail the same way on plainv4-5-0.shell-idle-spawnfailed in the full run and passes alone. It reads the live session, so that was session noise.🤖 Generated with Claude Code