Repository navigation
Conversation
[Medium impact] The PR appears safe to merge within its stated scope. SummaryFreezes checkout-backed Omarchy Mac installs and points users to Omarchy for Apple Silicon.
Reviews (7) · Last reviewed commit: "Describe the Mac upgrade as retired in t..." · Reviewed by Greptile |
maralcbr
left a comment
There was a problem hiding this comment.
Thanks Naeem, freezing in place rather than emptying quattro is the right shape. On a checkout-backed install, the last pull delivers both the frozen omarchy-update-dev and the migration, and the migration runs from the pulled checkout in that same update. The three legacy scripts also refuse before they touch anything. A few things need changing before merge:
- Fresh quattro installs aren't reached.
install.sh:409-410installs theomarchypackage and loadsenv-bootstrap. That setsOMARCHY_PATH=/usr/share/omarchyunless/etc/omarchy.confexists (default/bash/env-bootstrap:10-15). On those machinesomarchy-update-devexits at:7, and migrations come from the package, not this checkout. Only the 3.x upgrade (bin/omarchy-upgrade-to-quattro-mac:248-249),omarchy-dev-linkand the dev channel run from git. Please correct the body to say which installs this reaches. Reaching the rest needs a package release carrying the migration. - The guided setup still changes the disk before it hits the refusal.
README.md:97sends new users tobin/omarchy-mac-setup. It creates the user and locks root (:1307-1321,:804-831), moves/bootand encrypts the root (:715-731), and only then runsinstall.sh(:1047), which now exits 1. The resume unit (:456) then fails the same way on every boot. Please refuse inmain()before(( EUID == 0 ))at:1269, keeping--statusand--abortreachable (tests/test-mac-setup.sh:667-675and:710callmain). Please also add a test that a fresh run writes nothing. - Install VM can't pass on this tree.
test/vm/run-selective-edge:275runsbash install.sh, and:328fails on a non-zero exit. That covers thepull_requesttrigger too, not just push and the nightly run. Could you remove.github/workflows/install-vm.yml, or turn the harness into a check thatinstall.shrefuses? - The README needs to land with this.
README.md:17says the script install still works, and:48-166still documents the guided and by-hand installs. Please replace that section with the redirect, in this PR or a companion merged together.
Smaller things:
- The update badge never clears on a frozen checkout.
bin/omarchy-update-available:9-21still fetches and counts upstream commits, so oncequattromoves, the bar (SystemUpdate.qml:41) shows updates thatomarchy updatecan no longer apply. Worth skipping the checkout there the same way. #765 touches that file too. - Please add tests for the migration and the three refusals. For the migration, use a stub notifier that records its argv and once fails, and check that the update still succeeds.
Nits, fine as follow-ups:
- The summary at
bin/omarchy-update-dev:3anddocs/update-process.md:138,273still describe the pull. - A one-line comment in
install.shwould help the next reader. It would say the body stays because tests extract functions from it. - The migration makes one attempt per user. Over SSH there is no session bus, so the marker is set without a desktop notice. Worth a sentence in the body.
CI: only the bot check reported for 50d603a. There are no results for test/all, tests/all, shellcheck or Install VM, so there's no evidence the tests ran here.
Not code issues, and they need an explicit maintainer decision:
- Whether and when to freeze. The body already waits on Scott.
- How package-backed installs are told.
- Whether #765 lands first, so checkouts freeze on 4.0.5rc1.
- What happens to guided setups already in progress. They resume from their own installed copy of the script, so they'll hit the
install.shrefusal after encryption. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts track (including the Codeberg mirror) receive this commit. A live omarchy update on a checkout-backed install should also show the notice once and no pull afterwards.
Reviewed at 50d603a7c2f7.
— Marcelo's agent
maralcbr
left a comment
There was a problem hiding this comment.
Thanks Naeem. Re-reviewed 37b58f3: everything from the last round is fixed.
- The body now says which installs this reaches, and that matches the code.
omarchy-mac-setuprefuses at:1269-1277, after--statusand before the root check, the piped re-fetch or any step.--abortstill works and--stepis refused even alongside it.tests/test-mac-setup.sh:883-978covers each case, including that a refused run writes nothing.- The Install VM workflow and its guard test are gone, and
docs/testing.md:152says why. - The README redirects at
:17-18and:48-62. omarchy-update-availableno longer counts upstream commits, andupdate-available-test.sh:169-183checks that it makes no git call.- The migration and refusal tests are in, including a failing notifier.
- The summary, docs and
install.shcomment nits are done.
No code blockers. Two things I'd still like fixed:
- The 3.x→quattro upgrade still runs, and it lands on the frozen line.
docs/upgrade-to-quattro.md:45(linked fromREADME.md:138) still publishes thecurl | bashforomarchy-upgrade-to-quattro-mac. That upgrade is one-way: it switches the checkout toquattro(:125-134), removes the 3.x desktop (:317-351) and switches networking (:353-362). It then runsomarchy-migrate(:314), which tells the user to reinstall. Could it refuse the same way right after the argument loop (:84)? The environment checks, the confirmation prompt andsudo -vat:112all run beforemain(), so a refusal insidemain()would come too late.-hexits inside the loop, so help keeps working. Please also replace the "Run the upgrade" section and the README link with the redirect, and add the script tolegacy-install-refusal-test.sh. If keeping it open is deliberate, so 3.x users at least see the notice, please say so in the body instead. - Setups already in progress loop on the refusal. The old installed setup script re-fetches the ref (
bin/omarchy-mac-setup:1031-1036) and reaches the newinstall.shrefusal. It then says "fix it and run --resume" (:1049), and the unit retries on every boot (:456). Only--abortgets out, and nothing says so. Suggest one more line in theinstall.shrefusal: "If a guided setup is in progress, stop it with: sudo omarchy-mac-setup --abort".
Nits, fine as follow-ups:
docs/omarchy-mac-e2e-volunteer-checklist.md:82still sends volunteers to the guided installer. Worth a retired note at the top.
CI: only the bot check shows for 37b58f3. There are no results for test/all, tests/all or shellcheck, so there's no CI evidence for the new tests yet.
Still open, and not code issues:
- Whether and when to freeze. The body already waits on Scott.
- Whether the 3.x upgrade stays open.
- How package-backed installs are told.
- Whether #765 lands first. It touches
install.shandomarchy-update-availabletoo, so if it lands after this, please check it doesn't bring back the checkout fetch. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts track receive this commit. A live omarchy update on a checkout-backed install should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at 37b58f3fe6d1.
— Marcelo's agent
| # Omarchy Mac is no longer maintained. A setup can still be inspected or | ||
| # stopped, but not started, resumed or re-run. --step runs ahead of --abort | ||
| # below, so it is refused even alongside --abort. | ||
| if (( ! abort )) || [[ -n $step_only ]]; then |
There was a problem hiding this comment.
Fixed: refused before the root check and the piped re-fetch, with --status and --abort still reachable.
— Marcelo's agent
The upgrade is one-way and lands on a line that no longer updates. Refuse right after the argument loop, before any check, prompt or sudo, and point 3.x users at Omarchy for Apple Silicon.
An in-progress guided setup reaches the install.sh refusal on every boot. Name the way out: sudo omarchy-mac-setup --abort.
maralcbr
left a comment
There was a problem hiding this comment.
Thanks Naeem. Re-reviewed 987dffc: everything from the last round is fixed, and I found no blockers.
omarchy-upgrade-to-quattro-macnow refuses at:86-89, right after the argument loop and before the environment checks, the prompt andsudo -v.-hstill works. The guide (docs/upgrade-to-quattro.md:42-44) andREADME.md:138point to the Mac manual.legacy-install-refusal-test.shcovers the script, andupgrade-to-quattro-mac-test.sh:104-109checks that the guide no longer publishes the one-liner.- The
install.shrefusal now namessudo omarchy-mac-setup --abort(:15). An in-progress setup sees it, because the resume unit writes to tty1 (omarchy-mac-setup:457-460). The old installed copy at/usr/local/binstill handles--abortand removes the unit. - The volunteer checklist is marked retired at the top.
Nits, fine as follow-ups:
bin/omarchy-upgrade-to-quattro-mac:3and the usage at:12-21still describe the upgrade and print thecurl | bashline. Suggest a summary such as "Retired: points Omarchy Mac 3.x installs to Omarchy for Apple Silicon", plus one usage line saying the upgrade no longer runs.- A one-line comment after
:89, like the one atinstall.sh:11, would say the rest stays because tests extract its functions. - Worth a sentence in the body that the freeze covers any machine running from a checkout of this fork, including an x86 machine set up with
omarchy-dev-link, sinceomarchy-update-dev:7-10has no architecture guard. That's the intended effect, it just isn't stated.
CI: only the bot check shows for 987dffc. main.yml runs on pull_request, but there are no results for shellcheck, test/all or tests/all, so there's still no CI evidence for the new and changed tests. Could those jobs be run on this head before merge?
Still open, and not code issues:
- Whether and when to freeze. The body waits on Scott.
- Closing the 3.x upgrade is now part of the freeze and needs the same maintainer sign-off.
- How package-backed installs are told.
- Whether #765 lands first. If it lands after this, please check it doesn't bring back the checkout fetch in
omarchy-update-availableor move theinstall.shgate. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts and scripts fetch from receive this commit, including omarchy-mac/omarchy-mac (omarchy-mac-setup:44, omarchy-upgrade-to-quattro-mac:21). A live omarchy update on a checkout-backed install should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at 987dffc8faae.
— Marcelo's agent
| echo "Omarchy Mac is no longer maintained and can't start new installs." >&2 | ||
| echo "Install the official Omarchy for Apple Silicon: brew install omacom/mac/omarchy" >&2 | ||
| echo "Mac manual: https://omarchy.org/manual/mac-support/" >&2 | ||
| echo "If a guided setup is in progress, stop it with: sudo omarchy-mac-setup --abort" >&2 |
There was a problem hiding this comment.
Fixed: an in-progress setup now sees the way out on tty1, and the old installed copy's --abort removes the unit.
— Marcelo's agent
|
Thanks for the re-review, Marcelo. The nits are in now (800f199): the 3.x upgrade script is described as retired, the --abort line is in the install.sh refusal, and the volunteer checklist is marked retired. On CI: I checked, and no CI jobs (shellcheck, test/all, tests/all) have run on this PR. They haven't run on any omarchy-mac PR since 5 Nov, so it isn't specific to #804. Could someone with repo admin check why Actions isn't starting on new PRs? Once it does, I'll re-run them on this head. |
maralcbr
left a comment
There was a problem hiding this comment.
Thanks Naeem. Re-reviewed 800f199: the last round's nits are in, and I found no blockers.
- The summary at
bin/omarchy-upgrade-to-quattro-mac:3and the usage at:14-15now say the upgrade is retired and point to the Mac manual. Thecurl | bashline is gone. :85says the rest of the file stays for the tests. Those tests extract by function name and fixed strings (upgrade-to-quattro-mac-test.sh:9-11), so the shorter usage doesn't move anything they rely on. The refusal at:80-83is unchanged.
Nits, fine as follow-ups:
:17-21still lists--yes,--rebootand--refas working options. They're parsed and then ignored, which is harmless. A short "accepted but ignored" note would make that clear, or leave it.- Worth a clause in the body that "dev-link mode" covers any architecture, so an x86 machine dev-linked to this fork is frozen too (
bin/omarchy-update-dev:7-10has no architecture guard). That's the intended effect. The body just doesn't say it.
CI: only the bot check shows for 800f199. There are still no results for shellcheck, test/all or tests/all, so the new and changed tests have no CI evidence yet. Thanks for checking the wider Actions problem. That needs someone with repo admin, and the jobs should run on this head once it's fixed.
Still open, and not code issues:
- Whether and when to freeze, including closing the 3.x upgrade. The body waits on Scott.
- Once frozen, checkouts get no further fixes to Omarchy's own scripts and config, security fixes included. System and AUR packages still update (
bin/omarchy-update:49,61). Worth a line in the body so that's part of what's signed off. - How package-backed installs are told.
- Whether #765 lands first. If it lands after this, please check it doesn't bring back the checkout fetch in
omarchy-update-availableor move theinstall.shgate. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts and scripts fetch from receive this commit, including omarchy-mac/omarchy-mac (omarchy-mac-setup:44). A live omarchy update on a checkout-backed install should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at 800f199a05d7.
— Marcelo's agent
| echo "Mac manual: https://omarchy.org/manual/mac-support/" >&2 | ||
| exit 1 | ||
|
|
||
| # The rest of this file stays because tests extract its functions. |
There was a problem hiding this comment.
Fixed: the summary and usage now describe the script as retired, and this comment explains why the rest stays.
— Marcelo's agent
scottjones
left a comment
There was a problem hiding this comment.
Thanks Naeem, this is really thorough work: the VM runs, the per-install-type breakdown, and tests that fail on the old code. Thanks to Marcelo for the review rounds as well.
This is fine to merge once omacom#17735 is merged and https://omarchy.org/manual/mac-support/ shows the Apple Silicon section. Here's why the order matters. Every message in this PR links that page, and today it still says M-series Macs aren't supported. The migration notice fires once per install and sets its marker either way. If this lands first, a checkout user clicks the notice once, lands on a page saying their Mac isn't supported, and never gets the notice again. omacom#17735 is mine, so I'll get it in and check the live page. Could you add that dependency to the body, under "Not decided in this PR"?
Two small things, fine to fix here or as follow-ups:
legacy-install-refusal-test.shstill passes with theexit 1atbin/omarchy-upgrade-to-quattro-mac:83removed. The temp HOME has no Omarchy checkout, so the script's later checkout check stops it first, with the manual line already printed. Creating$work/home/.local/share/omarchy/.gitin the test would let it tell the gate apart from that check.bin/omarchy-upgrade-to-quattro:121still tells Apple Silicon users to runomarchy-upgrade-to-quattro-mac, which now refuses. Nobody gets stuck, since the refusal prints the manual link, but pointing straight at the manual saves a hop.
One wording thought: the notice and refusals open with "Omarchy Mac is no longer maintained". The README banner already says it better, that the Omarchy Mac team joined Omacom and Omarchy for Apple Silicon is the official version. Something like "Omarchy Mac is now part of Omarchy for Apple Silicon" in the notice would match it.
Matches the README banner. The retired upgrade's usage also says its options are accepted and ignored.
maralcbr
left a comment
There was a problem hiding this comment.
@malik-na re-reviewed at 30385dd. Thanks Naeem, the last round and Scott's points are in, and I found no blockers.
- The usage now says
--yes,--rebootand--refare accepted and ignored (bin/omarchy-upgrade-to-quattro-mac:17-20). The body now covers dev-links on any architecture and the end of script-level fixes, security fixes included. - Every refusal and the notice now say "Omarchy Mac is now part of Omarchy for Apple Silicon". The tests match on the manual URL rather than the wording, so nothing they check moved.
bin/omarchy-upgrade-to-quattro:121now sends Apple Silicon users straight to the manual. Only the aarch64 refusal text changed, and the x86 path is untouched.- The omacom#17735 dependency is in the body.
One thing worth fixing:
- The refusal test still passes on CI with the upgrade gate removed.
test/allruns on x86_64ubuntu-24.04(main.yml:74). With theexit 1atbin/omarchy-upgrade-to-quattro-mac:82gone, the architecture check at:89exits 1 there, after the manual line is already on stderr. Nothing reaches stdout, no stub runs and HOME is unchanged, so every assertion inlegacy-install-refusal-test.sh:41-45still holds. The new.gitin HOME fixes this on aarch64, where you ran it, but not on the runner. Could you add aunamestub tomock_binthat printsaarch64? It has to sit outside the record-only loop at:17-22, because$(uname -m)needs the output. Then a missing gate reaches the confirmation text at:93-107on stdout, and:43catches it.
Nits, fine as follow-ups:
- The comment at
bin/omarchy-upgrade-to-quattro:119-120still says the Mac upgrade "wires the source checkout instead", anddocs/upgrade-to-quattro.md:33still points toomarchy-upgrade-to-quattro-mac. Both are worth a refresh. - The docs and two comments still open with "Omarchy Mac is no longer maintained" (
README.md:17,48,50,docs/upgrade-to-quattro.md:44,docs/update-process.md:139,bin/omarchy-mac-setup:1269,tests/test-mac-setup.sh:885, among others).README.md:11-13already gives the context, so matching the new wording there is optional.
CI: only the bot check shows for 30385dd. There are still no results for shellcheck, test/all or tests/all, so the new and changed tests have no CI evidence. Those jobs should run on this head once Actions is working again.
Still open, and not code issues:
- Scott's go-ahead depends on omacom#17735 merging and the live manual page showing the Apple Silicon section.
- How package-backed installs are told.
- Whether #765 lands first. If it lands after this, please check that it doesn't bring back the checkout fetch in
omarchy-update-availableor move theinstall.shgate. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts and scripts fetch from receive this commit, including omarchy-mac/omarchy-mac (omarchy-mac-setup:44). A live omarchy update on a checkout-backed install at the final head should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at 30385dd8514a.
— Marcelo's agent
| --ref BRANCH Check out BRANCH instead of quattro | ||
| -h, --help Show this help | ||
|
|
||
| --yes, --reboot and --ref BRANCH are still accepted, and ignored. |
There was a problem hiding this comment.
Fixed: the ignored options are now stated.
— Marcelo's agent
maralcbr
left a comment
There was a problem hiding this comment.
@malik-na re-reviewed at cff45fd. Thanks Naeem, the last open item is fixed, and I found no blockers.
- The refusal test now needs the upgrade gate on x86 runners. The
unamestub atlegacy-install-refusal-test.sh:24-27sits outside the record-only loop (:17-22), so$(uname -m)getsaarch64. With theexit 1atbin/omarchy-upgrade-to-quattro-mac:82removed, the architecture check at:89passes,:91passes on the.gitin HOME, and the confirmation text at:93-107reaches stdout, which:48catches. The stub is only on the script'sPATH(:40). None of the other three scripts callsunamebefore its gate (install.sh:71comes after the refusal), so the stub doesn't change what those runs test.
The two wording nits from the last round are still open, and both are still fine as follow-ups: bin/omarchy-upgrade-to-quattro:119-120 and docs/upgrade-to-quattro.md:33 still describe the Mac upgrade as if it runs.
CI: only the bot check shows for cff45fd. There are still no results for shellcheck, test/all or tests/all, so none of the new or changed tests have CI evidence. Those jobs should run on this head once Actions is working again.
Still open, and not code issues:
- Scott's go-ahead depends on omacom#17735 merging and the live manual page showing the Apple Silicon section.
- How package-backed installs are told.
- Whether #765 lands first. If it lands after this, please check that it doesn't bring back the checkout fetch in
omarchy-update-availableor move theinstall.shgate. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts and scripts fetch from receive this commit, including omarchy-mac/omarchy-mac (omarchy-mac-setup:44). A live omarchy update on a checkout-backed install at the final head should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at cff45fdebf01.
— Marcelo's agent
|
|
||
| # On an x86 runner the upgrade's architecture check would otherwise stand in | ||
| # for the gate. | ||
| printf '#!/bin/sh\necho aarch64\n' >"$mock_bin/uname" |
There was a problem hiding this comment.
Fixed: with uname reporting aarch64, a missing gate now reaches the confirmation text on stdout, and :48 catches it.
— Marcelo's agent
maralcbr
left a comment
There was a problem hiding this comment.
@malik-na re-reviewed at fbf3fed. Thanks Naeem, both wording nits from last round are fixed. I found no blockers, just two small doc items below.
bin/omarchy-upgrade-to-quattro:118-120now says the Mac upgrade is retired. The refusal at:121is unchanged, and the static guard inaarch64-compat-test.sh:69-72still matches it.docs/upgrade-to-quattro.md:33-34calls the Mac upgrade retired, and:45uses the "now part of Omarchy for Apple Silicon" wording.
Smaller things, not blocking:
- The README overstates the freeze.
README.md:17-18and:59say existing installs "get no more updates", butomarchy updatestill updates system and AUR packages (bin/omarchy-update:49,61). Someone reading that literally might stop updating and miss system security fixes. Suggest: "Existing installs keep working, andomarchy updatestill updates system and AUR packages, but Omarchy Mac's own scripts and config get no more fixes." docs/upgrade-to-quattro.mdstill describes a running upgrade in a few places.:16-25and the "Back up first" bullet at:35-41sit above the "rest of this page describes what the upgrade did" line.:68says re-running is safe.:79-80saysomarchy update"now updates the checkout", which the freeze undoes.:86and:90say to re-run the script, which now refuses. Could those say the checkout is frozen, and point at the symlink loop at:92or the Mac manual instead? Fine as a follow-up.
CI: only the bot check shows for fbf3fed. There are still no shellcheck, test/all or tests/all results, so the new and changed tests have no CI evidence yet. They should run on this head once Actions is working again.
Still open, and not code issues:
- Scott's go-ahead depends on omacom#17735 merging and the live manual page showing the Apple Silicon section.
- How package-backed installs are told.
- Whether #765 lands first. If it lands after this, please check that it doesn't bring back the checkout fetch in
omarchy-update-availableor move theinstall.shgate. mainand the Pages site still serve the Omarchy 3 installer.
For release, not the merge: please confirm that the remotes existing checkouts and scripts fetch from receive this commit, including omarchy-mac/omarchy-mac (omarchy-mac-setup:44). A live omarchy update on a checkout-backed install at the final head should also show the notice once, pull nothing after that, and clear the badge.
Reviewed at fbf3fedd686f.
— Marcelo's agent
| # upgrade (omarchy-upgrade-to-quattro-mac) wires the source checkout instead. | ||
| [[ $(uname -m) != "aarch64" ]] || fail "This is the upstream x86 upgrade. On Apple Silicon run omarchy-upgrade-to-quattro-mac." | ||
| # mirrors and packages, which do not serve aarch64. The Mac upgrade | ||
| # (omarchy-upgrade-to-quattro-mac) is retired, so Apple Silicon goes to the manual. |
There was a problem hiding this comment.
Fixed: the comment now says the Mac upgrade is retired, matching the refusal at :121.
— Marcelo's agent
| Omarchy's x86_64 package mirrors, which do not serve aarch64. Running it | ||
| breaks pacman for the whole system. The command ships in this repo because we | ||
| track upstream — the Mac upgrade is `omarchy-upgrade-to-quattro-mac` below. | ||
| track upstream. The Mac upgrade, `omarchy-upgrade-to-quattro-mac`, is retired |
There was a problem hiding this comment.
Fixed: the Mac upgrade is described as retired.
— Marcelo's agent
Why
Omarchy Mac is no longer maintained. Omarchy for Apple Silicon is now an official Omacom solution. This PR freezes the installs that update from this repository's git history, and points their owners at the official version once. It also stops this repository's scripts from starting new installs. Nothing on an installed desktop stops working. This is the same change as maralcbr#289.
Which installs this reaches
quattrowithomarchy-upgrade-to-quattro-mac, and machines in dev-link mode (omarchy-dev-link) on any architecture, so an x86 machine dev-linked to this repository is frozen too. There,OMARCHY_PATHis the git checkout. The nextomarchy updatepulls this commit once, then stops pulling. The migration runs from the checkout in that same update. After that, a frozen checkout gets no more fixes to Omarchy's own scripts and config, security fixes included. System and AUR packages still update, becauseomarchy updatestill runs the package steps.quattroinstall made withinstall.shoromarchy-mac-setup. Those installs run from theomarchypackage, withOMARCHY_PATH=/usr/share/omarchy, and get their migrations from the package. To reach them, a package release has to carry this migration. That is a separate decision, outside this PR.main. They pull from the Codeberg mirror, not GitHub.omarchy-mac-setup, which still callsinstall.sh. They stop at theinstall.shrefusal, after the disk changes they had already made.What changes
omarchy updateno longer pulls the checkout.bin/omarchy-update-devprints that Omarchy Mac is now part of Omarchy for Apple Silicon and links https://omarchy.org/manual/mac-support/. It exits 0, so the package updates and migrations still run.migrations/1794628107.shsends a critical, persistent notification. Clicking it opens the Mac manual. A failure to notify never fails the update. The migration runs once per user. Over SSH with no session bus, the marker is still set, so that user only sees the line printed in the terminal.bin/omarchy-update-availableno longer fetches the checkout. A frozen install doesn't show updates it can't apply.bootstrap.sh,setup.shandinstall.shprint the brew command and the manual link, then exit 1.omarchy-mac-setuprefuses the same way before it changes anything, and--statusand--abortstill work. Theinstall.shrefusal also tells a guided setup in progress how to stop, withsudo omarchy-mac-setup --abort.omarchy-upgrade-to-quattro-macrefuses right after its argument loop, before any check, prompt orsudo. The upgrade is one-way, and it would land the user on a line that no longer updates. Its--yes,--rebootand--refoptions are accepted and ignored. The upgrade guide, the README link and the x86omarchy-upgrade-to-quattromessage for Apple Silicon now point to the Mac manual. 3.x installs that already moved keep working as checkout-backed installs..github/workflows/install-vm.ymlis removed. Its only job raninstall.shin a VM, andinstall.shnow refuses by design. The refusal is covered by shell tests instead.Verification
UTM VMs (Arch Linux ARM, aarch64)
These ran in disposable clones of the Quattro golden VMs, on 2026-11-14. The VMs are not Macs, so Apple-only paths did not run.
omarchy4.0.3rc4-91 built from this branch's head (800f199).omarchy update -yexited 0, installed the build and left no pending migration.1794628107.sh.omarchy-update-devexited 0 silently, the badge said "up to date" andomarchy-upgrade-to-quattro-macrefused../tests/allpassed../test/allhad 4 failures, and they also fail onquattrowithout this change.ascii,branding-about-animationandssh-reconnectfail the same way on 667500c.arm-channel-stagingneeds a disk-backed/tmp, which the VM lacks.quattrobefore this change (667500c). It was then dev-linked to a checkout of 667500c that trackseol/quattro.omarchy update -yexited 0 and pulled the checkout once to 800f199. It printed the notice and ran the notice migration once.install.shprinted the four refusal lines, including the--aborthint, and exited 1. No Omarchy package was installed, and the machine rebooted cleanly.Container
These ran in a Debian trixie aarch64 container with bash 5.2 and GNU tools, as uid 1000. CI here doesn't run them, so this is the only evidence that the tests ran.
test/shellsuite. The baseline (quattro667500c) passed 178 of 296 files. This branch passes 180 of 297. No file passes onquattroand fails here, or the other way round. The two extra passes are the new tests. The remaining failures happen on both trees and are environmental, such as a missingpacman-conf,vercmporjq.tests/test-mac-setup.sh. It passed 142 of 142 checks before and passes 168 of 168 now. 31 checks are new, and 5 obsolete finished-machine checks were removed. A fresh run,--resume,--stepand a finished machine are each refused, and none of them writes a file or runs a command.--statusand--abortstill work.|| true, the notice test fails. Against aninstall.shthat callspacmanbefore the gate, the refusal test fails.exit 1removed fromomarchy-upgrade-to-quattro-mac, the test fails, because the temp HOME now holds a 3.x checkout and the later checkout check can't stop the script first.-S erroris clean on every changed script.Not decided in this PR
mainand the Pages site still serve the Omarchy 3 installer.