Make omarchy toggle bar on/off match bar visibility

[WoodyWoodster]

Fixes #12082

omarchy toggle bar on hid the bar and off showed it, because the wrapper passed those arguments straight through to the bar-off flag. User-facing on/off now means bar visibility. omarchy-toggle-fullscreen-desktop hides the bar with omarchy-toggle-bar off so fullscreen on/off stay correct.

The on-disk flag is still bar-off. Menu and hotkey callers pass no argument, so they are unaffected.

Testing

• bash test/shell.d/toggle-test.sh — on clears bar-off, off sets it, and fullscreen still enters and leaves together
• env -u NO_COLOR PYTHONDONTWRITEBYTECODE=1 ./test/all — all 238 shell test files passed
• Live session with the checkout binaries on PATH: packaged omarchy-toggle-bar on still hides (the bug); checkout on shows, off hides; omarchy-toggle-fullscreen-desktop on hides the bar and off restores it
• Before/after visual evidence

Acceptance (test/acceptance.d/session-test.sh) is updated for the new mapping. That suite runs in a disposable VM, not the development session, so it is not in this run.

[WoodyWoodster]

Confirming the bug:

Testing on with fix:

Testing off with fix:

[sprajs]

There are several still-open implementations of this same bar-visibility inversion: #12022, #11807, #11849 and #11894 all update the bar/fullscreen helpers and associated tests. They address overlapping reports #12082, #11817 and #11769.

Could one be selected and the others closed as duplicates after consolidating the useful coverage? The explicit invalid-action handling here/#11807 and the implicit/explicit toggle coverage in #11849 are worth retaining alongside the corrected fullscreen and acceptance expectations.

[WoodyWoodster]

There are several still-open implementations of this same bar-visibility inversion: #12022, #11807, #11849 and #11894 all update the bar/fullscreen helpers and associated tests. They address overlapping reports #12082, #11817 and #11769.

Could one be selected and the others closed as duplicates after consolidating the useful coverage? The explicit invalid-action handling here/#11807 and the implicit/explicit toggle coverage in #11849 are worth retaining alongside the corrected fullscreen and acceptance expectations.

I'll go through these today

[kellen-xavier]

I reproduced this bug today on my Omarchy system: omarchy toggle bar on hid the bar, and omarchy toggle bar off was what actually brought it back — exactly the inversion described here. Confirms the packaged binaries (outside this PR) still have the issue.

[omarchybot]

Thank you for this, @WoodyWoodster. I reviewed it at 498fb44. The bug is real and is still on quattro: bin/omarchy-toggle-bar passes on/off straight to the bar-off flag, so omarchy toggle bar on hides the bar. Your fix is correct.

It makes the same change as #7023, which was opened earlier (2026-08-15) against the same base branch and is still open. Both pull requests remap on/off in bin/omarchy-toggle-bar and reject an unknown action with a usage line and exit 1. Both make bin/omarchy-toggle-fullscreen-desktop call bar off when it enters full screen and bar on when it leaves. Both make the same changes to test/shell.d/toggle-test.sh and test/acceptance.d/session-test.sh, including the half-hidden desktop setup at toggle-test.sh:75. #7023 also tests that an unknown action leaves the flag alone, so closing this one loses nothing. #7023 has also been brought up to date with quattro, and it was verified on a disposable Omarchy VM: a reproduction that fails on quattro and passes on its head.

I'm closing this as a duplicate of #7023. The bug reported in #12082 is fixed by #7023 as well. If I've misread how the two differ, please say so, and this can be reopened.

What was checked: I read both diffs against quattro, and Claude Opus 5.5 compared them. This pull request was not run on a worker, because its change is already verified in #7023. Codex Medium was asked separately, from the two pull requests alone, whether they are the same change, and it agreed. Its independence from this review is not guaranteed. The fix now waits on the maintainer, in #7023.