OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
omarchy-core critical common DISPUTE

Fix the screensaver closing itself after a second and leaving the session unlocked

omarchyomarchy-shellquickshellhyprlandwaylanddesktoplaptop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

Confirmed on this machine, omarchy 4.0.2-1 with Hyprland 0.56.2 and quickshell 0.3.1, by reading the shipped code rather than the thread. Claim 1 holds exactly: /usr/share/omarchy/bin/omarchy-screensaver line 44 is if read -n1 -t 1 || ! screensaver_in_focus; then exit_screensaver, and screensaver_in_focus at lines 5 to 7 is hyprctl activewindow -j | jq -e '.class == "org.omarchy.screensaver"'. Claim 2 holds: handleScreensaverWindowClosed at /usr/share/omarchy/shell/plugins/services/idle/Service.qml lines 129 to 139 calls cancelIdleCycle("screensaver-dismissed") for any screensaver window closing, and cancelIdleCycle at lines 100 to 111 stops lockTimer on line 103 and clears idledThisCycle on line 108, which is also the re-arm the record's 225-cycle overnight measurement needs and which the record never explained. Claim 5 holds: default/hypr/apps/system.lua lines 35 to 37 set fullscreen, float and an animation for org.omarchy.screensaver and no size, /usr/share/omarchy/default/foot/screensaver.ini sets no size, and man 5 foot.ini on foot 1.27.0-2 gives initial-window-size-pixels a default of 700x500. Claim 3, the one that matters, also holds in the shipped code: omarchy-launch-screensaver lines 13 to 15 exit before mapping a window when screensaver-off is set, so no closewindow arrives, and the 3 second grace timer at Service.qml lines 272 to 281 cancels only when !idleMonitor.isIdle, so an untouched machine keeps lockTimer and locks at idle.lock. Upstream state rechecked today: issue 6917 is open with 7 comments and pull request 7102 is open and unmerged against quattro, and the thread does support every claim, including the instrumented five runs and the read return of 142. I also checked v4.0.3, published 2026-09-08, because it is newer than this workstation: bin/omarchy-screensaver, bin/omarchy-launch-screensaver, bin/omarchy-toggle-screensaver and default/hypr/apps/system.lua at that tag diff clean against the installed 4.0.2-1 files, and the only idle service change is an unrelated idleConfig fallback, so nothing in this area moved at v4.0.3. Corrected for three things. First, the fix named omarchy-toggle-screensaver, which is omarchy-toggle screensaver-off with no explicit action and therefore a flip: on a machine where the screensaver is already off it turns the screensaver back on and silently restores an unlocked session, which is unacceptable in the one record whose failure mode is a machine that does not lock, so the fix now sets the flag with omarchy-toggle screensaver-off on and confirms it with omarchy-toggle-enabled screensaver-off. Second, the fix presented closing a bar panel as one of "two things that work now" while the record's own symptom says the defect also happens with no panel open, and the last reporter in the thread saw it with only omarchy-background and omarchy-bar present, so that workaround is now labelled as not reliable. Third, jq '.idle' ~/.config/omarchy/shell.json reports only what is written in the file and says nothing about whether idle is enabled, so it is replaced by omarchy-shell idle status, which I ran here and which returned "enabled": false because this workstation has stay-awake set. Verify now leads with omarchy-debug-idle, which I read in full and ran read-only: its Screensaver detector prints disabled exactly when the flag is set, and the journal grep now matches lock-system because the lock event is logged as lock-system: lock-timeout. On claim 4, the empty danger was wrong and is now filled, which follows the corpus precedent in walker-gtk4-vulkan-renderer-amdgpu-hard-freeze of using danger for the residual risk of the unfixed state as well as for the cost of the fix, and there is a genuine fix-side risk here in the toggle flipping back. Severity high is too low for a deterministic, silent loss of session locking measured over nine hours: critical already covers comparable security exposure with no data loss in docker-published-ports-bypass-ufw, and I recommend critical, but the verdict schema carries no corrected_severity and merge_gapfill.py applies only fix, cause, symptom, danger, verify and sources, so severity has to be changed in the record by hand before ingest or it will silently stay high. What I could not exercise: nothing was reproduced here. This workstation has stay-awake set, so omarchy-shell idle status reports "enabled": false and the idle service never runs a cycle, and I was instructed not to launch the screensaver, not to toggle it and not to lock the session. So the compositor declining the fullscreen rule while a layer surface holds keyboard focus, the activewindow false negative itself, the 1.42 second dismissal clustering and the self-sustaining wake loop are all from the thread and from reading the source, not from a local reproduction. The #7193 caveat in the thread, that a native screensaver binary would be execed ahead of the screensaver-off check and would make the toggle depend on that binary honouring the flag, does not apply yet: no native screensaver package is installed here and /usr/bin/omarchy-launch-screensaver carries no such exec.

The cause above was rewritten on 2026-09-11 to match this note; the fix was corrected by the audit itself.
RISK

omarchy-toggle-screensaver is a toggle, so running it twice, or running it on a machine where the screensaver is already off, re-enables the screensaver and silently puts the session back to not locking. Use omarchy-toggle screensaver-off on and confirm with omarchy-toggle-enabled screensaver-off. Until one of those is in place, treat the session as not locking: nothing in the journal reads as an error, and the only visible sign is the display blanking and flashing back every screensaver timeout. Do not treat closing a bar panel as the fix on a machine you leave unattended.

Symptom

The screensaver appears as a small floating terminal window in the middle of the screen instead of filling it, animates briefly, and closes on its own after about one second with no input. The session then does not lock on that idle cycle, and the journal records the screensaver as dismissed:


omarchy idle ... idle-cycle-cancel: screensaver-dismissed

It is deterministic with any bar widget panel open when the idle timer fires, and it also happens with no panel open.

One reporter measured the consequence over two days on omarchy 4.0.2-1 with lock: 300 configured: 239 dismissals, 233 of them between 1.396 and 1.560 seconds after the screensaver started, mean 1.420 seconds with a 0.03 second spread. One overnight stretch ran 225 idle cycles, every one of them dismissed, and the session stayed unlocked from 21:47 until 07:28. In cycles where no screensaver window existed, the lock fired at the deadline every time.

Reported on Omarchy 4.0.0-1 and 4.0.2-1, Hyprland 0.56.2, Quickshell 0.3.0 and 0.3.1, with foot and with Ghostty as the terminal, on Intel and on AMD graphics, single monitor and multi monitor.

Cause

/usr/share/omarchy/bin/omarchy-screensaver runs a poll loop whose exit test is line 44:


if read -n1 -t 1 || ! screensaver_in_focus; then

screensaver_in_focus, line 6, asks the compositor for the class of the focused window:


hyprctl activewindow -j | jq -e '.class == "org.omarchy.screensaver"'

The trigger is any layer surface holding exclusive keyboard focus, which is wider than the bar widget panel in the issue title. While one does, hyprctl activewindow keeps reporting the window that was focused before the screensaver mapped, so the class never matches and the first tick exits. Upstream reproduced it with a bar panel, with the Omarchy menu, and with the agents panel, and named the clipboard picker, emoji picker and polkit prompt as the same shape. One reporter hit it with only omarchy-background and omarchy-bar present and could not identify the surface. A collaborator instrumented the script on a disposable worker and measured five identical runs: exactly one tick at +1.09 seconds, with read returning 142, which is a timeout with no byte read, and the focus check returning false. No keypress is involved, and the one second is the read timeout.

Two details of the appearance follow from the same cause. /usr/share/omarchy/default/hypr/apps/system.lua lines 35 to 37 register fullscreen, float and an animation for org.omarchy.screensaver and no size, and /usr/share/omarchy/default/foot/screensaver.ini sets no size either, so the 700x500 box is foot's own default for initial-window-size-pixels as documented in man 5 foot.ini, and a different terminal gives a different box. The compositor declines the fullscreen rule specifically while a layer surface holds keyboard focus, while the float rule from the same block still applies.

The idle service is what turns a cosmetic problem into an unlocked machine. /usr/share/omarchy/shell/plugins/services/idle/Service.qml lines 129 to 139 treat any screensaver window closing as a user dismissal and call cancelIdleCycle("screensaver-dismissed"), and that function stops lockTimer at line 103. It also clears idledThisCycle at line 108, so the compositor's next idle edge starts a fresh cycle and the whole thing repeats at the screensaver timeout plus about a second, which is why the lock deadline is never reached on a machine nobody touches. Upstream corrected one detail of that loop: omarchy-system-wake is not what re-arms it and generates no input of its own. Nothing in the journal reads as an error, because the cancel is logged at debug level like every other idle event.

omacom/omarchy#7102, which only treats lost focus as a dismissal once the screensaver has actually held focus, is open and unmerged against quattro on 2026-09-11. Nothing in this path changed at v4.0.3: bin/omarchy-screensaver, bin/omarchy-launch-screensaver, bin/omarchy-toggle-screensaver and default/hypr/apps/system.lua at that tag are byte-identical to the files installed by omarchy 4.0.2-1, and the only change to the idle service there is an unrelated fallback for reading idleConfig.

Fix

There is no upstream fix yet, and only one of the two workarounds in circulation actually restores locking.

**Turn the screensaver off. This is the one to use if you care about the machine locking.** Set the flag explicitly rather than flipping it, so the result does not depend on what it was before:


omarchy-toggle screensaver-off on
omarchy-toggle-enabled screensaver-off && echo "screensaver off, lock path clear"

omarchy-launch-screensaver reads that flag at lines 13 to 15 and exits before mapping anything, so no window ever opens and no closewindow event can be misread as a dismissal. The 3 second grace timer at Service.qml lines 272 to 281 cancels the cycle only when the idle monitor has gone **active**, so on a machine nobody touches the lock timer survives and fires at idle.lock. The reporter's control measurement is exactly this: every idle cycle with no screensaver window locked on time, with the same configuration and the same idle service.

To put the screensaver back later:


omarchy-toggle screensaver-off off

The menu command omarchy-toggle-screensaver sets the same flag and sends a notification saying which way it went, but it is a toggle rather than an off switch, so running it on a machine where the screensaver is already off turns it back on and silently restores the defect.

**Closing a bar panel is not a reliable fix.** With no surface holding keyboard focus the screensaver does take focus, fill the output and stay up until you touch the machine, which is why it looks like one. It depends on you remembering before you walk away, and at least one reporter hits this with no panel open and an unidentified surface, so do not rely on it to lock a machine.

Check what the idle service is actually doing, because the screensaver and the lock are separate deadlines and the stay-awake toggle disables both:


omarchy-shell idle status | jq .

"enabled": false there means idle is off entirely and nothing will lock, whatever the timeouts say.

Verify


omarchy-debug-idle 200

That is the shipped diagnostic. Its **Screensaver detector** section prints disabled once the flag is set, and its **Idle IPC status** section reports the effective screensaver and lock timeouts and whether idle is enabled at all.

For the defect itself, read the cycle in the shell journal:


journalctl --user -t omarchy-shell -b --no-pager | grep -E 'process-start: screensaver|idle-cycle-cancel|lock-system'

A broken cycle shows idle-cycle-cancel: screensaver-dismissed about a second after the screensaver started and never reaches lock-system: lock-timeout. A healthy cycle reaches it. After turning the screensaver off, wait out the lock timeout once without touching the machine and confirm the session locks.

Sources

screensaver-self-dismisses-and-session-never-locks