Confirm a suspend actually locked the screen, since a stalled lock latches as locked with nothing engaged
Held every mechanical claim against /usr/share/omarchy on omarchy 4.0.2-1: Service.qml lines 37, 63 to 77, 135, 151, 255 and 513 to 521, omarchy-system-lock line 8 and its trailing || true, omarchy-system-sleep-lock lines 11 to 35 and 79 to 124, omarchy-hyprland-session-locked, omarchy-restart-shell lines 28 to 37, omarchy-system-lid-close line 17, utilities.lua lines 34 and 126, idle Service.qml line 68, and busctl reporting InhibitDelayMaxUSec 15000000, which derives the 12 s budget. All six cited blobs were fetched at the pinned sha and are byte-identical to the installed files, and Service.qml and the four scripts are also byte-identical at the v4.0.3 and v4.0.4 tags, so the latch ships in every 4.0.x release. Gate 1 holds: issue #10299 was opened 2026-09-05 by mchldotdev, no one connected to this repository, and its body and the collaborator comment contain every mechanism the record states. Gate 3 holds: the trigger is unknown to everyone and the record gives none. Four things were wrong. The symptom said the machine wakes straight to the desktop, but the reporter resumed into Hyprland's crashed-lockscreen failsafe and the unlocked-desktop observation belongs to the second reporter's idle and manual locks, so the symptom now follows the sources. The cause said no fix was proposed, but PR #11811 (2026-09-14) was, folded into #12064 and then into #12245, which is open and unmerged, so the record now says that. The fix and verify told the reader to run omarchy-hyprland-session-locked after a suspend or lock, which means from a TTY or ssh, where hyprctl exits with HYPRLAND_INSTANCE_SIGNATURE not set (hyprctl/src/main.cpp at v0.56.2, lines 512 to 516) and the script therefore returns 2 every time. The corrected fix exports the signature the way omarchy-restart-shell does, and separates the exit 0 case, which is a stranded lock and belongs to the existing record plugin-file-write-while-locked-strands-session, from the latch. The danger claimed there was no way back over ssh, which the corpus's own stranded-lock record contradicts, so it now says you cannot authenticate out of it from there. Severity stays high, against the brief's scale: the consequence is a session left unlocked while every check reports it locked, which is private data exposed to anyone with physical access, and critical is reserved for an unauthenticated remote or radio-range path. It is not medium, because the control is not merely off, it is reported present while absent. Frequency stays occasional: two reporters, no reproduction on demand. Nothing was exercised: no lock, no shell restart, no hyprctl beyond a read-only -j monitors to confirm solitaryBlockedBy exists on Hyprland 0.56.2. Re-checked against 4.0.4-1 2026-09-19: Re-checked on omarchy 4.0.4-1 (pacman -Q) against the v4.0.4 tag, commit c668141e9c42b13c80c9ca4ea108e11708c5e8a5. The latch is still present: /usr/share/omarchy/shell/plugins/lock/Service.qml is byte-identical to the v4.0.4 tag and has the same git blob (9ecb1cc) at v4.0.2 and v4.0.3, and the v4.0.2...v4.0.4 compare does not touch it. Traced in the QML: locked at line 37 is lockRequested OR the compositor's answer, beginLock sets lockRequested at line 135, the lock IPC at lines 513 to 516 returns ok without calling beginLock again while locked is true, isLocked at 519 to 521 reads the same property, and only finishUnlock (151) and onLockStateChanged (255) clear it, so one stalled request still makes every later lock request report success without locking. omarchy-system-lock, omarchy-system-sleep-lock, omarchy-restart-shell, omarchy-hyprland-session-locked and omarchy-system-lid-close are unchanged between the tags and match the installed files, and the restart-shell relock logic at lines 28 to 37 still behaves as the fix and danger describe. What 4.0.3 changed near this (the plugin authentication boundary in PR #10733, keepLoaded for the lock service across plugin reloads, and PATH hardening in omarchy-apply-lock) does not touch the latch, and no code calls the lock service through the shell's service map, so moving it out of that map changed nothing here. Three things needed updating. The idle service's screensaver gate moved from line 68 to line 69, because 4.0.3 rewrote two lines above it. PRs #11811 and #12064 are now closed unmerged, not only folded, and #12245 is still open. And quattro's Service.qml at e38c1d1 is no longer identical to v4.0.4, because video wallpaper support landed there, but the three latch sites are unchanged at lines 45, 86 and 585. Issue #10299 is open as of 2026-09-19. Nothing was exercised: no lock, no shell restart, no lock IPC call, no suspend.
Do not mute or dismiss the critical "Screen did not lock before suspend" notification that omarchy-system-sleep-lock sends. It is the only signal this bug gives before a laptop that looked suspended turns out to have been sitting unlocked. omarchy-restart-shell behaves differently depending on what omarchy-hyprland-session-locked reports at the moment it runs. When Hyprland holds a lock and the shell does not report one, for example a stale locker left behind by a crashed shell, the restart relocks the session, and over ssh that leaves you with a lock you cannot authenticate out of from there. Check that exit code first, and run the restart from the machine itself when you cannot. omarchy-system-sleep-lock fails open on purpose and caps its own wait at 12 seconds, so raising InhibitDelayMaxSec does not turn the lock into a veto on suspend.
Symptom
A lock that should have happened did not, and the shell says it did. The idle timer fires, or the lid closes, or omarchy system lock or Super+Ctrl+L is used, and no lock screen appears: the desktop stays where it was. omarchy-shell lock isLocked prints true all the same, omarchy-shell lock status shows "requested":true with "sessionLocked":false and "secure":false and never moves off lock-pending: screen-stabilizing, and every later lock request from any of those paths returns ok and does nothing, for hours. The screensaver stops appearing on idle as well, because it is gated on the same answer. Closing the lid then suspends the machine with the session unlocked, and the critical notification "Screen did not lock before suspend" is waiting on resume. The reporter on issue #10299 resumed into Hyprland's crashed-lockscreen failsafe rather than a password prompt, and it cleared on its own a few seconds later. A second reporter, whose shell had latched the same way after a stranded-lock recovery stalled, sat at an unlocked desktop for eight hours with isLocked reporting true throughout and two idle locks reported as exit 0.
Cause
Confirmed by reading /usr/share/omarchy/shell/plugins/lock/Service.qml on omarchy 4.0.2-1 and again on 4.0.4-1, byte-identical to the file at the v4.0.2 tag, commit 346e69e1cec6c4e8924531874af6ba010a1bc99e, and also byte-identical at the v4.0.3 and v4.0.4 tags. The shell's locked property at line 37 is true as soon as a lock is requested (lockRequested, set true inside beginLock() at line 135), before the compositor has confirmed anything. requestSessionLock(), lines 63 to 77, clears its own retry timer and marks the request done the moment it asks Hyprland for the lock, at lines 74 to 76, rather than when Hyprland confirms it. If that one request is declined or stalls, nothing retries it and nothing times out. From then on the lock IPC method sees root.locked already true and returns ok without doing anything, lines 513 to 516, and isLocked reports true off the same latched property, lines 519 to 521. The idle service gates the screensaver on that answer at shell/plugins/services/idle/Service.qml line 69 on 4.0.4 (line 68 on 4.0.2), so idling stops producing a screensaver too. Only two places ever clear lockRequested again, finishUnlock() at line 151 and the compositor's own onLockStateChanged at line 255, and both need a lock Hyprland has already confirmed, so a stall that happens before that confirmation cannot correct itself.
omarchy-system-lock, run by the lid switch through omarchy-system-lid-close (default/hypr/bindings/utilities.lua line 34) and by Super+Ctrl+L (line 126), discards the reply from the lock IPC at its line 8 and exits with the status of a trailing || true, so nothing downstream can tell a real lock from a latched one. omarchy-system-sleep-lock, run from the suspend path by omarchy-system-sleep-monitor, maps requested: true with secure: false to locking and deliberately leaves it alone, then gives up after its budget, 12 seconds on the shipped InhibitDelayMaxSec=15, and lets the suspend proceed, because a delay inhibitor is a timer rather than a veto and the alternative is a laptop that stays awake in a closed bag. The only signal on wake is a critical desktop notification.
Established in issue #10299, opened 2026-09-05 by a user unconnected to this corpus. A collaborator comment on the thread confirmed the latch line by line and produced the same status output on a test VM by simulating the stall, and a second reporter reached the same latched state through a stalled stranded-lock recovery. What makes the original request stall, a declined or slow response from the compositor, is not established there or here, and nobody has triggered it on demand. The issue is open as of 2026-09-19. Fixes have been proposed and none has merged: PR #11811 (2026-09-14) added an acquire timeout and was folded into #12064, and both were closed unmerged, #11811 on 2026-09-16 and #12064 on 2026-09-17. Their work continues in #12245, open and unmerged as of 2026-09-19. The 4.0.3 and 4.0.4 releases changed the shell's plugin and lock code around this file, keeping the lock service out of third-party plugins' reach and loaded across plugin reloads, but neither touched the latch. shell/plugins/lock/Service.qml on quattro at commit e38c1d1289252d2adb96372eeac48d02e489c5b7 (2026-09-19) has since gained video wallpaper handling and still carries the latch at the same three places, shifted to lines 45, 86 and 585.
Fix
No upstream fix has merged as of 2026-09-19. shell/plugins/lock/Service.qml is byte-identical at the v4.0.2, v4.0.3 and v4.0.4 tags. The fix PRs #11811 and #12064 were closed unmerged, and #12245, which carries their work, is still open. Until one lands, treat omarchy-shell lock isLocked, omarchy-shell lock status and a zero exit from omarchy-system-lock or omarchy-system-sleep-lock as unproven, and ask the compositor instead.
The visible test comes first: a lock that took shows a password field. If the desktop is still on screen after a lock request, confirm the latch from a terminal in the session:
omarchy-shell lock status | jq '{requested, sessionLocked, secure, lastEvent}'
omarchy-hyprland-session-locked; echo $?
requested: true with sessionLocked: false and secure: false that does not change within a second or two is the latched state. Exit 0 from omarchy-hyprland-session-locked means Hyprland itself holds an ext-session-lock. Exit 1 means it does not, whatever the shell says. Exit 2 means it could not tell, which happens when a monitor has no workspace yet or when hyprctl cannot reach the compositor, and is not a verdict either way.
From a TTY or over ssh hyprctl has no instance to talk to, so the check exits 2 every time. Give it the running instance first, the same way omarchy-restart-shell does, and run the omarchy commands from a login shell (bash -l) so OMARCHY_PATH is set:
export HYPRLAND_INSTANCE_SIGNATURE=$(ls -t "${XDG_RUNTIME_DIR:-/run/user/$UID}/hypr" | head -n 1)
omarchy-hyprland-session-locked; echo $?
If the compositor reports not locked (exit 1) while the shell claims otherwise, clear the latch by replacing the shell, then lock again:
omarchy-restart-shell
omarchy system lock
omarchy-restart-shell runs the same compositor check itself. On exit 1 it does not relock, so it only replaces the stuck shell with a fresh one whose lock service starts clean, and the omarchy system lock that follows is the real lock. A password field should now be on screen, and from a TTY or ssh omarchy-shell lock status should show "sessionLocked":true,"secure":true.
If instead the compositor reports locked (exit 0) and no password field is on screen, this is not the latch but a stranded lock: Hyprland is holding a lock whose client is gone. omarchy-restart-shell refuses while the shell still reports requested or secure, and relocks a fresh shell when it does not. That recovery, with its own risks, is the record plugin-file-write-while-locked-strands-session.
Verify
After a lock, from a TTY or over ssh as the session user, in a login shell:
export HYPRLAND_INSTANCE_SIGNATURE=$(ls -t "${XDG_RUNTIME_DIR:-/run/user/$UID}/hypr" | head -n 1)
omarchy-hyprland-session-locked; echo $?
omarchy-shell lock status | jq '.sessionLocked, .secure'
Exit 0 together with true and true means the compositor holds the lock and the shell's locker is live behind it, the only state in which the desktop is not exposed. Exit 1 or 2, or requested: true with both values false, means the session is not secured, whatever isLocked says.
Sources
- https://github.com/omacom/omarchy/issues/10299
- https://github.com/omacom/omarchy/blob/346e69e1cec6c4e8924531874af6ba010a1bc99e/shell/plugins/lock/Service.qml
- https://github.com/omacom/omarchy/blob/346e69e1cec6c4e8924531874af6ba010a1bc99e/bin/omarchy-system-lock
- https://github.com/omacom/omarchy/blob/346e69e1cec6c4e8924531874af6ba010a1bc99e/bin/omarchy-system-sleep-lock
- https://github.com/omacom/omarchy/blob/346e69e1cec6c4e8924531874af6ba010a1bc99e/bin/omarchy-hyprland-session-locked
- https://github.com/omacom/omarchy/blob/346e69e1cec6c4e8924531874af6ba010a1bc99e/bin/omarchy-restart-shell
- https://github.com/omacom/omarchy/pull/11811
- https://github.com/omacom/omarchy/pull/12064
- https://github.com/omacom/omarchy/pull/12245
- https://github.com/omacom/omarchy/blob/c668141e9c42b13c80c9ca4ea108e11708c5e8a5/shell/plugins/lock/Service.qml
- https://github.com/omacom/omarchy/blob/c668141e9c42b13c80c9ca4ea108e11708c5e8a5/bin/omarchy-system-sleep-lock
- https://github.com/omacom/omarchy/blob/c668141e9c42b13c80c9ca4ea108e11708c5e8a5/bin/omarchy-restart-shell
- https://github.com/hyprwm/Hyprland/blob/v0.56.2/hyprctl/src/main.cpp
- https://github.com/omacom/omarchy/blob/c668141e9c42b13c80c9ca4ea108e11708c5e8a5/shell/plugins/services/idle/Service.qml
- https://github.com/omacom/omarchy/blob/e38c1d1289252d2adb96372eeac48d02e489c5b7/shell/plugins/lock/Service.qml
- https://github.com/omacom/omarchy/compare/v4.0.2...v4.0.4