Skip to content

quickshell (Omarchy shell) aborts: "Tried to show lockscreen surfaces without active lock" #9727

Closed

Description

@lama7

Description

The Omarchy shell (quickshell) crashes with SIGABRT when lockscreen code attempts to show surfaces without an active lock session.

Trigger

Logged 2 seconds before crash:

FATAL: Tried to show lockscreen surfaces without active lock

Stack Trace (symbolized)

#0  raise (libc.so.6)
#1  abort (libc.so.6)
#2  QMessageLogger::fatal() (libQt6Core.so.6)
#3  QDebug::~QDebug() (libQt6Core.so.6)
#4  quickshell + 0x832b3 (omarchy shell QML)
#5  quickshell + 0x2240c2 (omarchy shell QML)
#6  QQmlPropertyPrivate::write() (libQt6Qml.so.6)
#7  QV4::QObjectWrapper::setProperty() (libQt6Qml.so.6)
... QML timer event \u2192 bound signal evaluation \u2192 property write \u2192 abort

System

Omarchy: 4.0.2
quickshell: 0.3.1 (Arch Linux)
Qt: 6.11.2
GPU: AMD amdgpu 3.64.0 (1002:15d8)

Reproduction

Unknown exact steps \u2014 occurred during normal desktop use. Likely related to lock/unlock cycle, idle timeout, or lid close/open triggering a race in lockscreen state management.

Impact

Shell auto-restarts (~1s downtime), no user data lost. Likely to recur under same conditions.

Debug Info

  • omarchy version output
  • omarchy debug --no-sudo --print (attaches /tmp/omarchy-debug.log)

Filed by nemotron-3-ultra-free via opencode.

Activity

  1. omarchybot commented on Sep 2, 2026

    @omarchybot
    Collaborator

    Triaged: your backtrace pins this to #6888, and PR #7169 is already up for the Omarchy half

    You did not have reproduction steps, but the stack is enough. This is #6888 (open, labelled bug), which has five machines confirming it and the same abort symbolised in a comment from 2026-08-14. #9654, #9709 and #9736 are the same bug too.

    What frames #4 and #5 are

    FATAL: Tried to show lockscreen surfaces without active lock is quickshell's own qFatal at src/wayland/session_lock.cpp:106, in WlSessionLock::updateSurfaces(), reached from realizeLockTarget():

    this->updateSurfaces(false);
    if (!this->manager->lock()) this->lockTarget = false;
    this->updateSurfaces(true, old);   // runs even though lock() just failed

    SessionLockManager::lock() (src/wayland/session_lock/session_lock.cpp:27) refuses when quickshell's process-global QSWaylandSessionLockManager::active is already set, and realizeLockTarget() ignores the refusal and shows surfaces anyway. That global is cleared only in QSWaylandSessionLock::unlock(); the destructor at session_lock/lock.cpp:16 calls destroy() and leaves it pointing at freed memory. Your "QML timer event → property write → abort" is Omarchy's side: one of the lock timers in shell/plugins/lock/Service.qml:434-446 fires requestSessionLock(), which writes sessionLock.locked = true at Service.qml:76.

    Why you could not name the steps

    That is the most useful thing about your report, and it has a concrete answer. The poison is process-wide and sticky. The event that causes it — Omarchy destroying and recreating its lock service while the compositor still holds the lock, which happened on every plugin hot-reload before commit d3d23fd — need not be anywhere near the crash. Once it has happened, the process aborts on the next lock: an ordinary lock/unlock cycle or idle timeout, possibly hours later, with nothing nearby in the log.

    Your instinct that it was "a race in lockscreen state management" is right, and the race is real rather than a figure of speech. SessionLockManager::isSecure() reads active->hasCompositorLock() off freed memory, so sessionLock.secure is undefined behaviour once a lock has been leaked: if it reads true, requestSessionLock() early-returns at Service.qml:64 and the session stalls silently instead; if it reads false, execution reaches line 76 and the process aborts. #6888 shows the same trigger producing an abort on one machine and a permanent silent stall on another for exactly that reason.

    Where the fix is

    If you can add one thing

    Not reproduced here — it needs a real compositor and a real session lock, so it would go on a disposable VM rather than the triage host. What would confirm which event poisoned your process is the journal for that whole boot rather than the two seconds before the crash: any Local plugin changed, reloading: line, any PluginRegistry: warning, and the earlier omarchy lock events. No need to chase it if the log is gone; the mechanism does not depend on it. Whether to close this in favour of #6888 is the maintainer's call.

    Reviewed twice: by Claude Opus 5, and independently by Codex at xhigh reasoning, which is what turned up PRs #7169 and #7572 and through them #6888 and #7106.

  2. bjarneo commented on Sep 21, 2026

    @bjarneo
    Member

    Automated duplication check: this issue looks similar to #9654, which covers the same Lockscreen surfaces abort without active lock. I keep that one open and close this one to keep the discussion in one place.

    If you feel this is the wrong decision, please open the issue again with a note on the difference.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions