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

Stop ttfx dumping core every time the screensaver is torn down by the lock

omarchyttfxlaptopdesktop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

Confirmed on this workstation: pacman -Qi ttfx reports 0.3.2-1 with build date Mon 17 Aug 2026, which is the date and version the record gives. The [omarchy] stable repo carries ttfx 0.3.2-1 today (fetched pkgs.omarchy.org/stable/x86_64/omarchy.db), so omarchy update is the right instruction. The cause is verified line by line in /usr/share/omarchy/shell/plugins/services/idle/Service.qml: lockSystem() at line 71 clears idledThisCycle at 76 and spawns omarchy-system-lock as a subprocess at 79, startIdleCycle() at 82 to 98 consults no lock state at all, and screensaverDelaySeconds is max(0, screensaver - min(screensaver, lock)) at lines 23 and 24, so it is 0 whenever screensaver <= lock and line 93 launches the screensaver immediately. The record's paraphrase of the collaborator triage is therefore more precise than the triage itself and is correct. /usr/share/omarchy/bin/omarchy-screensaver lines 9 to 14 send pkill -x ttfx and pkill -f '[o]rg.omarchy.screensaver' back to back with no wait, and lines 38 to 48 respawn a renderer as soon as the old one disappears, which are the other two routes the cause names. coredumpctl list ttfx --since '-1h' runs cleanly here, so the verify block is valid syntax. From sources: omacom-io/ttfx#18 "Stop dumping core when the terminal goes away" merged to master on 2026-08-17, and its diff matches the record exactly. It adds output_closed() treating EIO and BrokenPipe on tty output as the terminal going away and returning RunOutcome::OutputClosed, and it replaces println!/eprintln! with non-panicking outln!/errln! macros whose own doc comment reproduces the record's panic text and states that a release build aborts on panic. ttfx's Cargo.toml confirms [profile.release] panic = "abort", so the record's SIGABRT explanation is right for this binary rather than a wrong generalisation about Rust. Issue 6995 supports the symptom, the panic location, the three routes and the isLocked window. PRs 7131 and 7132 are both still open, checked today. One thing was wrong. Reported on Intel and AMD laptops and NVIDIA desktops is not supported: no NVIDIA report appears in issue 6995, or in 6762, 6764 or omacom-io/ttfx#10. What is attested is an Intel i7-4770HQ machine and a Framework 13 AMD laptop. ghostty is attested, but in issue 6762, which the cause names by number and the record did not cite, so 6762 and 6764 are added to sources along with the two open PRs. The unmerged-PR date was also refreshed from 2026-09-07 to 2026-09-11, which is the day it was rechecked. Not exercised: no crash was reproduced here. This machine already runs the fixed ttfx 0.3.2-1, and idling it into the lock to test the old behaviour is not possible read-only, so the before state and the one-core-per-monitor shape come from the reporters rather than from a local run. danger is left empty, which is right: the fix is omarchy update, the supported upgrade path, and the failure loses nothing but a core file.

The cause above was not rewritten and may still contain the error described. The fix below is the corrected version.

Symptom

On Omarchy 4.0.0-1 with ttfx 0.3.1-1, idling through the screensaver into the lock (or dismissing the screensaver) leaves a ttfx core dump behind and a "Process crashed" notification. coredumpctl list | grep ttfx shows SIGABRT entries, one per screensaver renderer, so one per monitor when the screensaver was up on all of them. The panic recovered from the core:


thread 'main' (7320) panicked at library/std/src/io/stdio.rs:1166:9:
failed printing to stderr: Input/output error (os error 5)

Reported on an Intel i7-4770HQ machine and a Framework 13 AMD laptop in issue 6995, with foot as the screensaver terminal, and on Ghostty 1.3.1 in the related issue 6762, all with idle: {screensaver: 150, lock: 300}.

Cause

ttfx, the screensaver renderer, is painting into a terminal that the lock script or the screensaver dismiss has already closed. Its next write fails with EIO, it reports that error with eprintln!, stderr is the same dead pty, and the report itself panics. Release builds abort on panic, hence SIGABRT and a core. Three routes into the same abort are documented across the thread and its linked issues: the lock script killing an already running screensaver (#6762, #6764), the screensaver supervisor loop respawning a renderer into a terminal that is being closed, and the one this issue reports, the idle service starting a fresh cycle and launching a second screensaver into a lock that is still being requested. The collaborator triage confirmed that last one in source: lockSystem() spawns omarchy-system-lock as a subprocess and about a second passes before lock-requested is set, startIdleCycle() consults no lock state, and with screensaver <= lock the delay is 0, so any idle re-assertion in that second launches a screensaver immediately.

Fix

The fix is ttfx 0.3.2 (omacom-io/ttfx#18, "Stop dumping core when the terminal goes away", merged 2026-08-17): a failed write to the tty is dropped instead of panicking, and EIO/EPIPE on the output tty ends the run quietly with exit 0. Omarchy's [omarchy] package repository carries ttfx 0.3.2-1 (built 17 August 2026, confirmed on this machine).


pacman -Q ttfx      # 0.3.1-1 is affected
omarchy update
pacman -Q ttfx      # 0.3.2-1

The Omarchy-side race, a second screensaver launched into a pending lock, is still open: PRs #7131 and #7132 are unmerged as of 2026-09-11. With ttfx 0.3.2 the terminal is still torn down abruptly, but the renderer exits instead of dumping core.

Verify

After the update, idle through the screensaver into the lock, unlock, then check that no new entry appeared:


coredumpctl list ttfx --since '-1h'

Sources

ttfx-screensaver-core-dump-on-lock