- Dikirim
- 23 September 2026 pukul 19.23 UTC
- Penulis
- Kamo
- Commit
- 06f8b1c
The first computer reported its idle time as 13.9, then 29.2, then 14.4 seconds over three minutes while nobody was touching it. That is not a person — it is a sawtooth capped by the thirty-second poll interval, and it is what poll() does when no display will answer: unreadable counts as ACTIVE and last_active is pushed to now on every pass. The agent runs as root and the X display belongs to the member, which is not something X is relaxed about. xprintidle was invoked with DISPLAY and no XAUTHORITY, so the display refused it, it exited non-zero, and the return value was None. Every poll. Forever. So the idle timeout could never fire on any computer. Hibernation readiness was the bug in front of it; this one was behind it, and would have kept every machine awake even once readiness came true. The cookie is read off the Xorg process that serves the display, via `-auth` on its command line. xorgxrdp passes that RELATIVE to the session's home, so a bare path is resolved through the process's own /proc/<pid>/cwd rather than guessed from the username — the process already knows, and a home directory that is not under /home would have made a guess wrong. Unreadable is no longer silent, and that matters more than the fix: it pins a computer awake, so it now says which display refused, what xprintidle printed, and which XAUTHORITY was tried — once per display and reason, not once per poll. stderr was going to DEVNULL, which is why three separate sessions of work never saw it. Five tests, including the relative-path resolution and the log-once behaviour.