Hi, On 2025-01-04 20:47 +03:00, Andrea Bolognani wrote: > On Fri, Jan 03, 2025 at 01:07:30AM +0100, Andrea Bolognani wrote: >> On Wed, Dec 11, 2024 at 05:47:42PM +0300, Alper Nebi Yasak wrote: >>> Thanks for the explanation. Indeed, I can connect over ssh and run >>> `systemd-inhibit --what=shutdown --mode=delay sleep infinity` and get >>> the same behaviour, so it's definitely about the inhibit delay. >>> >>> I guess the login screen showing up during the delay is a different bug >>> somewhere else, maybe in systemd/logind...? >> >> Since, according to your initial message, you have managed to >> reproduce this with both GDM and LightDM, I would rule out a bug in >> the display manager. systemd sounds like a reasonable target for a >> new bug report.
I have already filed a feature request [1] to systemd to disallow logins while a power-off is triggered but inherited, and explained the behaviour I see here. [1] https://github.com/systemd/systemd/issues/35728 Arguably, each display manager should itself detect a power-off is in progress and refuse to show its login screen, but perhaps show a simple message about the system powering off. > After thinking about this a bit more, I'm no longer sure there's even > anything worth reporting. > > I've reproduced the behavior and what I could see is that, upon > selecting the "power off" option in GNOME, my session was > successfully terminated and I was kicked back to the GDM screen. > > Honestly, this sounds like the sanest behavior that could be expected > given the circumnstances. Once the libvirt fix is in, you'll no > longer experience it anyway. A bit contrived, but: imagine someone sets InhibitDelayMaxSec to 12 hours, "powers off" their computer, goes to sleep, wakes up, finds their computer powered on at the login screen, logs in and starts doing work... Only to get hit with an unexpected shutdown. I think it should show the Plymouth splash screen instead, similar to an ordinary power off i.e. plymouth-poweroff.service. Although on recent computers that might happen too fast for the splash to be actually shown. (Anyway, not a hill I'd die on. As you said it'll go away soon.)
