| slangasek | pitti: so I've done some debugging of #537262, and I'm fairly convinced that this is simply a race condition resulting from sendsigs firing before the plymouth job has *started* on shutdown; there's little enough left in /etc/rc6.d on a stock system that this is very possible | 02:56 |
|---|---|---|
| slangasek | pitti: I think the right thing to do is, inside the 'for seq' loop, call initctl list again to check for any new jobs that have popped up; this is still not entirely race-free, but it is a fairly cheap fix (cheaper than waiting 10 seconds on shutdown for no reason) and should catch the vast majority of cases - what do you think? | 02:59 |
| slangasek | (even though we've stopped the crash reports now, we're still hanging around ~10s longer on shutdown than we need to because of this) | 03:00 |
| === ArneGoet1e is now known as ArneGoetje | ||
| pitti | sbeattie: thanks | 06:54 |
| pitti | slangasek: that sounds plausible, at least much more than initctl having a bug of randomly not displaying jobs | 06:55 |
| pitti | slangasek: but what I don't understand is why plymouth would cause a 10 s hang then? doesn't it react to SIGTERM? | 06:57 |
| pitti | slangasek: i. e. if plymouth starts slightly late, then it wouldn't be on $OMITPIDS, and the killall5 would (try to) TERM it | 06:57 |
| pitti | which was the original effect described in the bug, I believe, that plymouth is killed prematurely | 06:57 |
| slangasek | pitti: it /can/ cause a 10s hang because it also races the invocation of killall5 - and indeed that's the case bug #537262 would correspond to, since if it had been caught by the first killall, the apport hook wouldn't have reported it | 07:00 |
| ubottu | Launchpad bug 537262 in plymouth "plymouth pid missing from OMITPIDS and terminated by sendsigs - Please no more confirmations/comments" [High,Confirmed] https://launchpad.net/bugs/537262 | 07:00 |
| pitti | slangasek: hm, but wouldn't at least the third or fourth one kill it? | 07:01 |
| slangasek | pitti: the third and fourth ones use SIGCONT, which won't kill the processes :) | 07:03 |
| pitti | oh, indeed | 07:03 |
| pitti | right, of course; /me grabs some tea to wake up | 07:04 |
| pitti | yes, that makes perfect sense now | 07:04 |
| nigelb | can someone give me an ack for a UIFe for pitivi, bug 314885 | 16:54 |
| ubottu | Launchpad bug 314885 in pitivi "Don't show version number in titlebar" [Low,Triaged] https://launchpad.net/bugs/314885 | 16:54 |
| nigelb | got blessings from translations and doc | 16:54 |
| kirkland | slangasek: is a new bash-completion file for a command line utility subject to FFe? | 20:35 |
| slangasek | kirkland: it's a feature, not a bugfix, so yes | 20:47 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!