=== Pici is now known as TalkingMuffin | ||
brainwash | bug 1234469, once again all the duplicate reports have been removed... what up with that? | 10:28 |
---|---|---|
ubot2 | Launchpad bug 1234469 in linux (Ubuntu) "Network does not come up after resuming from suspend." [Medium,Incomplete] https://launchpad.net/bugs/1234469 | 10:28 |
brainwash | sorry, bug 1184262 is the main one | 10:28 |
ubot2 | Launchpad bug 1184262 in systemd (Ubuntu) "[logind] stuck in PrepareForSleep, causing network and other services to not resume" [Undecided,Confirmed] https://launchpad.net/bugs/1184262 | 10:28 |
penguin42 | brainwash: Hmm I could almost kind of understand that on the 'linux' one, because we try to avoid dupes on the 'linux' bugs unless we're 110% sure it's the same problem | 10:47 |
brainwash | penguin42: many have been converted to linux ones I think | 10:48 |
brainwash | it's a mess | 10:48 |
penguin42 | brainwash: Nod, tbh I'd have picked Linux as the source for a suspend/resume network problem; but 1184262 seems to be actively working it on the systemd/etc side of things so it does look it's being worked | 10:49 |
brainwash | for many people is was clearly a network-manager issue in the first place | 10:54 |
penguin42 | well, you say that but that's not where the debug process is going in that bug | 10:54 |
brainwash | it's just a bit annoying if someone messes around with reports | 10:54 |
brainwash | right, because people are not familiar with systemd yet | 10:55 |
brainwash | and most confirmed that the network-manager could be re-enabled manually | 10:56 |
penguin42 | Yeh it's a little odd, if there weren't any work around on it and it was just network-fails-after-suspend I would have put it against linux as well, but I agree with the bugs that show the workaround it does seem unlikely to be the kernel | 10:57 |
brainwash | but removing all duplicates without stating a reasons seems a bit wrong | 10:57 |
penguin42 | yeh should always state reasons for doing something | 10:58 |
brainwash | I'll try to contact him and ask about this, so I can understand why he did this | 11:00 |
brainwash | penguin42: thanks for clarifying some things :) | 11:00 |
hggdh | brainwash: pitti has been working on this bug. If he replaced n-m by systemd-shim it is because he believes systemd-shim is the actual guilty party | 13:16 |
=== Ursinha is now known as Ursinha-afk | ||
brainwash | hggdh: I know that, because I'm bothering him all the time with my observations and log files :) | 14:59 |
hggdh | heh | 14:59 |
=== Ursinha-afk is now known as Ursinha | ||
brainwash | and that's the reason I actually care about the removal of all the duplicates | 15:06 |
hggdh | brainwash: I am assuming that pitti removed them (I do not have details in the log). If he indeed removed them, it is because he considered them *not* the same issue. | 15:32 |
brainwash | hggdh: very unlikely, because it is the same issue in most cases, some reports don't provide enough details, but describe the same issue (saucy, resume does not wakeup network-manager and/or session is stuck in sleep state) | 17:01 |
brainwash | pitti's updated package does fix many cases of this issue, but we still investigate if the remaining people are eventually affected by another dbus timeout (systemd also) | 17:04 |
=== Ursinha is now known as Ursinha-afk | ||
brainwash | and some bug reporters have already marked their report as duplicate once again | 17:06 |
=== Ursinha-afk is now known as Ursinha | ||
hggdh | darn! I *hate* when people change bug status without explaining why. | 17:39 |
penguin42 | nod | 17:43 |
hggdh | OK. comment added on bug 1187005, which has seen a small flurry of wrong duplicates | 17:45 |
ubot2 | Launchpad bug 1184262 in systemd (Ubuntu) "duplicate for #1187005 [logind] stuck in PrepareForSleep, causing network and other services to not resume" [Undecided,Confirmed] https://launchpad.net/bugs/1184262 | 17:45 |
penguin42 | hggdh: I have a little sympathy that without noticing the work around people found, I'd have also guessed at 'linux' being the package for a suspend/resume bug, and once you've set it to linux you don't do dupes | 17:46 |
hggdh | penguin42: indeed. But, if Linux is not the correct package, undupping & dupping to a different bug strikes me as weird | 17:46 |
hggdh | brainwash: at least 1234469 was set as a dup to your bug by pitti himself... | 17:48 |
hggdh | penguin42: also, I would expect people to read the comments *and* look at the activity log. Although, I concede, if one does not know the other players, it get difficult to value their knowledge. | 17:52 |
* hggdh is now thinking this also applies to self... | 17:52 | |
penguin42 | hggdh: tbf I don't think the bugs actually been fixed yet (or hadn't yesterday) | 17:52 |
hggdh | I agree. Martin's fix does not get all, or there are different issues at play | 17:53 |
hggdh | and I am pretty sure that, right now, I do not want to look at systemd, it is rather confusing, docs lacking, API seems to change suddenly, etc | 17:56 |
penguin42 | hggdh: Yeh I mean it sounded like a hideous interaction of systemd/nm and a bunch of other stuff - wth knows who dropped the ball | 17:58 |
hggdh | <nod/> | 17:59 |
brainwash | it's not actually systemd's fault, but ubuntu's own wrapper for some dbus calls -> systemd-shim | 18:02 |
brainwash | systemd-shim quits after 10sec (that too fast), this has been fixed by pitti's new package | 18:03 |
hggdh | brainwash: I understand. But the fact we *need* a translator (shim) to interact with systemd says something is blowing out known APIs | 18:04 |
brainwash | well, we had to replace consolekit (deprecated) with logind (part of systemd) | 18:05 |
brainwash | and we don't ship every part of systemd, because it would replace upstart | 18:06 |
hggdh | indeed. But then systemd is getting to be a monolithic beast -- replacing a lot of known/used programs with its own, and incompatible with a lot (or everything else). | 18:07 |
penguin42 | brainwash: rule 24: Any fixed timeout is wrong | 18:07 |
brainwash | it seems that the 10sec timeout causes trouble, if the system is slow at suspending | 18:09 |
penguin42 | brainwash: Anything like that is a nightmare, things like suspend/resumes are really random times on different hardware | 18:09 |
penguin42 | brainwash: No timeout value will ever be right | 18:09 |
hggdh | indeed. Or, at least, fail gracefully | 18:10 |
brainwash | true | 18:10 |
=== wylde_ is now known as Guest32821 | ||
brainwash | 13.10 still feels somewhat broken and unpolished | 18:11 |
hggdh | well, this is sort of expected, 13.10 is a preparation for 14.04 | 18:12 |
hggdh | (so we could *sort* of look at it as a beta) | 18:12 |
brainwash | yes :) | 18:12 |
penguin42 | tbh I think it's time to throw the towel in and use systemd - it seems to work | 18:14 |
brainwash | canonical won't like this | 18:15 |
penguin42 | indeed, but then users don't like bugs like this | 18:15 |
=== Ursinha_ is now known as Ursinha | ||
=== rsalveti_ is now known as rsalveti | ||
hggdh | all in all, it will be a pity -- I like the idea of using events | 18:27 |
* hggdh remember the happiness when learning petri nets a long time ago | 18:30 | |
penguin42 | well nowt wrong with events as long as you never lose any | 18:32 |
hggdh | and you have a clear description of all events | 18:32 |
penguin42 | hggdh: It's like if I eat one of the M&Ms/smarties you were using to keep track of your petri net | 18:33 |
hggdh | LOL | 18:33 |
hggdh | indeed. And then try to prove you reach a final state... | 18:34 |
penguin42 | hggdh: maths/formal method stuff? | 18:34 |
hggdh | yes | 18:35 |
penguin42 | hggdh: I spent some time doing clock-less CPUs, there were a bunch of formal guys trying to do proofs against deadlock | 18:35 |
hggdh | faded to failure, IIRC | 18:36 |
hggdh | might still be possible if we had finite state (sub)machines; we might be able to prove the machine eventually enters a final state (which would be input to the next machine) | 18:37 |
=== wylde is now known as Guest77372 | ||
penguin42 | it's very difficult - the complexity just explodes as soon as you do anything useful | 18:45 |
=== maxb_ is now known as maxb |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!