| *** WSalmon has quit IRC | 05:42 | |
| *** WSalmon has joined #buildstream | 05:42 | |
| nanonyme | benschubert, if host has lower hard limit, then that ulimit command will fail. So that should prevent it from ever running bst with lower nofile than that | 06:15 |
|---|---|---|
| *** tristan has joined #buildstream | 06:33 | |
| *** ChanServ sets mode: +o tristan | 06:33 | |
| nanonyme | Any plans for 1.6.1? | 06:35 |
| tristan | Not specifically, ideally we'd give it a bit of time to accumulate more fixes as long as this issue is not an emergency | 06:39 |
| nanonyme | Hmm, maybe it's not. I was just seeing it about all the time yesterday on ppc64le (was triggered through timeout in grpc) but it doesn't seem to be happening anymore | 06:42 |
| tristan | As I understand, it will still be an error, just that the error will be reported better after the fix right ? | 06:49 |
| nanonyme | If it's still an error, then that sounds like another bug :) Why would grpcio timeout fail the build? | 06:50 |
| tristan | Oh then I'm mixed up, I thought I recall the stack trace saying that this exception happened during the "too many open files" exception | 06:51 |
| tristan | (which doesn't make much sense logically, it's just what I thought I recalled seeing on the issue) | 06:52 |
| *** tristan has quit IRC | 07:18 | |
| *** santi has joined #buildstream | 08:54 | |
| *** tristan has joined #buildstream | 09:25 | |
| *** ChanServ sets mode: +o tristan | 09:25 | |
| *** cs-shadow has joined #buildstream | 09:59 | |
| douglaswinship | is there such a thing as an @buildstream email address? | 11:04 |
| douglaswinship | I think we ought to set up a GitHub account called "BuildStream-Migration-Bot", and it'll need to have an email address associated with the account. | 11:04 |
| juergbi | douglaswinship: I don't know about @buildstream.build email addresses, however, definitely +1 on a dedicated GitHub account for migration purposes. import shouldn't happen with an account of a real person | 12:01 |
| *** tristan has quit IRC | 13:03 | |
| nanonyme | tristan: In case you read logs later: the fix was a generic issue in error handling code. The first place that broke was about running out of fd's but I just noticed yesteday also case where timeout error from grpcio hit this same broken codepath | 14:29 |
| nanonyme | juergbi, just wondering, maybe bst1 should actually do resource.setrlimit(resource.RLIMIT_NOFILE, ...) if the current one is set too low | 16:01 |
| juergbi | we do already set the soft limit to the hard limit | 16:02 |
| juergbi | more is not possible without privileges, afaik | 16:02 |
| nanonyme | Oh, okay | 16:02 |
| nanonyme | Yeah, it's not | 16:02 |
| nanonyme | juergbi, I don't believe what you're doing works though | 16:02 |
| nanonyme | ulimit -n ... # did the trick. Running bst1 manually did not | 16:03 |
| nanonyme | Where's this code that you're talking about? | 16:03 |
| nanonyme | The code looks fine so the question is whether this code is running early enough (before any child processes have been forked | 16:05 |
| nanonyme | ) | 16:05 |
| nanonyme | https://gitlab.com/BuildStream/buildstream/-/blob/bst-1/buildstream/_frontend/app.py#L122 | 16:06 |
| nanonyme | Hm, this is really weird. I can't see any reason why this code wouldn't do the right thing | 16:14 |
| douglaswinship | :q | 16:53 |
| douglaswinship | (wrong window) | 16:53 |
| *** santi has quit IRC | 18:20 | |
| douglaswinship | :q | 19:10 |
| douglaswinship | (and again!) | 19:10 |
| nanonyme | douglaswinship, using ircii? | 19:10 |
| douglaswinship | nanonyme: nope. Irssi | 19:10 |
| douglaswinship | but either way, the ":q" command is supposed to go the vim instance i have open in a different terminal | 19:12 |
| gitlab-br-bot | BenjaminSchubert closed issue #824 (Scheduler requires significant rework) on buildstream https://gitlab.com/BuildStream/buildstream/-/issues/824 | 19:35 |
| nanonyme | benschubert, yay | 19:36 |
| benschubert | I almost forgot about it, fell on it while cleaning my open tabs | 19:37 |
| benschubert | thanks for going through the issues yesterday, they tend to accumulate and drift | 19:37 |
| nanonyme | Yes, I know, in freedesktop-sdk we're also soon going over 200 issues | 19:38 |
| nanonyme | Any idea what the plan is for BuildStream vs buildbox-casd compatibility? Will you be able to update either without worries in bst2 days? | 19:46 |
| benschubert | That's a good question, juergbi might be more up to date with that | 20:14 |
| gitlab-br-bot | nanonyme closed issue #1281 (Downloadable file source is not setting a user agent; rejected by gitlab) on buildstream https://gitlab.com/BuildStream/buildstream/-/issues/1281 | 21:20 |
| *** cs-shadow has quit IRC | 21:50 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!