16:02:21 <tardyp> #startmeeting weekly
16:02:21 <bb-supy> Meeting started Tue Jun 27 16:02:21 2017 UTC and is due to finish in 60 minutes.  The chair is tardyp. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:02:21 <bb-supy> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:02:21 <bb-supy> The meeting name has been set to 'weekly'
16:02:28 <tardyp> #topic introductions
16:02:41 <djmitche> here :)
16:02:46 <tardyp> whos here?
16:02:48 <rjarry> hi o/
16:02:54 <tardyp> #link agenda http://bit.ly/2rup31x
16:03:23 <skelly> I'm here
16:04:07 <tardyp> #topic week in review
16:04:26 <tardyp> #info lots of activity this week
16:04:39 <tardyp> #info grid_view plugin: https://nine.buildbot.net/#/grid
16:04:47 <tardyp> #info secrets can now populate files on a worker (e.g for ssh keys)
16:04:48 <djmitche> awesome!
16:04:57 <tardyp> #info first graphQL patch in buildbot!
16:05:10 <tardyp> accessing the GH v4 API
16:05:46 <tardyp> graphql is not the most consise langage I have ever seen..
16:05:54 <tardyp> #info gitpoller python3 fixes
16:06:11 <tardyp> plus a few fixes and random stuff
16:06:40 <tardyp> but we are having some bigger contributions from newer people so I am pretty happy with that
16:07:11 <djmitche> link to the graphql patch?
16:07:13 <djmitche> I'm curious :)
16:07:32 <tardyp> https://github.com/buildbot/buildbot/pull/3371
16:08:03 <tardyp> https://github.com/buildbot/buildbot/pull/3371/files#diff-991b6f7e16962c66edd9caaeeda02557R204
16:08:12 <tardyp> jinja generated graphql queries...
16:08:34 <rjarry> haha that's cruel
16:09:14 <tardyp> #topic release
16:09:44 <tardyp> I think the grid view plugin is by itself a good reason to release, as this is a highly awaited feature
16:09:58 <wiking> yey
16:10:03 <wiking> ++ for grid_Vew
16:10:15 <tardyp> rjarry is not quite done on it, so I'll start the release as soon as it is ready
16:10:43 <rjarry> tardyp: I'm all over it :-)
16:10:45 <noc0lour> oh hello (:
16:10:54 <tardyp> no presure :)
16:11:08 <noc0lour> love the progress on secrets
16:11:28 <tardyp> me too. Remy is doing a good job here.
16:12:20 <rjarry> btw I still have that unicode PR in the pipe
16:12:46 <rjarry> i saw multiple prs fixing unicode related problems
16:12:56 <rjarry> but only parts of the problems
16:13:19 <tardyp> it is easier to fix the py3 problems one by one
16:13:34 <rjarry> I really should finish writing tests for it
16:13:52 <tardyp> doing it in one PR results in never finished PR
16:14:13 <rjarry> :-)
16:14:15 <noc0lour> yeah step by step, and once you learn how to write tests for buildbot it gets quite easy
16:14:31 <tardyp> we have a bunch of stalled PR. I have seen some stalled PR bots here and there, I think we should setup one
16:14:31 <noc0lour> you can run them on your dev machine until they work ;)
16:15:05 <tardyp> #topic AOB
16:15:13 <tardyp> any other topic you want to discuss?
16:15:25 <rjarry> tardyp: what do you mean by stalled PR bot ?
16:15:38 <djmitche> something that comments on, and soon closes, PRs without any activity
16:15:47 <tardyp> right
16:16:21 <rjarry> ok, about that, there was a bot that commented about who was likely suitable for a review
16:16:24 <tardyp> we have a stalled label, but we put those manually, and this is painful
16:16:39 <tardyp> mentionbot
16:16:43 <tardyp> he is still here
16:16:44 <rjarry> yes
16:17:07 <rjarry> how come he did not react to my last prs ?
16:17:12 <djmitche> i don't see mentionbot comments on https://github.com/buildbot/buildbot/pull/3371 for example
16:17:13 <skelly> haven't seen mentionbot lately
16:17:20 <noc0lour> probably we should go through the PRs offline and either reassign them or ping the people manually
16:17:42 <noc0lour> someone said something about mentionbot not being in integrations anymore
16:18:15 <djmitche> maybe with recent changes to apps/integrations
16:18:19 * djmitche shrugs
16:18:47 <tardyp> yeah it is not anymore in our webhooks...
16:19:29 <noc0lour> in the worst case we probably could selfhost it or replace it with something else
16:19:55 <tardyp> I just put it back
16:20:13 <djmitche> yay!
16:20:20 <djmitche> want to #info that :)
16:20:47 <tardyp> #info mentionbot is back
16:20:56 <tardyp> anything else?
16:21:00 <rjarry> yep
16:21:04 <rjarry> pypy
16:21:19 <tardyp> #topic pypy
16:21:29 <rjarry> does anyone have experience of buildbot running on pypy ?
16:21:46 <tardyp> I do have some pypy docker images
16:21:47 <rjarry> I made some live tests last week
16:22:09 <tardyp> https://hub.docker.com/r/buildbot/buildbot-master-pypy/
16:22:21 <rjarry> with 500+ builders, 40+ workers and psycopg2cffi
16:22:34 <noc0lour> speed improvements?
16:22:49 <rjarry> I experienced sudden crashes
16:22:54 <tardyp> I did see improvments by using pypy ~20%
16:23:02 <rjarry> without any logs, nothing
16:23:22 <tardyp> interresting..
16:23:30 <rjarry> yeah...
16:23:36 <tardyp> rjarry: do you run it in docker or anything else?
16:23:48 <rjarry> I switched back to cpython in the meantime
16:24:02 <rjarry> in an LXC container
16:24:09 <rjarry> on PROXMOX
16:24:47 <tardyp> the thing is we dont have any running  unit tests on pypy
16:24:58 <tardyp> so I dont know what is the exact status
16:25:00 <rjarry> I did not investigate much on the problem, so that may be irrelevant
16:25:29 <tardyp> a sudden crash looks weird. I do remember see a segfault once with pypy
16:25:36 <rjarry> do we have tests on postgresql btw ?
16:25:42 <tardyp> we do
16:25:55 <noc0lour> could we extend the running unit tests to pypy ?
16:25:58 <tardyp> it is well tested and has lots of users
16:26:13 <rjarry> yeah that's weird, twistd.pid was still here but the pypy process was dead
16:26:36 <tardyp> noc0lour: we can. just need to enable it in our metabuildbot docker image and then make the test pass
16:26:37 <rjarry> so it looks like segfault or something similar
16:26:45 <tardyp> rjarry: yes
16:28:02 <tardyp> https://bitbucket.org/pypy/pypy/issues/2454/resumedatadirectreader_getvirtual_ptr
16:28:09 <tardyp> that crash was with profiler
16:28:32 <noc0lour> rjarry: you could open an issue
16:28:55 <noc0lour> and then we can continue discussion and testing there
16:29:14 <rjarry> +1 for pypy integration tests if that's feasible
16:29:32 <rjarry> (sorry if I did double post)
16:29:39 <tardyp> agree
16:29:55 <tardyp> there are 23 installids of a master with pypy
16:30:04 <tardyp> but only with very few builders
16:31:06 <tardyp> ok shall we wrap up?
16:31:26 <rjarry> what kind of stress tests could we setup to see if there is a problem with high numbers of builders & workers ?
16:31:44 <noc0lour> maybe have an integration test with a lot of builders/workers ;)
16:31:49 <rjarry> yup time's up :-)
16:31:58 <noc0lour> we can generate them inside python
16:32:41 <tardyp> #endmeeting