16:01:25 <djmitche> #startmeeting weekly
16:01:25 <bb-supy`> Meeting started Tue Sep 26 16:01:25 2017 UTC and is due to finish in 60 minutes.  The chair is djmitche. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:01:25 <bb-supy`> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:01:25 <bb-supy`> The meeting name has been set to 'weekly'
16:01:26 <tardyp> ah
16:01:32 <djmitche> just finished the last meeting :)
16:01:43 <djmitche> #topic introductions
16:01:49 <djmitche> http://bit.ly/2rup31x
16:02:09 <tardyp> who's around? djmitche tardyp bdbaddog1
16:02:17 <djmitche> quick reminder we're limited to 30 minutes
16:02:20 <tardyp> frodox: around?
16:02:33 <tardyp> I know that you were disapointed we missed last weeks..
16:02:39 <djmitche> also, apologies for missing last week
16:03:48 <djmitche> #topic Week(s) in Review
16:03:51 <djmitche> what's new?
16:04:47 <tardyp> With tr4nce , we debugged race condition with locks https://github.com/buildbot/buildbot/pull/3650
16:05:04 <tardyp> locks have been broken from 7 month
16:05:23 <tardyp> if you had more that 3 things holding the lock, then you would dead lock
16:05:48 <djmitche> #info debugged locks in https://github.com/buildbot/buildbot/pull/3650 - deadlock with more than 3 users of the lock
16:05:55 <tardyp> I had a few reports already, but it was so strange as we didn't change the locks code for years
16:06:19 <tardyp> and it happen that a change of buildstep base class (making it comparable) did break the locks code..
16:06:35 <djmitche> I think I wrote that Lock code.. it was not pretty
16:06:36 <djmitche> sorry about that!
16:07:07 <tardyp> well, it was missing some integration tests.
16:07:26 <tardyp> I spend a few hours writting those, because its hard to synchronise them
16:08:09 <tardyp> Also rodrigc did fiinish the long lasting task of having py2 vs py3 compat for worker master protocol
16:08:24 <tardyp> his PR is still pending
16:08:31 <skelly> he said he has both ways working now too
16:08:42 <tardyp> right
16:08:55 <skelly> will be great to include in next release
16:08:56 <tardyp> that was a hard task involving twisted upstreaming
16:09:04 <tardyp> skelly: definitly
16:09:09 <djmitche> #info rodrigc finished py2/py3 compatibliity in the worker/master protocol (PR stil pending)
16:09:32 <djmitche> should we transition to talking about the 0.8.x builder, then?
16:09:44 <tardyp> and also he is working with the python guys so that they upgrade their buildbot instance to Nine
16:10:19 <tardyp> and they had some request that we support 0.8.x slaves
16:10:34 <tardyp> we do support them, but we don't have integration tests
16:10:43 <djmitche> #topic 0.8.x worker for metabuildbot?
16:11:02 <tardyp> rutsky did the test once, but we don't have automation
16:11:22 <djmitche> was the request to *run* a 0.8.x worker against our metabuildbot, and test that way?
16:11:27 <skelly> yeah
16:11:35 <tardyp> one concern is that we can make py2 0.9 master to py2 0.8 slave work
16:11:47 <tardyp> but we can never make py3 0.9 master to py2 0.8 slave
16:12:04 <tardyp> so for my python will have to upgrade their slaves to workers
16:12:28 <bdbaddog1> buildbot.scons.org has 0.9.11 py2 master, and 0.9 and 0.8 workers/slaves
16:12:29 <tardyp> yes, his request was that we dogfood.
16:12:45 <djmitche> bdbaddog1: those workers are all py2 as well?
16:12:49 <tardyp> for me  dogfooding makes few sense I prefer to have integrations tests
16:12:51 <bdbaddog1> yes. all py2
16:13:05 <skelly> I think dogfooding looks good practically
16:13:25 <skelly> it's one thing to say we have tests for it, even if that's better to prevent regressions, but seeing it work day-to-day is believing
16:13:38 <tardyp> nine.buildbot.net is a bbtravis, and thus does not test much of the features
16:13:49 <tardyp> its basically mostly shellcommands and git
16:14:13 <tardyp> no transfer or stat command, the kind of stuff that he fixed recently
16:14:36 <skelly> maybe we should just have it run random perforce commands...
16:14:50 <tardyp> also we are working with hyper, and that will mean making a docker image with 0.8 worker on it.
16:14:51 <djmitche> haha
16:15:04 <djmitche> yeah, I'd forgotten about all of that
16:15:44 <tardyp> so practically, I think it will be less annoying to actually create the test
16:16:02 <tardyp> which is already quite annoying and difficult
16:16:12 <djmitche> how would that test work? download a 0.8.x package and install it?
16:16:27 <tardyp> right
16:16:49 <djmitche> that does sound pretty annoying
16:16:57 <tardyp> setup a venv for master, setup a venv for worker, and run a integration test which involve all remote commands
16:17:22 <djmitche> that's a lot just for compatibility
16:17:52 <djmitche> what about just setting up a docker compose file that could allow testing this arrangement?
16:18:11 <tardyp> we don't have 0.8 worker docker image
16:18:13 <djmitche> and testing a few things likely to be problematic, then writing down in a blog entry "yep, they work" with a link to the compose file in a gist?
16:18:44 <tardyp> for me bdbaddog1 reporting that it works for him is also "yep, that work"
16:18:45 <djmitche> just don't have it, or is that especially hard to create?
16:18:50 <djmitche> yeah, that's true
16:19:16 <djmitche> I don't think this warrants *continuous* testing - just one-off testing (or a few one-offs and 'works for me's) is enough
16:19:53 <tardyp> at the moment, I would say so
16:20:17 <tardyp> and we have craig that did works for me for py2 vs py3
16:20:22 <djmitche> #info rodrigc looking for "dogfooding" with a 0.8.x worker connected to the metabuildbot
16:20:46 <djmitche> yeah
16:21:56 <djmitche> so i think we're agreeing that we should test this through one-off testing and anecdotal evidence?
16:22:36 <djmitche> skelly: sound reasonable? bdbaddog1?
16:22:40 <tardyp> for now, I guess it is what I can do
16:23:10 <skelly> sure
16:23:14 <djmitche> ok
16:23:19 <djmitche> I don't necessarily mean *you* need to test it tardyp :)
16:23:49 <djmitche> #info dogfooding is hard and not that effective -- nine.buildbot.net is a buildbot_travis instance, and is using Hyper, so it's not a good testbed for this configuration
16:24:19 <djmitche> #agreed since this is a historical-compatibility issue, we can rely on manual testing an "works for me" stories to check it
16:24:25 <skelly> okay, I need to go
16:24:30 <djmitche> thanks!
16:24:37 <djmitche> #topic Open Invention Network
16:24:58 <djmitche> #info (just to wrap up) Software Freedom Conservancy is already signed on with OIN, so they were basically wasting our time asking for this
16:25:08 <djmitche> #topic Expert Twisted Book Chapter
16:25:30 <djmitche> #info book revision has been moved to a private repo, as not every book author was comfortable writing in public (open source, eh??)
16:25:53 <djmitche> I've added a bit more about LRUs and done some editing
16:26:06 <djmitche> #info warner offered some early BB history that we should fold in somehow
16:26:09 <djmitche> anything else to add?
16:26:29 <tardyp> I am not sure I understood your last remarks about how to do the review process
16:26:44 <tardyp> what was wrong with merging text?
16:28:08 <djmitche> oh, only that I don't want to get into editing the chapter before we've finished the first draft
16:28:24 <djmitche> and I had some commits already written on top of what they were commenting on
16:28:37 <djmitche> so having to resolve conflicts seemed like needless work
16:28:47 <tardyp> so you did address the comments but inside your yet to publish branch?
16:28:50 <djmitche> since they were commenting on something I'd already redrafted
16:28:53 <djmitche> yeah
16:28:58 <djmitche> I think I addressed them in that PR?
16:29:13 <tardyp> I've not looked at the details.
16:29:24 <djmitche> ok
16:29:26 <djmitche> I'll complain again
16:29:43 <tardyp> but this looks like we need one file per paragraph
16:29:53 <djmitche> haha
16:29:55 <djmitche> I hope not
16:30:19 <djmitche> anyway, I'll keep plugging away on the parts I said I'd write, but those are slowly getting finished
16:30:24 <tardyp> I haven't starting working on it, but I hope we won't step on each other toes
16:30:41 <djmitche> I think we had seleted different sections
16:31:18 <tardyp> right
16:31:38 <tardyp> lets see how that work in the next few weeks
16:32:13 <djmitche> yeah
16:32:15 <djmitche> ok
16:32:17 <djmitche> I tihnk we're out of time
16:32:22 <djmitche> anything else?
16:32:43 <djmitche> #endmeeting