16:30:30 <djmitche> #startmeeting weekly
16:30:30 <bb-supy> Meeting started Tue Jun 21 16:30:30 2016 UTC and is due to finish in 60 minutes.  The chair is djmitche. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:30:30 <bb-supy> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:30:30 <bb-supy> The meeting name has been set to 'weekly'
16:30:55 <djmitche> #topic Introduction
16:30:58 <tardyp> Hi everybody
16:31:02 <djmitche> Hi!
16:31:11 <djmitche> This is an open meeting, and anyone is welcome to participate
16:31:24 <djmitche> we have an agenda at https://titanpad.com/buildbot-agenda and it's easiest to participate by adding things at the end that you'd like to bring up
16:31:38 <djmitche> Unlike last week, I am actually here this week
16:31:52 <djmitche> as a roll-call, who else is here?
16:32:01 <bdbaddog1> bdbaddog online..
16:32:08 * tardyp waves
16:32:56 <bdbaddog1> :D
16:33:18 <djmitche> great :)
16:33:52 <djmitche> #topic Week in Review
16:33:53 <tomprince> I'm here.
16:34:04 <djmitche> awesome, thanks
16:34:12 <djmitche> it's good to know I'm not talking to nobody :)
16:34:40 <djmitche> looks like we have a compatibility issue reported
16:34:46 <djmitche> http://trac.buildbot.net/ticket/3567
16:34:56 <djmitche> #info buildbot nine master trigger traceback in old 'slave'
16:35:15 <tardyp> its harmless actually. just annoying
16:35:24 <tardyp> the "slave" is working as expected
16:36:09 <tardyp> rutsky is pretty busy those day. Hopefully I can't talk with him about it
16:36:13 <tardyp> I can
16:36:45 <djmitche> ok
16:36:57 <bdbaddog1> So ideally the master would know what version the "worker" is and not try to call non-existent method on slave?
16:37:37 <tardyp> at the moment the method to figureout which version it is is to test getWorkerInfo command
16:37:40 <bb-trac> [trac] #3565/undecided (new) updated by dustin (Were these the warnings in the tests?  I think those are harmless, and have been ...) http://trac.buildbot.net/ticket/3565
16:37:45 <tardyp> which does not exist in ancient slaves
16:38:12 <djmitche> we've used the approach before of swallowing the no-such-method exception
16:38:13 <tardyp> so this will expectedly fail on those, but this pushes a stack trace in the logs
16:38:17 <bdbaddog1> gotcha.. so the error indicates old and you proceed appropriately
16:38:18 <djmitche> sounds like a good fix
16:39:07 <djmitche> ok
16:39:07 <bdbaddog1> +1
16:39:13 <djmitche> rutsky: are you around for a 2340 update?
16:39:51 * djmitche guesses no
16:39:59 <djmitche> #topic Bug triage for 0.9.0
16:40:14 <djmitche> http://trac.buildbot.net/query?status=accepted&status=assigned&status=new&status=reopened&milestone=0.9.0&group=status&order=priority
16:40:20 <djmitche> #info 12 bugs open for 0.9.0
16:40:36 <tardyp> new tickets are from me
16:40:38 <djmitche> one of which is mine :)
16:40:48 <tardyp> and all pending PRs
16:41:09 <tardyp> which require some love
16:41:25 <djmitche> ok
16:41:29 <djmitche> there are three un-owned tickets:
16:41:47 <bb-trac> [trac] #3567/enhancement (assigned) updated by dustin (empty comment) http://trac.buildbot.net/ticket/3567
16:42:03 <djmitche> http://trac.buildbot.net/ticket/2995 in particular -- getting rid of the unwired elements
16:42:19 <tomprince> #3336 doesn't look like it is a regression?
16:42:36 <bb-trac> [trac] #3565/undecided (new) updated by tardyp (with 2270, the code retry the requests undefinitly, making it very slow to just ...) http://trac.buildbot.net/ticket/3565
16:42:42 <tomprince> So no reason to block 0.9.0 on it.
16:43:02 <bb-trac> [trac] #3564/defect (new) updated by dustin (I think this would eventually be flushed when the "old" master fails to check in for ...) http://trac.buildbot.net/ticket/3564
16:43:18 <tardyp> as far as I know reconfig was working in eight for all the elements
16:44:09 <djmitche> I'd be comfortable punting that to 0.9.1 anyway
16:44:39 <tardyp> for me this is important feature. for example buildbot_travis will not work with that
16:44:44 <tomprince> It looks like it doesn't even support changing passwords.
16:45:07 <djmitche> ah, ok
16:45:37 <tomprince> That doesn't seem like a reason to block 0.9.0? We can just release a 0.9.1 when that lands.
16:45:37 <tardyp> as the users can reconfigure the workers, and the builders, and everything, it is important that reconfig is working as expected.
16:46:35 <tardyp> ok. I just want this release to be as good as possible :)
16:46:56 <tardyp> those last reconfig bugs are the last big things indeed.
16:47:09 <djmitche> ok
16:47:14 <tardyp> and they are complicated, so I need to find big chunk of free time in order to work on it
16:47:56 <bb-trac> [trac] #3336/undecided (assigned) updated by dustin (This still exists in master and represents a regression.  It needs some quality time ...) http://trac.buildbot.net/ticket/3336
16:48:08 <tardyp> we can talk about rc1 then, because there is nothing more in the backlog
16:48:24 <tardyp> all the remaining have pending PR, or are very easy last minute checks
16:48:36 <tomprince> It isn't a regression. eight only compares the classes of the slaves. nine will actually update the password.
16:49:09 <tardyp> tomprince: ok I though that the whole stuff was just swapped in eight
16:49:24 <tardyp> without real reconfig
16:49:26 <tomprince> tardyp: Only if the classes don't match.
16:49:35 * tomprince just went and looked at the code.
16:49:44 <tardyp> I belive you :)
16:50:07 <tardyp> lets go ship it then!
16:50:13 <djmitche> hehe
16:50:22 <djmitche> tomprince: can you comment the bug and re-milestone?
16:50:27 <tomprince> Sure.
16:51:30 <tardyp> I'll fix the remaining stuff, and ship rc1 this week end. sounds good?
16:51:44 <bb-trac> [trac] #3336/undecided (assigned) updated by tom.prince (This isn't actually a regression from eight. nine will actually update passwords, ...) http://trac.buildbot.net/ticket/3336
16:51:50 <djmitche> sounds great!!
16:52:03 <djmitche> #info should have a release candidate out this weekend, if all goes well
16:52:14 <djmitche> ok, up next..
16:52:18 <djmitche> #topic     http://buildbot.buildbot.net doesn't report to IRC
16:52:30 <djmitche> I think this means bb-meta?
16:52:33 <djmitche> is it just muted?
16:52:36 <djmitche> bb-meta: unmute
16:52:36 <bb-meta> You hadn't told me to be quiet, but it's the thought that counts, right?
16:52:40 <djmitche> aww
16:53:18 <tardyp> I took the action last week to fix the bug
16:53:28 <tardyp> about ramlfications
16:54:09 <bb-trac> [trac] #3568/task (v:master) created by tardyp (fix metabuildbot) http://trac.buildbot.net/ticket/3568
16:54:19 <bb-trac> [trac] #3568/task (assigned) updated by tardyp (empty comment) http://trac.buildbot.net/ticket/3568
16:54:56 <skelly> checking the log, there are a lot of failed login attempts by metrohm-win81
16:55:42 <djmitche> ok, I'll count that as moved to a bug then
16:55:49 <djmitche> #topic     Supporting upgrades from 8.x to 9.0.
16:56:14 <tomprince> There are still some tests for upgrading from old versions, but they don't support loading data from pickles. Is there any point in keeping that code around?
16:56:36 <tomprince> Is upgrading from 0.8 something we plan to support?
16:56:50 <tomprince> I think the answer is no.
16:56:53 <djmitche> We support it, but it means losing all of the pickle data (which is all of the interesting stuff0
16:56:55 <tardyp> This is something we plan to accept patches
16:56:59 <tomprince> But we sort of do, currently.
16:57:17 <tardyp> but I won't proactively work on it, as I think there is few ROI
16:57:18 <djmitche> I think we intend 'buildbot upgrade-master' to work
16:57:24 <djmitche> just at the cost of a lot of data loss
16:57:28 <tomprince> I think we should drop all the code to support upgrading, since we don't preserve any of the interesting data.
16:57:31 <djmitche> so those tests are testing intended functionality
16:57:44 <djmitche> scheduler state, change-source state, etc. are preserved
16:57:49 <djmitche> there's some value there
16:57:54 <tardyp> that is a good thing to do indeed.
16:57:55 <tomprince> Some of the code is to test loading changes from pickles.
16:58:02 <tardyp> we can just fail if the db is too old
16:58:10 <tardyp> its easy
16:58:15 <tardyp> and we remove those old tests
16:58:48 <tomprince> Do we care about upgrading from anything pre-0.8.12?
16:59:02 <tardyp> I think it makes sense, and will prevent mistakingly update with pip install -U buildbot
16:59:12 <djmitche> that's a good point
17:00:57 <bb-trac> [trac] #3569/task (v:master) created by tardyp (fail when buildbot upgrade-master is run from a <=0.8.x installation) http://trac.buildbot.net/ticket/3569
17:01:02 <djmitche> ok!
17:01:15 <djmitche> #agreed we will not support automated updates from <0.9.0
17:01:26 <tomprince> So, a related question, then.
17:01:26 <djmitche> #info see http://trac.buildbot.net/ticket/3569
17:01:43 <djmitche> #info Should the initial database still be created by upgrading from the begining of time, or could we simplify that to start with a database at some recent version?
17:02:03 <tardyp> there is already a feature to do that
17:02:21 <djmitche> I think it just uses MetaData.create_all(), right?
17:02:27 <tardyp> indeed
17:02:39 <tardyp> in model.py
17:03:24 <tardyp> https://www.irccloud.com/pastebin/QNqvHk7e/
17:03:30 <bdbaddog1> so no updates would imply just start with blank database right?
17:03:53 <tardyp> yes
17:04:17 <tardyp> you can backup your master dir, remove the db, and upgrade-master
17:04:43 <tomprince> Given that and the fact that we aren't supporting 0.8.12, we can drop old migrations then.
17:05:40 <djmitche> true!
17:05:48 <tardyp> yes
17:05:54 <djmitche> #info answer: this is already supported
17:06:03 <djmitche> tomprince: is that someting we should do before 0.9.0rc1?
17:06:14 <tomprince> Doesn't really matter, I don't think.
17:06:50 <tomprince> In any case, I think I'll get a chance to look at that this week.
17:06:55 <djmitche> ok, great
17:07:08 <djmitche> #topic Upgrade flake8
17:07:09 * tomprince was looking at testing against pypy and noticed some failures due to loading pickles .... :)
17:07:13 <djmitche> ahh
17:07:21 <djmitche> I'll be hapy to see the backside of pickles
17:07:23 <djmitche> *happy
17:07:37 <djmitche> so it looks like there's a flake8 PR already, and I just labeled it merge-me, so .. anything more to discuss?
17:07:38 <tomprince> #info https://github.com/buildbot/buildbot/pull/2280
17:07:53 <tomprince> I was wondering about incrementally adding some plugins.
17:08:06 <tomprince> https://pypi.io/project/flake8_tuple/ for example
17:08:30 <tomprince> And maybe https://pypi.io/project/flake8-commas/ (but there are currently some false-postivies with the later.
17:08:55 <djmitche> sounds good
17:09:04 <djmitche> although flake8_tuple doesn't have any docs.. aren't 1-tuples useful sometimes?
17:09:25 <tomprince> Yeah, but you can mark them explicitly as (thing,) when you need them.
17:09:36 <djmitche> oh
17:09:39 <tardyp> ah
17:09:40 <djmitche> so this finds x = 10,
17:09:41 <tomprince> I've often accidentally got them when copying something out of a list or keyword argument.
17:09:43 <djmitche> but not x = (10,)
17:09:47 <tomprince> Yeah.
17:09:47 <djmitche> right
17:09:52 <djmitche> that sounds great - I've done that before too :)
17:10:06 <tardyp> me too
17:10:07 <djmitche> OK, let's do it! :)
17:10:16 <djmitche> #topic 0.8.x release?
17:10:29 <djmitche> #info sa2ajj is officially the eight maintainer
17:10:34 <djmitche> so I think this would be up to him
17:11:53 <tomprince> I think there should probably be at least one last release with all the fixes that have landed.
17:11:53 <djmitche> we've said pretty firmly that there won't be another release
17:11:59 <djmitche> but not as firmly as guido has said "no" to 2.8
17:12:06 <djmitche> I'd be comfortable with that
17:12:10 <tomprince> Oh. I wasn't around for that.
17:12:14 <djmitche> do you want to start a conversation on devel@ about it, copying sa2ajj?
17:12:29 <tomprince> I can do that.
17:12:34 <skelly> it may prove useful to have another eight release to ease transition
17:12:36 <tardyp> also, the eight unit tests have been broken for a while
17:12:43 <tardyp> because of dependencies upgrade
17:12:57 <tardyp> so this is something that shal be taken care of before release
17:13:11 <djmitche> awesome
17:13:17 <djmitche> ok, last topic
17:13:37 <djmitche> #action tomprince will start a conversation about 0.8.13 on devel@
17:13:40 <djmitche> #topic ftp and docs ansible failures
17:13:53 <djmitche> these hosts are failing to run ansible.  I haven't done much but look at the emails and delete them
17:14:29 <djmitche> fatal: [localhost]: UNREACHABLE! => {"changed": false, "msg": "ERROR! Authentication or permission failure. In some cases, you may have been able to authenticate and did not have permissions on the remote directory. Consider changing the remote temp path in ansible.cfg to a path rooted in \"/tmp\". Failed command was: ( umask 22 && mkdir -p \"$( echo
17:14:29 <djmitche> $HOME/.ansible/tmp/ansible-tmp-1466261366.21-266510462889082 )\" && echo \"$( echo $HOME/.ansible/tmp/ansible-tmp-1466261366.21-266510462889082 )\" ), exited with result 1", "unreachable": true}
17:15:28 <djmitche> I guess the obvious answer is "well OK dustin why don't you look into that?" :)
17:16:23 <skelly> is there a bug about it?
17:16:44 <tomprince> The obvious answer to the obvious answer is that you want to move away from dealing with buildbot infra.
17:18:18 <djmitche> heh, very perceptive
17:18:22 <djmitche> yeah, I'll file a bug :)
17:18:37 <djmitche> #action djmitche to file a bug about ansible failures
17:18:44 <djmitche> any other business?
17:18:54 <tomprince> #info Agenda from this meeting: https://titanpad.com/ep/pad/view/buildbot-agenda/7RzHHjC1xi
17:19:05 <djmitche> oh, good thinking
17:19:11 <djmitche> #endmeeting