17:00:00 <djmitche> #startmeeting weekly
17:00:00 <bb-supy> Meeting started Tue Nov 14 17:00:00 2017 UTC and is due to finish in 60 minutes.  The chair is djmitche. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:00 <bb-supy> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:00:00 <bb-supy> The meeting name has been set to 'weekly'
17:00:03 <tardyp> yeah!
17:00:07 <djmitche> NAILED IT
17:00:17 <djmitche> #topic Introduction
17:00:23 <djmitche> http://bit.ly/2rup31x -- agenda
17:00:45 <djmitche> #nick bdbaddog tardyp
17:00:54 <djmitche> rjarry: you're around too?
17:01:14 <Frodox> hi o/
17:01:44 <djmitche> hi!
17:01:47 <djmitche> ok
17:01:51 <djmitche> #topic Week In Review
17:01:54 <djmitche> take it away tardyp :)
17:01:59 <rjarry> yup, here o/
17:02:24 <tardyp> 0.9.13 was released
17:02:56 <tardyp> but that was already in next week's meeting
17:03:10 <tardyp> apart from that, pretty quiet
17:03:11 <djmitche> it bears repeating :)
17:03:18 <djmitche> #topic 0.9.13 is officially out
17:03:22 <djmitche> haha
17:03:26 <tardyp> I did look at the 1.0 milestones
17:03:29 <djmitche> #undo
17:03:29 <bb-supy> Removing item from minutes: <ircmeeting.items.Topic object at 0x806961390>
17:03:33 <djmitche> #info 0.9.13 is officially out
17:03:40 <tardyp> and there is only two issues yet
17:03:47 <djmitche> #msg chanserv topic #buildbot  A Software Freedom Conservancy Project | Buildbot-0.9.13 | docs: http://docs.buildbot.net/current/ | tutorial: http://docs.buildbot.net/current/tutorial | https://irclogs.jackgrigg.com/irc.freenode.net/buildbot | meetings: Tuesdays 1700 London Time -- http://bit.ly/2rup31x
17:03:47 <tardyp> two issues remaining
17:03:58 <djmitche> #info only two major issues remaining for a 1.0 release
17:04:01 <tardyp> I've created to PRs for each of those issues.
17:04:07 <tardyp> the badge one was quite easy
17:04:08 <djmitche> what defines 1.0?
17:04:09 <bdbaddog> link to those issues?
17:04:25 <Frodox> I have deja vu :D
17:04:27 <tardyp> https://github.com/buildbot/buildbot/issues?q=is%3Aopen+is%3Aissue+milestone%3A1.0.0
17:05:13 <tardyp> badge is ready to be merged https://github.com/buildbot/buildbot/pull/3744
17:05:25 <tardyp> worker control needs more work https://github.com/buildbot/buildbot/pull/3745
17:05:28 <djmitche> In the run-up to 0.9 I had thought of 1.0 as "OK, our APIs are stable"
17:05:31 <djmitche> like semver
17:05:33 <djmitche> is that still the case?
17:05:56 <tardyp> yes. We once defined 1.0 milestones as do we have regression against eight
17:06:03 <djmitche> ah, ok
17:06:09 <tardyp> so those are remaining milestones
17:06:28 <djmitche> #info 1.0 will mean stable APIs and possible regressions against 0.8.x
17:06:29 <tardyp> then API stability is probably something we either mostly have or will never have
17:06:34 <djmitche> yeah
17:06:40 <tardyp> depending on the definition of API stability
17:06:43 <djmitche> #info where "stable" means "as stable as we can get" :)
17:06:58 <rjarry> tardyp: I saw this https://nine.buildbot.net/#/bbtravis/deploy
17:07:06 <rjarry> is that a bbtravis plugin ?
17:07:10 <djmitche> so someone could conceivably build some other automation around the BB REST API and expect it to be compatible over several versions
17:07:23 <tardyp> I think since a few release we have few report of people having reggressions on their customization because of api breakage
17:07:44 <rjarry> about that, do we have a policy about master-worker api ?
17:07:45 <tardyp> rjarry: yes deploy is a feature of bbtravis
17:08:16 <tardyp> our general policy is to keep backward compatibility
17:08:59 <tardyp> i.e keep old worker work, and gently disable feature if it requires newer worker
17:09:18 <tardyp> #topic only one command per workerforbuilder
17:09:27 <rjarry> yeah, I was thinking about that :)
17:09:41 <tardyp> So we had a discussion with rjarry lately and discovered that you can only run one command per workerforbuilder
17:10:01 <tardyp> this is the kind of issue which requires change to master/worker protocol
17:10:13 <tardyp> this is required because the protocol does not have a handle for command
17:10:33 <tardyp> a command is associated to a wfb and you need to use the wfb to stop a previous command
17:10:35 <rjarry> while that is unfortunate, it is not a blocking issue, the main need was to be able to run a command outside of the context of a builder
17:10:49 <tardyp> if we want to implement say a parrallelShellCommand, we cannot
17:10:56 <djmitche> hm
17:11:12 <djmitche> is that an issue with the PB protocol. or more generally?
17:11:17 <tardyp> then there is the twisted rjarry usecase which wants to run command on the worker in nextWorker callback to detect if a worker can do a build
17:11:43 <tardyp> it is not related to pb, but rather out own protocol
17:11:49 <rjarry> "twisted" as in "crooked", not like the python lib
17:11:57 <tardyp> :)
17:12:31 <tardyp> twisted guys like to play with the double sense
17:12:46 <rjarry> say no more :)
17:13:25 <rjarry> tardyp: btw, I have double checked, and I cannot do this in nextWorker(). Even with caching
17:13:41 <tardyp> anyway imho, it will be hard to change our pb based protocol by keeping compat. I think this might be done in a new 2.0 iteration which probably should be based on websocket + amp
17:13:54 <djmitche> that makes sense
17:13:58 <tardyp> but that would be a fair ammount of work (not as big as nine though)
17:14:05 <rjarry> so, dropping pb altogether ?
17:14:07 <djmitche> I think dropping protocol compat would allow generation of a much simpler protocol
17:14:18 <rjarry> I like that :)
17:14:22 <tardyp> right
17:14:35 <rjarry> does that mean droping twisted too ?
17:14:39 <rjarry> dropping*
17:14:41 <tardyp> ahah no
17:14:51 <djmitche> #info current master/worker protocol does not support parallel execution of commands for the same builder
17:14:55 <tardyp> dropping twisted means rewrite buildbot entierely
17:14:57 <rjarry> well, this was worth trying :)
17:15:03 <rjarry> yeah
17:15:07 <djmitche> #info fixing that would probably require a breaking change (so, 2.0 in semver)
17:15:17 <rjarry> we could have a look at python 3 asyncio
17:15:22 <djmitche> it would mean having less involvement with PB, which is a good thing
17:15:28 <tardyp> I think we can probably implement a master with support with both protocol
17:15:32 <djmitche> my understanding is twitsed and asyncio are converging
17:15:46 <tardyp> but the pb one won't be able to use certain features like parrallel commands
17:15:54 <djmitche> makes sense
17:16:00 <rjarry> djmitche: where did you get this ?
17:16:15 <tardyp> I don't think that would be a good idea to make a 2.0 master dropping support with older worker
17:16:16 <djmitche> haha, I'm not sure
17:16:36 <rjarry> that would be actually cool
17:17:07 <tardyp> you can use asyncio with twisted
17:17:24 <tardyp> but I am not sure you can do the other way around
17:17:27 <rjarry> yeah that's what I gathered, it is just a backend
17:17:35 <rjarry> for twisted
17:17:40 <tardyp> maybe at somepoint we can drop py2. then we can convert everything to async await
17:17:47 <djmitche> woo!
17:17:55 <djmitche> ok
17:18:01 <djmitche> #topic Debian Packaging
17:18:07 <tardyp> it was already quite long and painfull to move from waitfordeferred to inlineCallbacks
17:18:22 <rjarry> I have contacted the debian packaging guys
17:18:28 <rjarry> bcc: tardyp
17:18:55 <djmitche> #info rjarry has contacted debian packaging folks
17:19:06 <djmitche> ok, that was quick :)
17:19:11 <djmitche> #topic Expert Twisted chapter
17:19:15 <rjarry> :)
17:19:18 <djmitche> I had some time to review the open PRs this weekend
17:19:21 <djmitche> I think they're good to merge
17:19:30 <djmitche> and I really appreciate the rewrite of some of my stuff :)
17:19:37 <djmitche> Frodox: anything to add here?
17:19:37 <tardyp> cool
17:19:56 <Frodox> not yet, I think :)
17:20:05 <tardyp> I need to take the reviews in account. I think we can merge them like that, and then edit the review in a followup PR
17:20:22 <tardyp> I won't be able to update it this week, probably
17:20:29 <rjarry> I need to run, kid@nanny
17:20:34 <Frodox> trying to keep whole model/chapter/book in mind
17:20:39 <rjarry> have a good one
17:20:45 <Frodox> rjarry: good luck !
17:21:24 <djmitche> tardyp: that seems like a good idea.  Should I click the "merge" buttons now?
17:21:30 <tardyp> ok
17:22:13 <djmitche> https://github.com/buildbot/expert-twisted/pull/7 has conflicts..
17:22:50 <djmitche> anyway
17:22:57 <djmitche> anything else on that topic .. or any other topic?
17:23:59 <tardyp> looks good
17:24:06 <tardyp> I fixed the conflict online
17:24:11 <djmitche> hah, so did I
17:24:31 <djmitche> feel free to click merge when it's green :)
17:24:34 <djmitche> ok
17:24:37 * djmitche wraps up..
17:25:37 <djmitche> #endmeeting