16:30:26 <djmitche> #startmeeting weekly
16:30:26 <bb-supy`> Meeting started Tue Feb 16 16:30:26 2016 UTC and is due to finish in 60 minutes.  The chair is djmitche. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:30:26 <bb-supy`> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:30:26 <bb-supy`> The meeting name has been set to 'weekly'
16:30:32 <tardyp> hi all
16:30:35 <djmitche> #topic Introduction
16:30:35 <koobs> hi all
16:30:36 <djmitche> hi!
16:30:42 <djmitche> https://titanpad.com/buildbot-agenda
16:30:43 <bdbaddog1> Hi!
16:30:43 <infobob> http://paste.pound-python.org/show/GMgYgrOLAsTFCw8FwNIZ/ (repasted for djmitche)
16:30:44 <rutsky> hi!
16:31:13 <djmitche> Quick reminder: anyone's welcome to participate here; if you want to add a topic, just put it on the etherpad I just linked
16:31:28 <djmitche> #topic Buildbot-0.9.0beta7 just released
16:31:43 <rutsky> any known issues with beta7 UI?
16:31:46 <djmitche> #info Pierre shipped a new beta yesterday, fixing some UI issues
16:32:00 <rutsky> bdbaddog1: do you find source of your issues with UI?
16:32:49 <bdbaddog1> Nope. Every now and then when I click on waterfall I get a blank page which never populates and has no spinning wheel.
16:33:21 <rutsky> this should be reported somehow...
16:33:21 <djmitche> anything in the browser console?
16:33:26 <rutsky> I can't reproduce it
16:33:28 <bdbaddog1> nope.
16:33:39 <djmitche> yeah, I wonder if browsers have a way of intercepting errors that otherwise end up on the console
16:33:43 <bdbaddog1> it was blank until force reload. and since then I can't get it to come back.
16:33:45 <koobs> djmitche: Would you like buildbot ports/packages updated to 0.9.0b* on FreeBSD ?
16:34:00 <djmitche> koobs: no, they're still betas and best not deployed to the masses
16:34:00 <rutsky> bdbaddog1: have you tried to clear browser cache? e.g. reload with ctrl+shift+r?
16:34:15 <koobs> djmitche: OK. Is there an ETA for RC's ?
16:34:35 <djmitche> koobs: I think they're really close now - the betas are squashing last-minute things one by one
16:34:43 <djmitche> no more major dev work, really
16:34:47 <tardyp> for me next beta is with worker stuff
16:34:51 <tardyp> and then rc
16:35:06 <koobs> djmitche: Can you action item to heads-up me (and/or other OS packagers) about the forthcoming RC's. Perhaps on the mailing lists?
16:35:39 <koobs> djmitche: Only if you think it's valuable (it is for me, so i/we can ensure our users get fresh stuff and get you feedback if you need it
16:36:00 <djmitche> #action djmitche to send heads-up message about pending RCs for OS packagers on the mailing lists
16:36:07 <koobs> Gracias
16:36:14 <djmitche> y tu
16:36:41 <djmitche> I'll take another sweep through the 0.9.0 milestone
16:36:48 <djmitche> http://trac.buildbot.net/query?status=accepted&status=assigned&status=new&status=reopened&milestone=0.9.0&col=id&col=summary&col=status&col=type&col=priority&col=milestone&col=version&order=priority
16:37:10 <djmitche> I think we can just admit that a lot of that is 0.9.1 :)
16:37:18 <djmitche> #topic GSoC 2016
16:37:24 <tardyp> koobs: note that the fact that buildbot is building his own coffeescript into javascript does not help packaging
16:37:27 <koobs> action item, clean up the milestone list for 0.9.0 ;)
16:37:42 <koobs> tardyp: releases will contain processed assets, no ?
16:37:43 <tardyp> I know that the debian maintainers were not very happy about it
16:37:45 <djmitche> #action djmitche to triage 0.9.0 into 0.9.1
16:38:15 <tardyp> debian has a policy were they want to be able to rebuild everything from source
16:38:20 <tardyp> inside their system
16:38:29 <koobs> thats expected/normal
16:38:35 <koobs> release <> development
16:38:46 <tardyp> so this needs to pull the node.js/npm dependencies into the build system
16:38:53 <koobs> oh wut.
16:38:56 <koobs> no.
16:39:00 <rutsky> tardyp: what is expected time that mentor should allocate to work with student per week?
16:39:06 <tardyp> they wanted to make a deb package for every npm package needed to do the build
16:39:13 <koobs> tardyp: yes, so will we
16:39:16 <djmitche> ouch!
16:39:22 <koobs> tardyp: npm/node is a NIGHTMARE of vendoring.
16:39:28 <tardyp> indeed
16:39:44 <koobs> that is close to, if not a deal breaker
16:39:57 <koobs> buildbot ceases to be a python package at that point
16:40:03 <koobs> and instead a product.
16:40:05 <tardyp> right
16:40:07 <koobs> 'so to speak'
16:40:49 <koobs> yeh, i can confirmn, thats huge and unwelcome news
16:40:50 <tardyp> so I expect the packaging to be a big topic once we do the rc
16:40:58 <djmitche> rutsky: we'll come back to that question, hang on :)
16:41:00 <koobs> ill leave that for that topic
16:41:19 <djmitche> ok :)
16:41:25 <koobs> #idea separate into non compulsory UI component
16:41:41 <koobs> dont know whether feasible/possible, but just ideating.
16:41:51 <tardyp> its already the case
16:41:58 <djmitche> what does "non-compulsory" mean?
16:42:01 <tardyp> buildbot-www is a separate package
16:42:02 <koobs> buildbot-web
16:42:25 <koobs> djmitche: i mean, can run a master without a ui, possibly opening the way for other 'build status' display methods
16:42:33 <djmitche> right
16:42:34 <koobs> cli, export to html, i dunno
16:42:35 <djmitche> that's already the case
16:42:36 <tardyp> its the case
16:42:41 <djmitche> but people are going to want the UI too :)
16:42:44 <koobs> of course
16:42:50 <koobs> but that allows downstream packagers to avoid it
16:42:53 <koobs> punt the problem
16:42:58 <koobs> allowing master to be packaged well
16:43:02 <djmitche> haha
16:43:10 <koobs> 'separate the concerns'
16:43:14 <djmitche> so FreeBSD will run the 'buidlbot' package, but you're gonna have to run Ubuntu to get the UI? :)
16:43:35 <koobs> well ubuntu is going to have to deb the package anyway, which means solving the npm dep problem anyway
16:43:37 <rutsky> is it possible to run UI completely separate from Python server part, asking for data and controlling stuff over Data API?
16:43:43 <koobs> my guess is you wont see -www packaged for a while.
16:44:08 <djmitche> rutsky: yes
16:44:08 <koobs> in freebsd, we're "trying" out best to make npm deps per port much easier to port
16:44:12 <koobs> but its not a solved problem
16:44:40 * djmitche (aside) wishes there was some way to tell meetbot you went back to a previous topic :)
16:44:47 <djmitche> we do it EVERY WEEK :)
16:44:47 <koobs> #note would rather have seen a django ui/package
16:44:55 <koobs> sorry.
16:45:06 <djmitche> s'ok, I'm glad to have the discussion, just sad meetbot doesn't get it
16:45:17 <djmitche> django's not browser-side
16:45:26 <koobs> so? :)
16:45:35 <tardyp> no auto update
16:45:37 <djmitche> we need a browser-side app
16:45:37 <koobs> lets not bikshed stack stuff in this topic ;)
16:45:37 <tflink> djmitche: can't you just undo? or did you mean go back after a bunch of other things happened?
16:45:37 <djmitche> right
16:45:46 <djmitche> tflink: after other things happened, yeah
16:45:57 <djmitche> also, hi :)
16:46:07 <djmitche> ok, let's wrap up the packaging thing and discuss when we get there
16:46:32 <koobs> djmitche: one last thing. is node/npm a hard requirement for 0.9 ?
16:46:38 <koobs> "for the record"
16:46:47 <djmitche> yes
16:46:50 <tardyp> koobs: for the web ui, yes
16:46:55 <koobs> ok
16:47:13 <djmitche> ok :)
16:47:20 <tardyp> it is helping us a lot
16:47:34 <tardyp> and we know it does not help the packager :-/
16:47:44 <tflink> fwiw, the node stuff in -www is mostly what's kept me from updating the fedora packages. it's not just debian
16:47:53 <koobs> Back to GSoC ?
16:48:03 <djmitche> #topic GSoC 2016 (for real this time)
16:48:09 <djmitche> so first up
16:48:21 <djmitche> tardyp: do you have an estimate of the time commitment of mentoring?
16:48:25 <djmitche> or any other previous mentors?
16:48:43 <koobs> Id factor in N hours per week, where N can be 1-a few
16:48:50 <bobhood> There's no way to talk in absolutes there.  It depends a great deal on the student.
16:48:57 <koobs> djmitche: Might be worth asking some of our FreeBSD GSoC Mentors
16:49:07 <tardyp> it really nepends on the student
16:49:11 <koobs> And the task
16:49:19 <koobs> But there ought to be a minimum expectation for mentors
16:49:22 <tardyp> if you get good student, 1-2 hour per week
16:49:25 <bobhood> I spent 1-2 hours a week, but I had a great studend.  Motivated and knowledgable about Buildbot.
16:49:41 <bobhood> That was as an "project manager."
16:49:46 <djmitche> rutsky: ^^ helpful?
16:49:49 <tardyp> but in the gsoc questions they ask you how much time, and this go from 3 hours /w to like 20hours
16:49:50 <rutsky> ok, and estimation for upper limit? :)
16:49:51 <koobs> Goal: Student knows they have dedicated time with mentor. Results: ideally consistent progress, no blocking
16:50:38 <djmitche> that's a good perspective: 1-2h of scheduled time, and probably 1-10h of unscheduled time depending on the student, the project, and the co-mentor
16:50:44 <tardyp> a weekly meeting is highly reocmmended
16:50:50 <koobs> Agreed
16:50:54 <tardyp> + 1 or two quick sync
16:50:58 <koobs> Even if not taken. Knowing you have the time is key
16:51:01 <bobhood> Yes, I had one with my student and technical mentor.
16:51:07 <koobs> Able to prepare (on both sides) and anticipate
16:51:22 <tardyp> also make sure TZ is not too different is key
16:51:37 <koobs> It also sets expectations (should i ask again? oh, ill help if the student asks)
16:51:39 <tardyp> and look whether student works late or early
16:51:44 <koobs> Want to avoid that
16:51:46 <djmitche> #info discussion of time commitment for mentors - some scheduled time (1-2h/wk) plus a much more open-ended amount of unscheduled time (reviews, emails, technical chats, etc.)
16:51:50 <tardyp> sometime student work like 10PM to 3 AM
16:52:03 <tardyp> which is fine, but difficult to collab
16:52:11 <bobhood> I had an advantage in that area.  My "day job" team is all over the planet, so I had to deal with timezones daily anyway.
16:52:15 <koobs> djmitche: agenda during weekly catchup (even if simple) would also help make good use of limited time
16:52:20 <djmitche> #info also important to look at timezone "fit"
16:52:29 <rutsky> ok, then I would like to apply as technical mentor, if there is free position
16:52:33 <koobs> djmitche: "progress, questions, blockers, next week?"
16:52:35 <djmitche> ok, great!
16:52:40 <tardyp> bobhood: a student wont have this experience
16:52:40 <djmitche> koobs: good point
16:53:03 <bobhood> But it's important those mentoring be sensitive to it, as well as cultural differences, if needed.
16:53:10 <tardyp> ye
16:53:30 <koobs> who mentors the green mentors
16:53:33 <koobs> ahh!
16:53:52 <koobs> djmitche: Is there an organisational policy/guidelines on mentorship at all?
16:53:54 <djmitche> koobs: since we do two mentors per student, we can pair new and old
16:53:57 <bobhood> Quis custodiet ipsos custodes?  :)
16:53:58 <koobs> And i dont mean '50 page formal doc
16:54:02 <djmitche> haha
16:54:04 <koobs> I just mean, the basics
16:54:08 <djmitche> koobs: that's the main bit of it
16:54:14 <djmitche> one mentor is organizational, one technical
16:54:18 <koobs> cool
16:54:33 <djmitche> so the technical might do review, help with code-related questions, etc., but maybe not make it to the weekly meetings
16:54:52 <djmitche> while the other does weekly meetings, keeps track of progress on the schedule, helps the student find help and communicate, etc.
16:55:16 <djmitche> also, I think we should include student reports in this weekly meeting
16:55:20 <djmitche> we did last year and it helped quite a lot
16:55:21 <bobhood> Yes, my technical mentor was more often unable to attend.  Which was fine, because most of what I needed had to do with the student.
16:55:25 <djmitche> to make them part of the project
16:55:26 <rutsky> how many students expected?
16:55:38 <djmitche> rutsky: that depends on number of mentors, really
16:55:42 <koobs> What are the open organisational GSoC questions ?
16:55:44 <djmitche> I think we had about a dozen non-spam entries last year
16:55:56 <djmitche> koobs: you mean http://trac.buildbot.net/wiki/ProjectIdeas ?
16:56:14 <koobs> djmitche: I mean, whats left to be figured out for the program?
16:56:24 <djmitche> https://public.etherpad-mozilla.org/p/buildbot-gsoc2016
16:56:34 <djmitche> that's the application, and there are a few things I need filled in
16:56:40 <djmitche> and it's open for edits too
16:57:22 * djmitche pauses for folks to read
16:57:48 <koobs> djmitche: How many mentors have applied so far and what are the target?
16:57:52 <koobs> targets*
16:58:19 <djmitche> rutsky just volunteered, so 1 so far :)
16:58:31 <djmitche> and pierre and I will be co-mentors, I think (nobody else has stepped up)
16:58:46 <koobs> djmitche: Have you put a call out on the ML ?
16:58:50 <bdbaddog1> when does this start? (calendar wise?)
16:59:01 <djmitche> koobs: I did
16:59:02 <koobs> djmitche: Or to known organisations that use buildbot?
16:59:16 <koobs> djmitche: How long since the last? Reminder worthwhile?
16:59:20 <djmitche> https://developers.google.com/open-source/gsoc/timeline
16:59:28 <koobs> Any people from related/aligned (python) projects who might be interested?
16:59:30 <djmitche> koobs: it was thursday I think
16:59:45 <koobs> djmitche: twisted people in particular?
16:59:49 <djmitche> most of those projects have their own GSoC's - Twisted does
16:59:53 <koobs> right
16:59:56 <djmitche> so I wouldn't want to poach their mentors :)
17:00:01 <koobs> or!
17:00:03 <koobs> :)
17:00:05 <djmitche> last year, in fact, one of our students worked quite a bit on Tuesday
17:00:08 <djmitche> *Twisted
17:00:19 <koobs> Do you have libcloud latent buildslave in the ideas list?
17:00:34 <koobs> we could "make" it 12 weeks of work :D
17:00:49 <djmitche> it's not on the list, no
17:00:52 <tardyp> koobs: please feel free to add your own ideas!
17:00:55 <djmitche> it's a good one, though want to add it!
17:01:01 <koobs> "tested buildbot in N public IaaS Services"
17:01:22 <tardyp> this is a great modern topic
17:01:25 <bobhood> Did twisted ever finally get ported to Python 3?
17:01:36 <koobs> bobhood: theyre in the middle of it
17:01:40 <koobs> bobhood: some things ported
17:01:45 <bdbaddog1> build/metrics tracking/graphing?
17:01:46 <bobhood> They were in the middle of it last summer.  :)
17:01:54 <koobs> bobhood: theyre past the middle of it?
17:02:03 <djmitche> bdbaddog1: I took metrics off the list this year
17:02:12 <koobs> if buildbot python3 compat? #GSOC #Idea
17:02:16 <koobs> s/if/is
17:02:21 <bb-github> [13buildbot] 15adityadivekar03 commented on issue #1995: @rutsky , If I'm not wrong I should look into how to include the frames from inside eval into the logfile so as to make the logfile complete?... 02https://git.io/vg5dE
17:02:23 <bobhood> That was our project last summer.
17:02:24 <djmitche> http://trac.buildbot.net/ticket/3234
17:02:38 <koobs> miche has that url in muscle memory
17:02:40 <bdbaddog1> djmitche: metrics not a good project for gsoc?
17:02:54 <djmitche> bdbaddog1: I think it's too large
17:03:05 <djmitche> we really struggled last year
17:03:29 <koobs> what about a single metrics page, most of the work is in tracking stats ?
17:03:38 <koobs> fastest bot
17:03:38 <bdbaddog1> gotcha: maybe a more limited scope?
17:03:40 <djmitche> I think, at best, we'd end up with something like user support, that "sort of" works and just leads users to disappointment
17:03:40 <koobs> average build times
17:03:43 <koobs> bdbaddog1: yeh
17:03:53 <djmitche> yeah, if we could scope that down
17:03:56 <koobs> djmitche: descope it for the 80/20
17:04:05 <bdbaddog1> for scons project something like: memory and runtime tracking is important
17:04:06 <koobs> whats the main value prop for metrics?
17:04:13 <djmitche> yeah, if one of the two of you want to adjust that bug or make a new one with a much more limited scope, that'd be good
17:04:19 <koobs> djmitche: 'slowest steps' maybe
17:04:21 <bdbaddog1> so we can see if a commit improves or not.
17:04:24 <koobs> could be performance oriented
17:04:30 <koobs> step time deltas
17:04:34 <djmitche> yeah
17:04:40 <djmitche> so this was part of the problem:
17:04:50 <djmitche> some people wanted metrics of the things they are doing with buildbot
17:04:50 <bdbaddog1> gotcha.
17:04:51 <koobs> make a huge list, sort it, pick the top 5.
17:05:00 <djmitche> while others want metrics about buildbot itself, for monitoring in production
17:05:08 <bdbaddog1> Generic metrics storage?
17:05:19 <koobs> djmitche: thats not a problem, its an abundance of ideas, the issue is selection criteria, which is an organizational question
17:05:34 <djmitche> right
17:05:36 <bdbaddog1> allow builds to store any metric along with build in db? then pick metric to graph
17:05:38 <koobs> djmitche: what are the goals, what are the current GAPS
17:05:41 <djmitche> well, and telling people "no we're not solving that problem" :)
17:05:47 <djmitche> koobs: you are on fire today :)
17:05:51 <koobs> djmitche: where is buildbot behind, two whom do you want to add the most value?
17:05:56 <koobs> djmitche: this is me every day mate
17:06:09 <koobs> given a good problem space.
17:06:17 <djmitche> right, and I think adding new features in general is not the best use of dev time on Buildbot right now
17:06:28 <koobs> State it, then state where *is*
17:06:33 <koobs> Help people narrow down their idea space
17:06:37 <djmitche> having user tracking seemed like a good value prop
17:06:45 <djmitche> but having a half-baked user trakcing system is turning out to be horrible
17:06:53 <koobs> reminds me of that MVP image.
17:07:29 <djmitche> a lot of buildbot is half-baked
17:07:29 <koobs> djmitche: So what areas require the most focus? Is that on the top of the ideas page to help people orient their brainstorming ?
17:07:30 <djmitche> locks
17:07:33 <djmitche> latent workers
17:08:18 <djmitche> it's not, because the fixing doesn't boil down into projects very well
17:08:52 <koobs> s/fixing/replacing X with creative/challenging/innovative Y
17:09:08 <bb-github> [13buildbot] 15rutsky commented on issue #1995: > @rutsky , If I'm not wrong I should look into how to include the frames from inside eval into the logfile so as to make the logfile complete?... 02https://git.io/vg5F6
17:09:36 <koobs> djmitche: If its not new features, and its not fixing, what could it be?
17:09:52 <koobs> Innovation in the CI space?
17:10:26 <koobs> What about solving the builder concurrency problem in N permutation build_class CI test matrixes?
17:10:40 <bb-trac> [trac] #2977/project-idea (closed) updated by dustin (I think this would tie Buildbot in too many knots.  Django, for example, doesn't ...) http://trac.buildbot.net/ticket/2977
17:10:58 <djmitche> that sounds good
17:11:03 <djmitche> insofar as I don't know what it means
17:11:21 <koobs> My builders go from zero builds to N*M builds for any project that has many variations (with, without debug, with,wthout clang/gcc, with/without shared)
17:11:41 <koobs> Jitter/Build Offsets taking into account known average build times
17:11:47 <koobs> "Smoothing out load/concurrency spikes"
17:11:47 <rutsky> any plans/ideas to add ability to configure builds from codebases? as Travis does
17:11:50 <koobs> "Resource utilisation"
17:11:56 <djmitche> yeah
17:12:03 <djmitche> these are all great ideas to add more complexity to Buildbot
17:12:15 <tardyp> rutsky: https://github.com/tardyp/buildbot_travis
17:12:18 <koobs> s/complexity/make a better Ci framework/product
17:12:22 <koobs> o_O
17:12:31 <djmitche> koobs: no, not really
17:12:32 <tardyp> this can be done as an external buildbot plugin
17:12:38 <djmitche> we can bolt-on a bunch of poorly implemented features
17:12:44 <djmitche> and eventually what we have is an unusable mess
17:12:56 <rutsky> tardyp: perhaps we can find some tasks for GSoC from buildbot_travis?
17:12:58 <djmitche> also you need to escape your /'s
17:13:03 <koobs> djmitche: but you're assuming a) poorly implemented and then b) unusable
17:13:11 <koobs> djmitche: thats a function oif design/planning/spec/leadership
17:13:14 <djmitche> koobs: I have lots of experience
17:13:16 <tardyp> rutsky: yes I was thinking about it as well
17:13:21 <koobs> fear based decision making helps noone
17:13:21 <djmitche> yeah, I'm leading right now :)
17:13:29 <tardyp> or have some gsoc projects about creating plugins
17:13:48 <koobs> djmitche: im not saying you dont. just because failures have occured in the past, of that 90% of stuff thats ever been implemented is mediocre, doesnt make tomorrows or the next thing necessarily so
17:13:53 <djmitche> asking students to take on large, open-ended problems that will need excellent design, lots of implementation work, and ongoing maintenance is not effective
17:13:53 <bdbaddog1> Isn't GSOC mainly good for "nice to haves"
17:14:02 <koobs> bdbaddog1: ideally no
17:14:13 <bdbaddog1> +1 on unconstrained projects bad for students.
17:14:15 <koobs> bdbaddog1: its an R&D opportunity to push the boundaries of things
17:14:24 <djmitche> koobs: it can be
17:14:28 <koobs> bdbaddog1: it tests the capabilities of many things in a project
17:14:37 <djmitche> koobs: but the way Buildbot, as an organization, exists, that's not a good idea
17:14:43 <koobs> bdbaddog1: some projects cop out by doing 'obvious things'
17:15:03 <koobs> djmitche: i dont know what that 'effectively' means though
17:15:30 <djmitche> we have a very small group of core developers who are *massively* overwhelmed by the size of the project
17:15:32 <djmitche> in terms of complexity
17:15:50 <djmitche> and we're about to create a massive amount of new complexity (0.9.0)
17:16:03 <djmitche> with less of the old complexity deleted than we'd like
17:16:24 <djmitche> I don't have a good way to say it: buildbot is in real danger of collapsing under its own weight
17:16:35 <koobs> youre not creating complexity though, youre trading off growth for organizational scaling
17:16:52 <djmitche> look at it this way
17:16:58 <rutsky> djmitche: Pierre's idea about GSoC in BB plugins should decrease complexity
17:17:02 <bdbaddog1> organizational scaling assumes you have more resources..
17:17:03 <koobs> djmitche: sure, that might be the case, but misunderstanding the root cause (or actual contributing factors) is as key to getting through it
17:17:14 <koobs> bdbaddog1: no.
17:17:19 <djmitche> rutsky: right, that's a direction I'd like to go
17:17:24 <koobs> bdbaddog1: it *can*, but it doesnt necessarily
17:17:45 <koobs> bdbaddog1: Fundamentally this is something thats either ignored, conflated or misunderstood. "productivity" is a RATIO of "effort" to output.
17:18:12 <koobs> its not a thing you pay for
17:18:14 <djmitche> wow this got academic
17:18:21 <djmitche> this isn't a business
17:18:32 <bdbaddog1> We're probably out of scope for the meeting…
17:18:34 <djmitche> there are 0 employees, and ~2.5 part-time developers
17:18:43 <djmitche> I think this is an important conversation about the project though
17:18:46 <koobs> djmitche: this is about success, not business :)
17:18:54 <djmitche> right, but what does success mean?
17:18:56 <koobs> you said yourself you were at risk of collapse
17:19:02 <koobs> great question, action item that.
17:19:07 <djmitche> lol
17:19:11 <koobs> :)
17:19:16 <bdbaddog1> success means no contributor burns out.
17:19:40 <djmitche> ok
17:19:40 <bdbaddog1> The project adds reasonable features and fixes bugs at a reasonable rate in ways that aren't duct tape/bailing wire.
17:19:40 <koobs> great start
17:19:51 <bdbaddog1> + adds more core contributors.
17:19:54 <tardyp> needs to be more drunk for this kind of discussion..
17:20:08 <koobs> djmitche: My point on all of this was on topic to #GSOC. Orgnizationally, you can help people inovled in ideas to narrow down the focus to the things you believe are important
17:20:10 <djmitche> success mean (1) continue to serve our existing users, (2) fix bugs causing users problems
17:20:32 <djmitche> tardyp: me too, and it's 5 hours earlier here
17:20:38 * djmitche considers dosing his coffee
17:21:00 <djmitche> koobs: yeah, I can agree to that
17:21:02 <djmitche> so let's set this aside
17:21:04 <djmitche> for the moment
17:21:09 <djmitche> and look at the application
17:21:28 <djmitche> In particular, edits for "How will you help your students stay on schedule to complete their projects?" would be helpful
17:22:20 <djmitche> #info Great discussion of Buildbot's priorities and how those connect to choice of GSoC project ideas
17:22:26 <koobs> djmitche: answered (at least partially) prior topic. Consistent mentoring sessions, dedicated time, alignment (culturally, technical expertise, timezone)
17:22:37 <djmitche> #info you'll have to read the logs (click the links!) for moe
17:23:02 <koobs> djmitche: Also add, a specific "scoping" sessions prior to student accepting task, to ensure that project is well specified and deliverables are clear/understood
17:23:06 <djmitche> #info (from koobs) answered (at least partially) prior topic. Consistent mentoring sessions, dedicated time, alignment (culturally, technical expertise, timezone)
17:23:21 <koobs> djmitche: "ie; Project/Task Pre-Initiation Session"
17:23:23 <djmitche> yep, that's part of the community-bonding
17:23:43 <koobs> djmitche: "You'll vet, refine,iterate on tasks/ideas until they are "ready"
17:23:43 <djmitche> oh I see, that feeds into the application response
17:23:46 <koobs> yes
17:25:06 <koobs> djmitche: Also add something like "Each task will have defined N "key" progress milestones that will assist in re-orienting/scope if necessary"
17:26:10 <koobs> tailor as needed. the point is "student wont be going off for 12 weeks, then big bang assumed to have all deliverables at the end"
17:26:21 <djmitche> right, that's part of GSoC in general
17:26:52 <koobs> many projects liaise only at start and end, not many do it more often
17:27:10 <djmitche> huh
17:27:43 <koobs> djmitche: many projects laisie with their students only at the beginning, and then at the end with very few sessions during
17:27:57 <koobs> djmitche: they find out late that stuff is too open-ended, or progress has waned
17:28:06 <djmitche> well, I guess we're ahead of that curve :)
17:28:10 <koobs> then its too late to recover, which results in a negatory outcome for both student, and project
17:28:18 <djmitche> yep
17:28:23 <djmitche> OK, any other feedback on this application?
17:28:27 <koobs> djmitche: yeh, im just saying that explicitly stating that you'll do that in the app, will set you apart
17:28:35 <djmitche> rutsky: tardyp: can you guys sign up with the new GSoC and send me your usernames?
17:28:43 <djmitche> koobs: yep, I added it - thanks!
17:29:21 <koobs> GSoC Idea: Do all the work to remove the need for any <= *_requires
17:29:35 <koobs> buildbot: 'twisted >= 11.0.0, <= 12.1.0', - current version 15.x
17:29:36 <djmitche> what does that mean
17:29:39 <djmitche> oh
17:30:00 <koobs> djmitche: the deps are all horribly/unecessarily constrained, remove those constraints, freedom!
17:30:14 <djmitche> we don't have any right now :)
17:30:14 <koobs> djmitche: and bonus: less complexity.
17:30:22 <koobs> djmitche: setup.py says otherwise :)
17:30:33 <djmitche> where?
17:31:02 <koobs> I speak of 0.8.12, so if its gone in 0.9
17:31:03 <koobs> great
17:31:05 <djmitche> yeah
17:31:07 <koobs> backport!
17:31:11 <djmitche> I like the idea though
17:31:15 * koobs goes silent to help agenda
17:31:16 <djmitche> haha, right, me and whose army?
17:31:19 <djmitche> thanks :)
17:31:30 <djmitche> yeah, sorry to get into such a long run on one topic -- but I like the energy
17:31:37 <djmitche> no intent to exclude anyone else!
17:31:39 <djmitche> let's move on
17:31:43 <djmitche> #topic Bug 2340
17:31:52 <djmitche> rutsky: I have a PR to look at
17:32:12 <djmitche> https://github.com/buildbot/buildbot/pull/1962#issuecomment-184016202
17:32:17 <koobs> trac.buildbot.net/ticket/2340
17:32:18 <djmitche> #info a PR is up for the main bits of this project
17:32:32 <koobs> djmitche: Im glad this is a topic, coz i was going to ask about it earlier ;)
17:32:48 <koobs> djmitche: Whats the open question on the topic?
17:33:04 <djmitche> looking for an update from rutsky mostly
17:33:10 <djmitche> but I think that was it -- ready for review
17:33:41 <rutsky> in general, bug2340 branch is ready
17:34:12 <rutsky> I have some minor tasks to do with it, but they can be done after branch will be merged
17:34:31 <djmitche> awesome
17:34:37 <djmitche> and this has brought a lot of other nice fixes with it
17:34:42 <rutsky> so if you review it quickly and I quickly fix issues arised, I will do last changes in separate PRs
17:34:43 <djmitche> not least a lot better DB compatibility
17:34:48 <djmitche> ok
17:34:58 <koobs> rutsky: downstream packer question: buildslave script is being renamed im assuming ?
17:34:59 <djmitche> I'll make some time tonight
17:35:07 <rutsky> koobs: not yet
17:35:18 <rutsky> this PR renames only master-part
17:35:20 <koobs> rutsky: will there be a deprecation warning prior to the rename?
17:35:24 <rutsky> (with docs and tests)
17:35:26 <koobs> for a few versions?
17:35:31 <koobs> rutsky: that will help us downstream
17:35:38 <djmitche> koobs: lots of deprecation warnings, yes
17:35:41 <djmitche> he's done a great job of that
17:35:45 <koobs> djmitche: cool
17:35:57 <rutsky> koobs: all changes don in backward compatible way (at least should be so)
17:35:57 <djmitche> the new master will be compatible with both slaves (0.8.x) and workers (0.9.0)
17:36:03 <koobs> so i imagine there will be a few versions where buildslave buildworker will coexist?
17:36:13 <rutsky> koobs: yes, they should work
17:36:13 <koobs> in terms of what console_Scripts are installed?
17:36:20 <koobs> ok. sweet
17:36:34 <djmitche> koobs: have a look at the PR (see also the email to devel@)
17:36:36 <rutsky> worker part will be later, after bug2340 is merged
17:36:39 <djmitche> I'd love more feedback than just mine :)
17:36:47 <koobs> djmitche: you just got mine ;)~
17:36:53 <koobs> customer not coder
17:37:00 <sa2ajj> is it about the pr above?
17:37:03 <djmitche> haha, well, it was worth asking
17:37:06 <djmitche> sa2ajj: largely
17:37:26 <djmitche> https://pastebin.mozilla.org/8859931 <-- devel email
17:37:38 <djmitche> #info ^^ pastebin is the devel@ email seeking review
17:38:03 <djmitche> anything else for bug 2340?
17:38:09 <rutsky> one question more
17:38:15 * sa2ajj . o O (interesting read)
17:38:25 <koobs> djmitche: just a comment that the issue should always reflect current state (im bug admin for freebsd)
17:38:29 <rutsky> is anyone know public system where large PR can be reviewed?
17:38:36 <rutsky> i.e. 300+ files
17:38:45 <koobs> email? :)
17:39:06 <sa2ajj> probably e-mail would be the best for this kind of review
17:39:09 <rutsky> I would be happy to provide direct links on some of my changes, but looks like there is no service that can do it
17:39:18 <djmitche> koobs: just for perspective, you ask for a lot of things to be done (in the passive), yet there's nobody to do them
17:39:42 <djmitche> I'm not sure a 100,000 line email is any more useful than `git diff`
17:39:54 <sa2ajj> is it just one commit?
17:39:54 <bdbaddog1> rutsky: maybe google docs post the diff?
17:40:00 <djmitche> I like the format you've selected, with specific things to look at, and git diff instructions
17:40:08 <djmitche> sa2ajj: no, a few hundred commits
17:40:14 <djmitche> it's quite well-organized, just a massive change
17:40:20 <djmitche> I think we can make this work
17:40:26 <sa2ajj> then 1000000 line e-mail is not going to happen
17:40:47 <sa2ajj> anyway, i'll just follow the instructions (if/when i have time :))
17:40:51 <djmitche> ok, thanks
17:40:59 <rutsky> bdbaddog1: google docs will not be integrated with rest of codebase...
17:41:19 <rutsky> I would expect to see some review service to be able to work with such diff
17:41:19 <djmitche> and I suspect that google docs would crash your browser well before it got to this point
17:41:32 <djmitche> reviewboard might, dunno
17:41:38 <rutsky> djmitche: I tried - no
17:41:41 <djmitche> hah, ok
17:41:47 <bdbaddog1> :) just a thought.. I have 197 page doc in google docs..
17:41:57 <rutsky> it has pager, but no way to provide link on specific file in diff
17:42:12 <djmitche> ah, I see
17:42:21 <djmitche> ok, well, moving on to wrap up
17:42:26 <bdbaddog1> how many line chagnes are we talking about +/-
17:43:24 <rutsky> djmitche: I tried GitHub, Gitlab, BitBucket, Phabricator's Differential, Reviewable, Review Board.
17:43:37 <sa2ajj> wow
17:44:19 <rutsky> $ git diff buildbot/master...buildbot/bug2340 | wc -l
17:44:20 <rutsky> 38257
17:45:08 <sa2ajj> 334 files changed, 17232 insertions(+), 11714 deletions(-)
17:45:09 <rutsky> best option that I found, is to actually patch Gitlab to have option to show diff for a single file :)
17:45:13 <koobs> djmitche: usually its the assignee who should takes care of triaging/updating each issue they have. its the only way to scale. normally one doesnt need a single person/group to do it
17:45:29 <koobs> djmitche: we have the same issues in freebsd (and i imagine in many other projects)
17:45:33 <djmitche> yeah
17:45:42 <djmitche> #topic Development week in review
17:46:05 <djmitche> I just briefly want to note that we had a really active week
17:46:06 <koobs> djmitche: so the pattern we take is making it easier/quicker/more obvious for each person with assigned issues to do that
17:46:10 <rutsky> this week in short for me: I hate SQLite vs PostgreSQL :)
17:46:15 <djmitche> lots of DB fixes
17:46:19 <djmitche> heh, exactly :)
17:46:33 <djmitche> #info Lots of fixes and work this week - DB compatibility in particular (dustin & rutsky)
17:46:48 <djmitche> #info And a lot of fixes that went into the beta7 release
17:46:54 <rutsky> djmitche: what about zombie-apocalypse?
17:47:01 <rutsky> will you integrate dumb-init?
17:47:25 <sa2ajj> any particular pr to look at about this?
17:47:29 <djmitche> #info We also fixed the "zombie apocalypse" where twistd running in a Docker container accumulates hundreds of zombie children, since it is not an init process
17:47:33 <djmitche> rutsky: it's landed
17:47:42 <rutsky> oh, great, thanks!
17:47:42 <djmitche> sa2ajj: the zombie apocalypse?
17:47:49 <sa2ajj> yup
17:48:05 <djmitche> sa2ajj: https://github.com/buildbot/metabbotcfg/blob/master/docker/metaworker/Dockerfile
17:48:11 <sa2ajj> well, if it's landed i'll check history
17:48:34 <djmitche> sa2ajj: http://trac.buildbot.net/ticket/3454
17:48:38 * djmitche re-titles that
17:48:39 <sa2ajj> thanks
17:48:48 <djmitche> anything else to highlight this week development-wise? or other business?
17:48:57 <djmitche> this has been our longest meeting ever!
17:49:06 <rutsky> I have few
17:49:16 <rutsky> I would like to discuss https://github.com/buildbot/buildbot/pull/1994
17:49:19 <rutsky> quickly :)
17:49:21 <bb-trac> [trac] #3454/defect (closed) updated by dustin (empty comment) http://trac.buildbot.net/ticket/3454
17:49:28 <bb-github> [13buildbot] 15adityadivekar03 commented on issue #1995: Could you tell me how can I check this issue locally, so that I can check my fix locally first? 02https://git.io/vgdfo
17:49:45 <djmitche> #topic BR 1994 - install service_identity in Travis to silence Twisted's warnings
17:50:07 <rutsky> In short: Twisted complains about missing service_identity - should we add new dep to BB or should we ask end user to do so?
17:50:20 <sa2ajj> the latter
17:50:51 <djmitche> we do this for a number of other tools, yes
17:51:05 <djmitche> we should have it installed for tests, and we should include something in requirements.rst about it
17:51:06 <sa2ajj> rationale: more dependencies, more likely things might go wrong for a novice
17:51:09 <djmitche> right
17:51:12 <djmitche> and more for packagers
17:51:17 <sa2ajj> indeed
17:51:22 <rutsky> tardyp: do you agree with djmitche, sa2ajj?
17:51:51 <sa2ajj> since i do not create packages for quite some time, i forgot about my sorrows :)
17:52:30 <sa2ajj> rutsky: i think we'd rather leave a comment in PR
17:52:40 <sa2ajj> so tardyp would see it when he has time
17:52:41 <rutsky> ok
17:52:45 <rutsky> another question
17:52:51 <rutsky> idea: gather questions from StackOverflow with tags #buildbot at translate them in irc channel
17:53:02 <djmitche> #topic idea: gather questions from StackOverflow with tags #buildbot at translate them in irc channel
17:53:11 <rutsky> what do you think, guys?
17:53:11 <sa2ajj> translate how?
17:53:16 <sa2ajj> post notifications?
17:53:20 <rutsky> yes
17:53:21 <sa2ajj> like for prs and trac?
17:53:25 <rutsky> yes
17:53:35 <rutsky> a lot of user come here for simple questions
17:53:53 <rutsky> they can ask the on SO and that info would be potentially reused later
17:53:58 <koobs> if you dont have a supybot already, it has lots of plugins for these things - checkout the fork of supybot limnoria
17:53:59 <sa2ajj> here -- channel or SO?
17:54:03 <koobs> we're setting one up for freebsd
17:54:15 <rutsky> here - this IRC channel
17:54:16 <tardyp> looks cool
17:54:28 <koobs> github, twitter, stackoverflow, faq's, knowledge base plugins, etc
17:54:37 <rutsky> so that dev could be able to quickly respond on questions on SO
17:54:41 <tardyp> its always anoyring to start answering a question and see the guy closed irc client 5min after asking
17:54:47 <djmitche> so the idea would be to alert here when a question is asked
17:54:52 <djmitche> ?
17:54:52 <rutsky> djmitche: yes
17:54:57 * sa2ajj nods
17:55:00 <djmitche> I'm happy to try that
17:55:04 <djmitche> koobs: we have a supybot set up
17:55:09 <djmitche> koobs: do you want to add that plugin?
17:55:11 <rutsky> and perhaps somehow tell user that this is a good idea to post question on SO
17:55:41 <djmitche> koobs: https://github.com/buildbot/buildbot-infra/blob/6363303edda3c49cbcba50ffcece47262d934eff/jail-supybot.yml
17:55:42 <tardyp> this will boost our SO karma :)
17:55:42 <koobs> djmitche: presumably you could use the rss plugin
17:55:57 <sa2ajj> well, posting questions at SO requires registration...
17:55:58 <koobs> djmitche: im just getting started with it too
17:56:06 <rutsky> also SO has it's own policies about "dumb" questions and formatting rules, so user would think more when asking question
17:56:13 <djmitche> yeah
17:56:14 <sa2ajj> yup
17:56:15 <koobs> sa2ajj: do you want to post from here too, or just notify so someone can jump on and answer?
17:56:21 <djmitche> so I like this idea, just not sure who's going to actually do it :)
17:56:33 <djmitche> I don't think irc's line length is compatible with SO's guidelines
17:56:37 <rutsky> and there are plenty of SO users that can answer questions, when devs are not around in the irc channel
17:56:44 <sa2ajj> i think just posting a notification here would be enough
17:56:46 <koobs> djmitche: summary + link
17:56:57 <sa2ajj> at least, to see how often this happens
17:56:58 <djmitche> maybe
17:57:01 <djmitche> let's start with just the notification
17:57:06 <koobs> heading to bed, its 5am argh
17:57:09 <bdbaddog1> +1 notification.
17:57:11 <djmitche> where by "lets" I mean the following set of people: []
17:57:16 <sa2ajj> koobs: good night :)
17:57:19 <djmitche> anyone want to be on that list?
17:57:31 <koobs> djmitche: is the bot online here?
17:57:39 <djmitche> bb-supy`: <--
17:57:39 <bb-supy`> djmitche: Error: "<--" is not a valid command.
17:57:41 <rutsky> I'm not in at this moment
17:57:52 <djmitche> ok
17:57:54 <rutsky> at least I would like to see ticket or task created for this
17:58:06 <sa2ajj> i can't commit at this time [yet], getting there though.
17:58:07 <djmitche> #info idea is that a bot would post to the irc channel when a new question is asked on stackoverflow
17:58:07 <rutsky> if you like this idea in general
17:58:15 <djmitche> I like the idea, and a ticket doesn't hurt
17:58:25 <koobs> djmitche: ill give it a go, i want to solve it from freebsd too
17:58:28 <djmitche> #agreed the idea is sound, but nobody has signed on to do it
17:58:35 <koobs> djmitche: i cant do the config management thing though
17:58:42 <djmitche> cool! rutsky, want to create a bug assigned to koobs for now?
17:58:44 <koobs> djmitche: ill prototype it on MattDaemon
17:58:49 <djmitche> koobs: we can help with that
17:58:50 <djmitche> yeah
17:58:59 <rutsky> ok, I'll create ticket
17:59:01 <koobs> djmitche: ill get it to match #buildbot here and #freebsd in the other channels for So questions
17:59:03 <djmitche> sweet!
17:59:11 <rutsky> nice!
17:59:17 <djmitche> #action rutsky to file bug
17:59:20 <koobs> two birds one stone
17:59:26 <djmitche> #action koobs to prototype this using MattDaemon
17:59:26 <djmitche> yep
17:59:34 <djmitche> awesome, ok, well, we're now at 90 minutes!!
17:59:35 <djmitche> so
17:59:39 <djmitche> last call
17:59:57 <rutsky> no more questions from me :)
18:00:08 <koobs> Does meetbot have time limits?
18:00:10 <koobs> per topic?
18:00:18 <djmitche> haha
18:00:21 <koobs> propose: N minutes per topic for next time
18:00:24 <djmitche> I should have been more forceful
18:00:29 <koobs> #action tell meetbot devs we want that feature
18:00:30 <djmitche> with you  :)
18:00:34 <koobs> djmitche: yes :)
18:00:38 <djmitche> #endmeeting