mozilla :: #conduit

16 Mar 2017
00:21mcotesmacleod: work in progress: https://docs.google.com/document/d/1ao95IYt1vTpBT256DvtNZGzOLHzWP9xjgi0faJJQ4WY/edit
00:22mcoteI'll probably add another paragraph outlining some ideas of how we're identifying what constitutes a topic (e.g. bug number), and then another paragraph or two about the review-request generator or whatever we're calling it
00:22mcoteprobably will finish this tomorrow
14:02dklso if i run a test script from tests/test_bmo_attachments.py using pytest, it cannot find commitindex/reviews/bugzilla.py.. but if i move test_bmo_attachments.py to the current directory, 'pytest test_bmo_attachments.py can find commitindex/reviews/bugzilla.py fine
14:02dklis there something happening in the docker environment that allows the modules to be found properly?
14:03dklthat i am not doing locally?
14:07imaduemesmacleod: Finally got the review request up if you want to race to the 11th hour to land it. No rush though, better to get it right :) https://reviewboard.mozilla.org/r/121056/diff/1#index_header
14:07imaduemeFor the hg client extension ^
14:08imaduememars might know what's up with that dkl
14:09dklyeah
14:10imaduemeAlready updated to diff 2: https://reviewboard.mozilla.org/r/121056/diff/2/
14:16dklarg mountebank why you no work
14:19dklmountebank returns the proper status code but is not returning the body that i configured it to
14:19* imadueme
14:22marsdkl: I'm online if you want to have a look in a screen share
14:22dklmars: yeah. let's try it
14:23dklshould i come to your office?
14:23marsyep
14:23* dkl puts on a pair of pants
14:25marsdkl: pip install -e .
14:31marsimport pdb; pdb.set_trace()
14:33marsdkl: http://www.mbtest.org/docs/api/mocks
14:48marsdkl: https://httpie.org/
14:50* smacleod wishes https://github.com/gabrielfalcao/HTTPretty had python3 support. He always wanted to use it
15:01smacleodimadueme: gave you some feedback, looks like you're on the right track though :)
15:05* imadueme thanks
15:06smacleodmars: you broke us https://circleci.com/gh/mozilla-conduit/conduit/50
15:07smacleodstupid flaky irc bot fails to join half the time
15:07marssmacleod: 8|
15:15smacleodmars: see my comment in https://trello.com/c/FyHY3b5k/114-3-figure-out-what-the-best-way-is-to-pass-errors-from-backing-services-through-the-app-and-out-through-the-api-log-and-return-5x
15:15janxsmacleod: for what it's worth, we considered Swagger for Janitor API, because we wanted something that ties implementation + docs + tests together in order to stay current and reduce maintenance cost, but deemed it too opinionated / weird / painful (especially with our Node.js server), so we re-implemented our thing
15:15janxhttps://github.com/JanitorTechnology/selfapi/ (forces you to write meta-data and examples right next to handlers, and can serve live API docs and run your handlers against its own examples for tests)
15:16* janx apologizes for the verbose product placement
15:16smacleodjanx: heh no worries - I think swagger works for us, as this is all new APIs so we can make sure they fit well in swagger
15:18janxsmacleod: ok nice, it just sounded like you'll need to maintain both swagger.yml alongside your actual implementation going forward, which could diverge
15:19smacleodjanx: nah, the actual implementation will generate our swagger.yml (using the python library Flakse-RESTPLUS). We just write a swagger definition for the endpoint *before* implementing it, that way we can make sure we actually build what we said we would, and consumers can start working against the API before it's implemented
15:20janxsmacleod: ah ok, makes sense. Flask sounds really powerful for creating APIs (definitely more so than Node.js's express)
15:20smacleodmars / davidwalsh: https://www.mercurial-scm.org/wiki/Phases#What_is_a_.22publishing_repository.22.3F
15:20smacleodjanx: https://github.com/noirbizarre/flask-restplus
15:45dklmars: for after meeting https://gist.github.com/dklawren/94052484df648364bb163a5e8a7e200c
16:19smacleodimadueme: sweet camera angle
16:19* imadueme
16:23janxglob: I agree about requiring a bug even for trivial changes (incidentally, we recently moved the janitor repo under a new github org, and I lost the ability to directly push small changes. I briefly tried to restore that access, but then thought I might as well create small pull requests for everything which is much cleaner)
16:24imaduemeyou gotta admit janx, PRs are a bit more streamlined :) Still agree w/ bugs though
16:25janximadueme: agreed! but my point was that it's great to have a process that prevents direct pushes (fewer risks, more tracability/accountability, etc)
16:26janximadueme: also, I'm trying to make Firefox bugs + patches as streamlined as PRs with Janitor :) (with buttons to file bugs / upload patches for review)
16:27janxnot sure how "easy" this whole process can be made though
16:28imaduemethat would be awesome! never used janitor before but i hear good things :)
16:29janximadueme: it's not as convenient yet as it can be soon, but it already does the bootstrap part fairly well (spawns a full up-to-date contribution environment in 1.5s, and you can create as many as you want/need)
16:30janxnext up is an IDE button to push commits to your review system :)
16:30imaduemewow, it took me like an hours to download moz-central
16:30imaduemean hour*
16:32globimadueme: no comment
16:32* imadueme maybe i should upgrade to hg 4.1
16:33janximadueme: it takes people between several minutes and several days (sometimes weeks, or even worse, some contributors never manage to set up because of a weird OS or environment)
16:34imaduemelooks like I feel on the good side of things then :P
16:34janx`git clone https://github.com/mozilla/gecko-dev` completes in a few minutes out of the box, `hg clone` is a lot slower out of the box but you can configure it to be extremely fast (you just have to know what to do)
16:34janxbut you can't beat 1.5s :P
16:35janxsource: "spawn" times on https://janitor.technology/projects/firefox/
16:35imaduemeoh i remember, when i try to use janitor i can't. I have to be invited that's why I clone moz-central :P
16:35* imadueme
16:36janximadueme: I can send you an invite (I currently send them out when I see requests, but we used to be maxed-out server side)
16:37imaduemejanx: I can't promise I will use it so do save it for someone else if you're at capacity. Otherwise I'll at least check it out
16:37* janx will stop de-railing the topic, but just mention that invites are going away soon (in a few months or so)
16:39* imadueme can't tell if that's a good thing or bad thing but it sounds good
16:40janx"used to" be maxed-out :) glob's AWS instance really helps by taking developer load, and also it's now super easy to add more instances as needed
16:41janximadueme: no invites = everyone can use janitor instantly, and not even have to sign in (except to save any work done in the IDE)
16:41imaduemenice!
16:41imaduemethat could be a huge boost for contributing
16:43janxdefinitely (we'd like to offer a URL that opens a privte IDE on any Firefox repository + sha, so that you can quickly hack on prototypes, try out reviews, debug failed taskcluster containers, etc)
16:47smacleoddavidwalsh: duuuuuude get a new laptop
16:51janxsmacleod: isn't the network connection partly to blame for such big checkout times ?
16:52smacleodjanx: clone times for mercurial?
16:53janxsmacleod: yes
16:53janxalso it's a bit weird that you have to `./mach bootstrap` to make hg fast, but "mach" is part of what you're actually trying to clone with hg in the first place
16:53smacleodjanx: not if your client is properly configured - it should be using bundles downloaded from S3, through a CDN - should be quite quick
16:54smacleodjanx: http://mozilla-version-control-tools.readthedocs.io/en/latest/hgmo/bundleclone.html
16:55janxsmacleod: that still requires you to find a way to `./mach bootstrap` before having actually cloned anything right?
17:01smacleodjanx: no, you could just configure the extension in the base image
17:01smacleodsorry, it's not an extension anymore, it's builtin to recent mercurial
17:08circleci-botFixed: davehouse's build (#51; retry by smacleod) in mozilla-conduit/conduit (master) -- https://circleci.com/gh/mozilla-conduit/conduit/51?utm_campaign=chatroom-integration&utm_medium=referral&utm_source=irc
17:59smacleodckolos: just finishing up our team meeting, might be 2 minutes more or so, no confidential stuff or anything so don't worry about lurking
17:59ckoloskk
18:01dklmars: for after meeting: test_bmo_attachments.py https://gist.github.com/dklawren/94052484df648364bb163a5e8a7e200c and bugzilal.py https://gist.github.com/dklawren/de2e8904171e696b13e6061f60839dab
18:01dklgotta do something
18:04ckoloshttps://autoland.dev.mozaws.net/scripts/environment.js
18:35ckolosYou guys should check the site now :)
18:38imaduemeckolos: \o/ Awesome thanks a million
18:39ckolosnp yo
18:57* smacleod goes to take a break after the 4th hour of meetings
19:12marsdavidwalsh: I'm in sococo, let's plan out that card
19:13marsdkl: I'll look into the mountebank stuff next
20:44davidwalshmars: Our plan has a minor issue in that we can't extend a service that has a "depends_on"
20:44davidwalshLooking for the best way around that now
21:00imaduemesmacleod: Updated the hg client extension, I think it is good to go now: https://reviewboard.mozilla.org/r/121056/diff/2-4/. I left a comment on the issue I'd like to drop. If you're not going to be around right now feel free to fixup and land that so that noone is blocked.
21:04marsdavidwalsh|afk: maybe the sidecar container was the wrong way to go. It might be better to pass in an env var with a list of repos that the container will initialize itself. BUT - we could have the same problem all over again with commitindex, or autoland, or any other depends_on service. :/
21:17imadueme|awayalso smacleod (or whoever): In that rr above I wanted to ask someone to look over it for any GPL violations, since you guys probably have a better grasp of it than I
17 Mar 2017
No messages
   
Last message: 8 days and 18 hours ago