mozilla :: #conduit

16 May 2017
11:22zalunis there a reason we're lowercasing all commands sent by `docker-compose run phabricator`?
11:33zalundkl|afk: seems like you're doing something similar to my task
11:34zalunI've been sharing this before it's working with the old conduitdemo - from Mercurial, I'm working on making it work with the new containers
13:29imaduememcote / glob / mars (i.e. l3 people) can one of you try going here:, log in with the red button, then on the resulting page scroll down to user info, and tell me if there's a way for you to determine scm level from that?
13:29imaduemeI either can't see my l1 access info there or it's not available
13:29imaduemeor I'm overlooking it
13:30globimadueme: looking
13:31globi don't see the scm group there
13:31globmaybe because it's a posix group?
13:31imaduemewhat the difference between a posix group and a normal one?
13:32globimadueme: you can look it up by hitting ldap using HTTP_OIDC_CLAIM_ID_TOKEN_DN
13:32globnot sure to be honest. i know that posix groups are not self-service, so i need to file bugs to get them updated (which is annoying)
13:32globi'm surprised they aren't there to be honest
13:32globchat with mr kang
13:32mcoteyeah I don't see anything
13:33imaduemeYikes, and how exactly can i use that token to look up LDAP? (Appreciate the help btw)
13:34globimadueme: look in vct - pylib/mozreview/mozreview/ldap (you want query_scm_group)
13:34globyou'll need credentials to bind to ldap too
13:35globi assume you have somewhere to store and deploy secrets
13:35globi'm not sure if there's a staging ldap that you can use for playing
13:35imaduemenot as of yet, besides ops puppet configs. But I will take a look at pylib and ask kang if there's an alternative :)
13:35globping jabba on irc and let him know. but also chat with kang - it may be possible for them to expose the scm_level_* groups
13:37imaduemeon it :)
13:37* imadueme waits 3 hours for the sun to move to the west coast
13:37globstupid sun
13:37imaduemecan't live with it, can't live without it
13:37marsimadueme: glob, is the ldap built into the vct dev environment suitable for testing against?
13:38marsclose enough to production?
13:38globmars: it should be
13:38* glob checks the ldap schema
13:39globmars: yup, it's the same schema as prod
13:39marsimadueme: ^ would that unblock you?
13:40marsglob: that's good new for setting up our demo and dev environments. IIRC it was trivial to dockerize. cool.
13:41imaduememars: ideally I wouldn't need to query ldap directly if auth0 exposes scm info. My plan is to poke around to see if that's possible, and if not we can use Auth0 to confirm the user is who they say they are and from there manually query ldap using the HTTP_OIDC_CLAIM_ID_TOKEN_DN glob spoke about
13:43globman, the HTTP_OIDC_CLAIM_ID_TOKEN is on the long side
13:43glob2035 chars
13:44imaduemegotta make those hackers work for it
13:47globi wrote my first webextension today; mostly as a trial run for my other more complicated addons i need to port at some stage
13:47globyou need flash installed to view vidyo replays. this adds a quick way to download the mp4 instead
13:48mcotegot an appt; back in ~40 min
13:49imaduemeneat, how hard was it to write?
13:50globpretty simple. not sure if that's because i'm already used to the addon/webex model
13:51globwebextensions are so very limited though; there's going to be a lot of stuff that won't be remotely possible
13:52globlike, look at
13:52globtrivial addon, does the same job as stylish but without all the crud
13:52globproblem is webextensions don't allow direct file access, so this won't be able to load the css files to overlay them
13:56imaduemethat's unfortunate I wonder which of: trying to not break addons with each new ff version vs making sure addons are secure vs being compatible with chrome addons was the biggest factor in the move to the new extension api
14:01globaddons within firefox are great for causing issues, because of the extensibility. when someone reports an issue in the #firefox channel, the first thing anyone says is to try without addons enabled
14:01globi think it's a move in the right direction, but our addon ecosystem is a massive differentiator vs. chrome
14:02globimadueme: if you're interested in what's going on with webextensions i can point you at a vreply which has the executive summary and plans
14:05globimadueme: i'm over 40, i don't know if that emote is "yes, link me up" or "nah, i'll look the other way" :)
14:10imaduemeglob: emoji's can take on many different meanings . In this case ==
14:11imaduemeI feel sorry for future historians
14:12imaduemeThis is possibly one of the hardest languages to decipher
14:18dklnice breakdown on the differences of Phabricator and Gitlab.
14:18dklthe gnome project is considering moving away from Bugzilla to either Gitlab or Phabricator
14:18dklwe obviously could not pick Gitlab anyway but interesting comparison
14:20imaduemenice find
14:37zalunwe force lowercase on all commands sent by `docker-compose run phabricator`. Is there a reason?
14:38mcoteprobably for easy string comparison?
14:40zalunmaybe there is a switch?
14:40zalunDifferentialBugzillaIdCustomFieldTestCase != differentialbugzillaidcustomfieldtestcase
14:44mcotezalun: best thing to do is figure out who wrote the piece of code that lower-cases it and ask them directly :)
15:03zalundkl: do you want to chat about it?
15:04zalun && ARG=$(echo ${1:-start} | tr [A-Z] [a-z])
15:05dklzalun: which part? the mozphab container or my extension issue with overriding the edit/view policies of new revisions?
15:05zalunis it to have the start command? I think we can remove this line then
15:06zalunI'm running it like docker-compose run --rm phabricator "../arcanist/bin/arc unit /app/phabricator/src/extensions/differential/customfield/__tests__/DifferentialBugzillaIdCustomFieldTestCase.php"
15:08dklright we can just add a new command called 'test-ext', etc. that will run the extensions test suite
15:09dklactually does phab need to be running for the tests to work?
15:09zalunthe problem is that running arc unit will not run all tests
15:09zalunit asks for the db
15:09zalunI tried to run the tests on unconfigured repo locally
15:10zalunand I needed to basically make everything except of web presence to make the test run
15:10dklwhat if we had a script that looped over the tests and ran arc unit for each?
15:11zalunthat would be good, but it would require to edit the phabricator extension on every filename change
15:11dklwell if we adopted a standard naming scheme it could just do glob find or something
15:12globnah, do it yourself
15:13zalunalso it would block devs from running a specific test
15:13dklissue is we do not want to adding any testing requirements to the mozphab production container. so in demo we create a new phab container that inherits from mozphab, adds the extra testing code
15:13zalunOK, so we can copy this file
15:13zalunit may even reside in phabricator-extensions
15:15zalunSo mozphab would have a command which would run a file from phabricator-extensions dir. That would work only on a new phab container, as mozphab would not have required environment (lack of bash, make, etc)
15:18davidwalshmars: imadueme: I don't find this helpful (; is a flag really what we want to do?
15:18imaduemeno, there's something else... let me look
15:19zalunthe test command should have an argument so one would be able to test specific files. `test current-stuff-im-working-on` would run tests specified by dev
15:19zalundkl: ^
15:19dklsounds good. we can put something together and try it out
15:20dklit should be in the demo container. or we could create a new repo that is specific for testing
15:20dklit can't be the mozilla-conduit/phabricator-extensions repo though
15:20dklsince that will be used in prod
15:20imaduemedavidwalsh: honestly, click should infer the type of the parameter based on the default variable ( You can try adding `type=boolean` but I suspect the problem may lie with how the environment variable is being set.
15:20zalunissuing the command should be in mozphab
15:21zalundkl: but it would work only in special circumstances (it might even check for bash)
15:21zalunat least configuration should be there
15:22zalunso someone working on an extension could change the tested files
15:30* zalun has to fly, but I'll be back
16:24imaduemeUpdate: Kang says that Auth0 should be providing the scm levels and if it doesn't then that's a bug that will be fixed
17:46zalundavidwalsh mars: I've added there ideas how to implement the unit testing to demo repo
17:52dklzalun: will take a look
18:15mcoteyou know it's funny, spending all this time trying to figure out how to parse a bug out of a commit message
18:15mcotewhile users of bzexport have been explicitly providing the bug number on the command line the whole time
18:47dklmars: couple points. 1) conduit-demo should use a 'phab' user in the demo shell instead of root maybe? 2) getting error when running 'hg init': *** failed to import extension conduit from /root/version-control-tools/hgext/conduit-client/ [Errno 2] No such file or directory: '/root/version-control-tools/hgext/conduit-client/'
18:48dklwill file pull requests for those
18:48dklsince they are not related to the demo.sql PR
17 May 2017
No messages
Last message: 8 days and 19 hours ago