mozilla :: #taskcluster

20 Apr 2017
04:41pmoore|awaygarndt: grenade: testing 8.3.0 release in https://treeherder.mozilla.org/#/jobs?repo=try&revision=be69d0d48ad53f661ee7de6ce34ecad19e86d9c0
07:14pmooregah
07:14* pmoore investigates
07:30pmooreweird, https://queue.taskcluster.net/v1/task/Rfdymk6_RyeSCVmBGdnsMA/artifacts/public/build/firefox-55.0a1.en-US.win32.test_packages.json seems valid - but logs say no json could be found
07:32pmooregrenade: Tomcat|sheriffduty: have you seen this type of error before? ( https://treeherder.mozilla.org/#/jobs?repo=try&revision=be69d0d48ad53f661ee7de6ce34ecad19e86d9c0 )
07:33Tomcat|sheriffdutygood morning pmoore
07:33pmooreTomcat|sheriffduty: good morning! maybe this could be same as bug 1306865? although that happened 3 months ago, only one day :/
07:33firebothttps://bugzil.la/1306865 NEW, nobody@mozilla.org Intermittent-infra Funsize ValueError: No JSON object could be decoded
07:34pmooreah no, i guess that is funsize related
07:34pmooreah well
07:34Tomcat|sheriffdutyyeah
07:34pmoorei've retriggered one of the jobs, maybe ok now
07:50pmoorelol, ok i see the problem - ironically it is because the generic worker isn't respecting the content encoding of the artifact it downloads, when it is the one that set the content encoding :p
07:51pmoorei can fix that ....
07:51pmooreah no, it is mozharness that isn't respecting it
07:51pmooreso not so ironic
07:51pmoorebut i'll see if i can fix mozharness
08:27pmoorestrange, even when sending `Accept-Encoding: identity` cloudfront still insists on returning `Content-Encoding: gzip` :/ (also with ;q=1)
08:57pmooregrenade: just a heads up - we won't be able to roll out 8.3.0 until i finish testing a mozharness change, which i'll also need reviewed by a US-based person, which means probably won't be until tomorrow we can roll out 8.3.0 globally
08:57pmoorefwiw this is the change i'm currently testing: https://hg.mozilla.org/try/rev/ca9a86020187816eedb65aa8448cd3a5b9f99343
08:58pmoorewhich is needed for bug 1347956
08:58firebothttps://bugzil.la/1347956 NEW, nobody@mozilla.org Some public-artifacts.taskcluster.net files are not served gzipped
09:03pmooregrenade: so i'll make a PR for 8.2.0 instead
09:03grenadeok
09:11pmooregrenade: https://github.com/mozilla-releng/OpenCloudConfig/pull/56
09:18grenadecheers pmoore! staggering the rollout to try and win10 first. the rest when i don't get shot.
09:19pmooregrenade: sounds sensible - thanks!
10:50allstarschhI got a Gecko Decision Task opt failure on try-server,
10:51allstarschhbut I have no idea why
11:21pmooreochameau: i've granted permission to github.com/ochameau/ff-dt to use win2012r2 worker type - let me know how you get on! :)
13:57catleewcosta: do you still need a review on https://github.com/mozilla-releng/buildbot-bridge/pull/45/files ?
13:57catleeor did you figure out a way to use the upload task id?
13:57wcostacatlee: oh, no, found another route :)
13:57catleewcosta: excellent!
16:10pmooregrenade: garndt: armenzg_brb: how does this look? https://public-artifacts.taskcluster.net/X4CZTXqWToadLs3X1SOdEA/0/public/test_info/mozilla-test-fail-screenshot_sadien.png
16:10* pmoore takes a bow
16:10garndtthat sure looks like a screenshot!
16:11garndt\o/
16:11pmooreand do you see what it says? i think we have the reason for the failures too! \o/
16:11garndtwooooo
16:11* garndt does a dance even with his broken ass foot
16:12pmoorehahaha
16:12garndtI wonder if that prompt is ac ommon one for the test failures. Do you know what it would take to disable that prompt?
16:12garndtI don't know much about the firewall exceptions
16:14grenadepmoore ++
16:15grenadeThat is sweet!
16:15armenzg_brbpmoore: excellent!
16:16armenzg_brbfyi a lot of known pop messages will now be seen
16:16armenzg_brbthere are modifications on the releng machines for all of them
16:16pmooreaki: many thanks for the quick review btw :)
16:17armenzg_brbthere's going to be few for certain utilities that are needed for running unit tests
16:17akipmoore: no worries. this may be completely seamless, or you may get bounced out of m-c
16:17pmoorehahaha
16:18pmoorearmenzg: so i think grenade already went through most of these things with win7 - i suspect most of these issues will be solved already, this might be a random straggler due to some minor difference in the setup - but we went through all this with win7, and used that knowledge when setting up win10, so i don't expect a flood of problems
16:19armenzgpmoore: that is good!
16:19armenzgless pain
16:19pmoorehopefully!
16:20pmoorebut i know you've been burnt a lot from this in the past - i remember you were our platform guy :)
16:46pmoore|awayand this is the tracking bug for win10 stuff: https://bugzilla.mozilla.org/show_bug.cgi?id=1326419
16:46firebotBug 1326419 NEW, nobody@mozilla.org [tracking] green up Windows 10 tests on Taskcluster
16:50Aryxhi, "Error calling 'link' for taskclusterProxy : (HTTP code 500) server error - shim error: context deadline exceeded" https://treeherder.mozilla.org/logviewer.html#?job_id=93060165&repo=autoland
16:51* dustin shakes fist at docker
16:54armenzgdustin: garndt once you read my comment and you agree on the change of direction I will close the bugs and refile properly
16:55armenzg_lunchpmoore|away: /me still has nightmares from it
16:57pmoore|away:D
19:26erahmWhat's the magic URL for discovering sub-routes again? Like I'm trying to figure out why gecko.v2.mozilla-central.revision.c0ea5ed7f91a6be996a4a3c5ab25e2cdf6b4377e.firefox.win7-debug doesn't exist
19:27erahmSo I want to look at gecko.v2.mozilla-central.revision.c0ea5ed7f91a6be996a4a3c5ab25e2cdf6b4377e.firefox
19:29garndterahm: is this what you're looking for: https://tools.taskcluster.net/index/#gecko.v2.mozilla-central.revision.c0ea5ed7f91a6be996a4a3c5ab25e2cdf6b4377e.firefox/gecko.v2.mozilla-central.revision.c0ea5ed7f91a6be996a4a3c5ab25e2cdf6b4377e.firefox
19:29erahmgarndt: yeah I think so, now why isn't win7 in there...
19:34erahmHmmm I think maybe it's b/c there isn't a win7 build per se
19:34garndtyea I was just looking, and it seems it might be win32-debug?
19:35garndtnot one specific for win7
20:30dustinjonasfj: what's the design idea behind plugins being disabled, instead of enabled?
20:30dustinwon't that make it difficult to deploy new versions of tc-worker with new plugins defined?
20:31jonasfjyeah,
20:31jonasfjdustin: I changed it..
20:31dustinoh
20:31dustinfun :)
20:31jonasfjdustin: checkout -> https://docs.taskcluster.net/reference/workers/taskcluster-worker/docs/configuration
20:32jonasfjdustin: I changed config quite a bit...
20:32jonasfjdustin: but, at-least it should now be reasonably documented... and it's much easier to write config against the schema from the docs site :)
20:32dustinawesome :)
20:33dustin"A plugin is enabled if it has an entry in this mapping, and isn't explicitly listed as 'disabled'. "
20:33dustinI like htat
20:33jonasfjyeah
20:33jonasfjplugins: {disabled: [], myplugin: {}} <-- myplugin is enabled
20:33jonasfjplugins: {disabled: []} <-- myplugin is disabled
20:34jonasfjplugins: {disabled: [myplugin], myplugin: {}} <-- myplugin is disabled
20:34jonasfjplugins: {disabled: [myplugin]} <-- myplugin is disabled
20:34dustindisabled is a good default state :)
20:34jonasfjyeah,
20:34dustinI&#39;m setting up tc-worker 0.0.13 to run on linux hardware
20:34dustinhence the interest :)
20:34jonasfjthe disabled: [] <-- array, is just to allow you to disable it without deleting stuff from your config file..
20:35jonasfjdustin: did you make a new version?
20:35tedyaml seems neat, but i can never actually remember how to format things properly
20:35jonasfjdustin: you&#39;ll probably want to make a new version, just hit the tag thing on github
20:36dustini didn&#39;t
20:36dustinI just want to see this one run a task, then i can
20:36dustinted: I suppose you know protobuf by heart though
20:36jonasfjlol,
20:37jonasfjYAML is only problematic if you do all the crazy things like multi-line strings, which you mostly don&#39;t need
20:38dustinOMG, it Just Worked(tm) - https://tools.taskcluster.net/task-inspector/#GJqzcK9NSMe08dJG_gJ-jQ/0
20:38dustinjonasfj: and there&#39;s a whole reference/referent thing with * and &
20:39dustinit&#39;s super-useful for Ansible
20:39dustinbut I&#39;ve avoided using it in the in-tree stuff because it&#39;s also not very legible
20:39jonasfjyeah, YAML is too much...
20:39garndtdustin: why would you be surprised? everything just works! world is full of rainbows
20:40jonasfjactually JSON5 is pretty neat...
20:40tedhand-authoring JSON is kind of awful too, don&#39;t get me wrong
20:40dustinYAML lets you have comments
20:40tedbut like, i cannot for the life of me write the artifacts line in a YAML task definition properly
20:40dustinthat, alone, is enough for me
20:40tedwhere it&#39;s a list of objects
20:40jonasfjJSON5 == JSON but with ES6 syntax
20:41dustinted: JSON is a subset of YAML, so you can write it as [{..}, {..}] if you want
20:41tedyeah?
20:41dustinjonasfj: that makes no sense :)
20:41tedi might just do that
20:41dustinI suppose JSON6 has ES4 syntax?
20:41jonasfjlol
20:42jonasfjokay, ES5
20:42jonasfjJSON5 is a proposed extension to JSON that aims to make it easier for humans to write and maintain by hand. It does this by adding some minimal syntax features directly from ECMAScript 5.
20:42jonasfj^ cite http://json5.org/
20:42dustinthat does sound pretty nice
20:42jonasfjgranted they just made it super hard to write a parser for it :)
20:43tedi kind of dig toml that Cargo uses
20:43garndt&quot;Both inline (single-line) and block (multi-line) comments are allowed.&quot;
20:43garndtwheeeeee
20:43tedit seems moderately sensible, but i also haven&#39;t had to deal with parsing it or anything
20:44dustinI feel like toml is YANINI
20:44dustinWindows already has 17.2 .INI formats
20:44dustinthen there&#39;s a few Python formats
20:44dustinnow Rust is in on the game
20:44tedhah
20:44tedi don&#39;t think rust invented it
20:44jonasfjtoml was unwritable until I sat down and read the spec...
20:44jonasfjI could read cargo files, but didn&#39;t understand the data model
20:44AutomatedTesterJSON does allow comments but few parsers understand it (or at least thats how I remember it)
20:45sfinkother way around
20:45sfinkJSON does not allow comments but many parsers understand them
20:45dustinright
20:45tedbut also i don&#39;t know what the spec looks like or anything
20:45AutomatedTesterahhh
20:45dustinand those are Parsers That Are Not To Be Trusted
20:45dustine.g., eval()
21:02tedhum
21:02tedResult: IDLENESS_LIMIT_EXCEEDED
21:02tedis there a way to override that?
21:04tedoh wait, nvm
21:04tedi just set my maxRunTime too low hah
21:18dustinjonasfj: want to chat briefly about tc-pulse?
21:18jonasfjdustin: sure, let&#39;s bring bstack..
21:18dustinyep
21:18bstackNO
21:18bstackok
21:18dustinhaha
21:18dustinmy rm
22:28garbasdustin: just before i forget - busy working - thank you a lot for proof of concept on tc+auth0
22:28dustinsure
22:28dustindid you give it a shot?
22:41garbasdustin: i will next week after tuesday. too much urgent stuff until then
22:41dustinok
22:41dustinit won&#39;t be running by then :)
22:42dustinit really only takes 30 seconds
22:42garbasi&#39;m definetly giving it a sping locally
22:46dustinyou&#39;ll need to ask kang for credentials
22:46* dustin recommends clicking http://taskcluster-login.ngrok.io/ in lieu of checking irc for 30 seconds :)
22:47dustinsorry https://taskcluster-login.ngrok.io/some-tool
22:49* garbas clicks :)
22:51* dustin heads off for the evening so I won&#39;t bug you further :)
21 Apr 2017
No messages
   
Last message: 6 days and 16 hours ago