mozilla :: #taskcluster

13 May 2017
18:21dustinhttps://treeherder.mozilla.org/#/jobs?repo=try&revision=d2365e23f35737f53852940f63721809a704c3fd
18:21dustin20 green linux talos runs (the full set) </humblebrag>
18:31garndtNice dustin!
18:31dustinand, at least for that limited circumstance, tc-worker is doing just fine
18:31dustinso, Nice jonasfj :)
18:32dustinand wcosta :)
18:32dustindid you want me to wrap up https://bugzilla.mozilla.org/show_bug.cgi?id=1363409 ?
18:32firebotBug 1363409 NEW, nobody@mozilla.org unable to schedule e10s specific talos jobs on tc scheduled platforms
18:34garndtI think one of us needs to, I&#39;m happy to work on it on Monday, but I won&#39;t be doing it over the weekend.
18:35garndtI just saw the dev platform email
18:35dustinyeah
18:35dustinit&#39;s been broken since March so it seems a little over the top, OTOH we should just fix it
18:36dustinon the plus side, I asked MattN for help with one of the Win10 bugs, so I can mark that one P1/critical too :)
18:36* dustin waves
18:37garndtHah
18:38garndtWe&#39;re just helping each other out
18:39garndtSo I&#39;m happy to help out with this if you want to put up with my annoying questions. I don&#39;t mind learning about this stuff, welcome it actually
18:39garndtMore hands in it, the better
18:39dustinI have some time now, so I might just knock it out
18:40garndtUp to you
18:40dustinseems like it&#39;s higher priority than we thought
18:40garndtYea. I knew it was important to fix, but this seems like it should be on the top of the list
18:41garndtI didn&#39;t realize it was that high :(
18:41garndtI like the suggestion of notifying others if we know of a regression like this, even if it&#39;s not fixed right away
18:42garndtNot everyone watches the bugs for things like this (nor should they really)
18:50MattNdustin, garndt: Maybe the bug I&#39;m seeing is even worse than that bug and maybe the workaround in my email doesn&#39;t even work I&#39;m not getting talos results on my new pushes with `-t all` yet https://treeherder.mozilla.org/#/jobs?repo=try&author=mozilla@noorenberghe.ca&fromchange=41c565a75384041b90a7d4d4d5a0b5c3f5f04117&group_state=expanded&tochange=25da5ac341fee01146b8cb4febf2d239cac16e7e
18:51MattNTalos results were working fine on linux64 and osx a week or two ago
18:52dustinindeed, there are no talos jobs in https://public-artifacts.taskcluster.net/WPDk6gtoQ5GBmqQmqYnB7A/0/public/task-graph.json
18:52MattNnot having talos results on try is a critical issue
18:52garndtHrm when that bug was opened &quot;-t all&quot; was working....
18:53dustinyeah
18:53dustinlet me take a look
18:54MattNWe&#39;re told to address talos regressions within 3 days but every piece of talos automation (even local ./mach talos-test) is broken for me so you can hopefully understand my frustration
18:55garndthere&#39;s one for linux 64 https://treeherder.mozilla.org/#/jobs?repo=try&revision=6323f32a287856621f1a76abf240312f3ef4b9d1&filter-searchStr=talos
18:56garndtMattN: yea I definitely can understand that, we&#39;ll dig in and try to figure it out with you. When was the last time you rebased against m-c? I&#39;m wonder if any scheduling changes went out and stale try pushes might be affected
18:56MattNjust yesterday
18:56MattNmy parent rev is https://hg.mozilla.org/try/rev/96b36c5f527d
18:56MattNthanks
18:57dustinoh, they all got optimized away -- https://public-artifacts.taskcluster.net/WPDk6gtoQ5GBmqQmqYnB7A/0/public/logs/live_backing.log
18:57dustinwtf
18:57MattNwas there a mistake with bug 1364421? I rebased on top of that to avoid optimizing out
18:57firebothttps://bugzil.la/1364421 ASSIGNED, bstack@mozilla.com unable to backfill or add new jobs for buildbot bridge job (linux64 talos, OSX *)
18:58MattN(I mean the temp. workaround)
18:58MattN(though I thought SETA didn&#39;t affect Try but idk what I&#39;m talking about)
18:59dustinyeah, signs are pointing to SETA
18:59dustinhmm
18:59garndtMattN: but your &quot;-t all&quot; push didn&#39;t include windows talos either, right?
18:59MattNthe windows builds aren&#39;t done yet
19:00MattNso not sure
19:00garndtoh ok
19:00MattNI just pushed that after sending the email
19:00garndtbased on some people getting talos results today on other pushes, I&#39;m wondering if this is that patch + SETA related
19:00dustingarndt: those aren&#39;t run by TC so we wouldn&#39;t see them in the taskgraph either
19:00garndtlinux and OS X are affected by this patch
19:00bstackIt&#39;s quite possible that we messed up the seta change.
19:00bstackAlthough it was quite small. Maybe we got it backwards though?
19:00garndtdustin: good point
19:01* dustin looks
19:03dustinno, that should be correct
19:04garndtugh, and I just screwed up my local checkout...
19:05dustinoh
19:05dustinsorry
19:05dustinI totally misread that patch
19:05dustinno, that&#39;s flat-out wrong
19:05* dustin shouldn&#39;t code on saturday :)
19:05MattNsorry
19:05dustin`return True, None` means, yep, optimize this away and nope, no need to replace it with another task
19:05dustinand that&#39;s what it did
19:06dustinso, we should back that out, bigly
19:06dustinsomeday I&#39;ll learn not to read comments and ignore the code
19:06dustin# return true
19:06dustinreturn False
19:07garndtcan be backout, and submit a new patch with it fixed up so that BBB backfilling still works?
19:07dustinwell, backout requires a sheriff
19:07MattNOr land a correct workaround on top?
19:07dustinmaybe we should just land a fix
19:07dustinyeah
19:07dustinone sec
19:07MattNsince the bug is leave-open anyways
19:07garndtah ok, didn&#39;t know if we like to fix the original patch if it was wrong, or submit a fix on top
19:08dustinbstack: r?
19:08dustinhttps://irccloud.mozilla.com/pastebin/oUlkvzBG/
19:08garndtthanks dustin
19:08MattNyes, thank you!
19:09bstackdustin: r+
19:09dustintx
19:10dustinlol, hg unshelve just destroyed those changes
19:10dustinI love how often hg deletes my work, keeps me sharp
19:12garndtI like how right now it&#39;s telling me I have outstanding merge conflicts, but my working directory is clean...I&#39;m not sure what I did wrong
19:12dustinit&#39;s just being hg
19:12dustinunshelve apparently merged my changes with someone else&#39;s commit
19:12* dustin shakes fist
19:13dustinand it&#39;s lost track of my working copy.. seriously hg?
19:13dustinit&#39;s a 4-line patch
19:16garndtit&#39;s just making sure you really, really want to submit that patch
19:17dustinyeah
19:17dustinMattN: ok, landed on inbound
19:17dustinnow while I&#39;m doing so well, I&#39;ll try to fix the e10s thing (but probably leave an r?garndt until Monday)
19:17MattNdustin, garndt, bstack: thank you very much!
19:17dustin:)
19:18garndtsomeday I&#39;m going to learn how to fix these problems without blowing away my checkout and recloning :)
19:21KWierso|afkgarndt: `hg strip --no-backup &#39;not public()&#39;`
19:22dustingarndt: my experience suggests, no, you never really do
19:22KWierso|afkshould blow away any not pushed changes
19:22dustinI hate that whereas &#39;git checkout&#39; will say &quot;hey you have uncommitted changes&quot;, &#39;hg up&#39; just gives you conflict markers with no way to go back
19:22garndtyea
19:22garndtI&#39;m just not smart enough for hg
19:23* MattN has only had to re-clone an hg repo once in 8 years
19:23garndtyea, I&#39;m not blaming hg, this is purely a problem with my mental model and knowledge
19:24dustinyeah, i haven&#39;t had to re-clone in a while (aside from data corruption)
19:24dustinbut it breaks in some pretty obscure ways
19:24MattNok, just checking
19:24* MattN does 2 new try pushes for talos
19:25MattN(with the inbound patch)
19:28dustinheh, that&#39;ll help :)
19:33garndthttps://tools.taskcluster.net/task-group-inspector/#/bdnWiwvWQTaNc13JyE2iBw?_k=hoxhat
19:33garndtI see talos in there!
19:34garndtthat&#39;s the graph for https://treeherder.mozilla.org/#/jobs?repo=try&revision=c59ad9e9cc37364f0e6c521cabfc4a4403bcf4c7
19:35garndtdustin: https://bugzilla.mozilla.org/show_bug.cgi?id=1363409#c10 What about triggering non-e10s talos tests?
19:35firebotBug 1363409 NEW, nobody@mozilla.org unable to schedule e10s specific talos jobs on tc scheduled platforms
19:36dustinit will be somehting like -t g1-e10s to trigger the e10s test
19:36dustinwhich is how -u works for unittests
19:36dustinof course now you&#39;re going to tell me that -u and -t work differently in buildbot
19:37dustinbecause if things made sense in try, the universe would implode on itself
19:37dustinand we wouldn&#39;t want that, so we keep try and commonsense isolated by a 16&#39; concrete wall
19:39garndtyea, but what about &quot;-t g1&quot; to trigger the non-e10s?
19:40garndtwouldn&#39;t that mean the try names don&#39;t match so it won&#39;t get scheduled?
19:40* garndt should just probably go back to cleaning the house and stop interrupting :)
19:41dustin-t g1 will trigger the non-e10s talos (only)
19:41garndtok
19:42dustinpreviously all talos jobs, e10s or not, had the same name
19:42dustinso -t g1 would trigger both
19:42dustinand -t g1-e10s would trigger nothing
19:43garndtright, ok
19:44dustinok
19:44dustinI think I should probably close irc :)
19:44* dustin waves
19:47garndtlater sir
19:47garndtthanks for your help
19:56MattNyay! talos results are working on try including the --rebuild-talos 5!
14 May 2017
No messages
   
Last message: 129 days and 23 hours ago