mozilla :: #ateam

15 Mar 2017
09:57marcoekyle: ping
10:43ekylemarco: pong
10:43marcoekyle: I received an alert email
10:44marcoekyle: I think it's from a Windows installation, so I guess it's a false positive?
10:44ekylemarco: that was me testing that it actually sends, and since windows has a problem, I can ensure I get an email
10:45marcook :)
10:45ekyleso, I turned you off right now, but I will turn you back on if/when I see errors from grcov.
10:48marcoekyle: sounds good
10:48marcoekyle: is it in production now?
10:50ekyleit might be; I got to the point of having the code on the etl branch (which is picked up by workers) I did not push the changes to existing workers, which I am doing this morning.
10:55marcoekyle: btw, you forgot to create the weekly meeting
10:59ekylemarco: yes, thank you, I will make the meeting
11:22AutomatedTesterdecamping to coffee shop for a bit
11:23* AutomatedTester needs caffeine to wake me up
11:24jgrahamAutomatedTester needs to learn to talk in the third person ;)
11:25* AutomatedTester puts on his autocratic hat
11:25AutomatedTesterthe beating will continue until moral improves
11:28jgrahamWhich moral in particular are you loking to improve? :p
11:38* AutomatedTester makes note of jgraham's insubordination, especially since the end of the quarter is near
11:40jgrahamAh, so the moral is to not question authority
11:56jgrahamAnyone want to do some very quick test fixup review?
11:58AutomatedTesterjgraham: sure
12:02jgrahamAutomatedTester: Thanks!
12:02Tomcat|sheriffdutyDexter: ping
12:03DexterTomcat|sheriffduty, pong
12:03Tomcat|sheriffdutyhey Dexter
12:03Tomcat|sheriffdutyDexter: i'm hitting a merge conflict when merging m-i to m-c
12:03Tomcat|sheriffdutyi guess this is a trivial one so no need to backout etc
12:04Dexterlet me see
12:04Tomcat|sheriffdutybut could you tell me what i need to fix here :)
12:04Tomcat|sheriffdutyits in the .flake8 file
12:05Dextersure! So, it's not that bad (you scared me :-P)
12:05Tomcat|sheriffdutyits because landed before from autoland on central
12:09DexterTomcat|sheriffduty, the E111, E231, E261 and F841 rules should not be in the resulting .flake8 file
12:09Dexter(darn one line full of rules!)
12:09DexterIf you feel like it would be faster to backout, don't worry, back it out
12:10Tomcat|sheriffdutyDexter: so from this pastebin taking the working copy (from autoland to m-c merge) and removing this E things you mentioned or ?
12:10DexterTomcat|sheriffduty, yes!
12:10AutomatedTesterjgraham: done
12:14Tomcat|sheriffdutyDexter: \o/
12:15DexterTomcat|sheriffduty, awesome :D Thanks for doing that and for reaching out ;=
12:15Tomcat|sheriffdutynp :) if this are such trivial things i can fix them :)))
12:49jgrahamAutomatedTester: Thanks!
13:22maja_zfjgraham: is this what will fix the wpt harness, etc? Bug 1345490
13:22bugbotBug web-platform-tests, normal, nobody, NEW , Update web-platform-tests to revision 7c7db1b516b01394e19fad80094fbb81ab02ce75
13:24jgrahammaja_zf: Yeah c.f. the Nth try run for this update:
13:24jgraham(some updates are harder than others. This was not an especially easy one)
13:25jgraham(maybe the fact there's 20ish coommits in the branch is a clue there)
13:29maja_zfjgraham: yeah, wow. and there's still more to fix? I see actions tests are still disabled. Let me know if I can help with fixing them.
13:34jgrahammaja_zf: Oh I think that's more of an oversight, but I would prefer to land that in a different commit
13:34jgrahamI'm *hoping* this try run is the last
13:34jgrahams/commit/patchset/ or whatever
13:44jgrahammaja_zf: (btw artifact builds are a *huge* win for this kind of thing. It literally saves me days)
13:46maja_zf^ chmanchester
14:01maja_zfekyle: sorry for the delay. be there in a minute.
15:05Tomcat|sheriffdutyAutomatedTester: did the comment now in
15:05bugbotBug 1345904: General, normal, nobody, NEW , Frequent failures in windows 7 vm pgo tests with Automation Error: Received unexpected exception while running application
15:22jgrahammaja_zf: Landed geckodriver 0.15 and test fixups on inbound
15:23jgrahammaja_zf: Should be possible to reenable the actions tests now. Do you want to take that or shall I?
15:23maja_zfjgraham: i'm not here tomorrow + friday. you want to give it a try?
15:23jgrahammaja_zf: Sure
15:27jgrahamRan 44 tests (9 parents, 35 subtests)
15:27jgrahamExpected results: 43
15:27jgrahamUnexpected results: 1 (PASS: 1)
15:27jgrahamPromising start :)
15:54jgrahammaja_zf: opt looks good, debug is still a little broken
15:54maja_zfjgraham: debug never worked in the first place, right?
15:55maja_zfor hasn't in a long while
15:55jgrahamNot sure. I have a debug build and it works, but I guess my machine is quite a lot faster than AWS
15:55jgrahamSo I think it's just about speed
15:56maja_zfjgraham: yeah, tests have been disabled on debug before I did any work on them.
15:57jgrahamOK, well I'll try a --disable-optimise build and see if I can reproduce or just re-disable the offending tests on debug
15:57jgraham(is there a way to get just wdspec tests with try syntax?)
15:57maja_zfjgraham: i don't think so
15:58maja_zfjgraham: oh, just wdspec, yes there is
15:59* maja_zf looks at old try pushes
15:59maja_zfjgraham: try: -b do -p linux64 -u web-platform-tests-wdspec,xpcshell -t none --artifact
15:59jgrahamOh, that's straightforward
16:00jgrahamThe magic patch fairy should update trychooser
16:00* jgraham hopes the magic patch fairy is real
16:00maja_zfface: who is the magic patch fairy?
16:00* face biked 96km+ this weekend.
16:00jgrahamOr, I guess if my parents are willing to pretend, that works too ;)
16:00maja_zfalways just changes the subject when I ask about something important
16:06jgrahamOK, a disable-optimise build is plenty slow enough
16:42zwolIs this still the right place to ask questions about Marionette? I am wondering whether there is any way to detect _from content JS_ whether the browser was started with Marionette active. (I'm hoping there isn't.)
17:10AutomatedTesterzwol: are you asking if a website can see if its under automation?
17:16jgrahamzwol: With WebDriver you can see it, I dunno if marionette enables that by default
17:18AutomatedTesterjgraham: in html, an id can't have a space right?
17:18AutomatedTester for context
17:20AutomatedTesteryes, found it
17:25jgrahamAutomatedTester: That's an authoring requirement
17:26AutomatedTesterjgraham: that the author not use a whitespace?
17:30jgrahamAutomatedTester: Yes
17:30AutomatedTesteryea, cool
17:30AutomatedTesterI closed the bug
17:31jgraham suggests that implementations don't care
17:33jgrahamAutomatedTester: Yeah, UAs should just handle a space
17:34AutomatedTesterjgraham: "handle" ?
17:34AutomatedTesterso if we dont find the element that is fine?
17:36jgrahamdocument.querySelectorAll("#a\\ b")
17:36jgrahamNodeList [ foo#a b ]
17:36jgrahamIt should work
17:37jgrahamWe just need to make sure that it's escape correctly if it's in a CSS selector
17:42stephendhrm, is jgriffin out today?
17:43stephendah, thx
17:43stephendcamd: I'll be sure to checking with Jonathan too, but you don't have any qualms about me filing a "Move MozTrap to Graveyard" bug, do you?
17:46camdstephend: nope, that sounds great. Thanks! :)
17:46AutomatedTesterstephend: anything I can help with?
17:46stephendthx; I'll file and need-info him
17:46camdjgrifffin is on PTO at the moment and you probably saw he's resigning. Or perhaps not. :)
17:47stephendAutomatedTester: don't think so; just wanted to check in with him and ensure it's OK to move MozTrap bugs -> graveyard
17:47stephendoh, no, I did not....
17:47* stephend puts a note to buy and consume Scotch
17:47camdstephend: yeah. Huge bummer...
17:47camdYeah, i've been downing chocolate like there's no tomorrow...
17:48stephendTheobromine can only do so much
17:48camdof course... I always do that... but still...
17:48* camd looks for that chocolate bunny he stashed...
17:50AutomatedTesterstephend: if moztrap has been decommisioned it should be fine to put it all to graveyard
17:50* stephend nods
17:50AutomatedTesterstephend: happy to process the ni? if you want
18:29marcoekyle: could we schedule the meeting either thirty minutes earlier or one hour later? I have a conflicting meeting in the middle
19:30jgrahamIs anyone able to get a one click loaner from ? I get "insufficient scopes error"
19:31tedAutomatedTester: FWIW zwol is a former mozilla employee
19:31KWiersojgraham: works here
19:32jgrahamKWierso: I wonder what special powers you have
19:32jgrahamdustinm`: ^
19:37maja_zfjgraham: try signing into TC with the LDAP account that you use for pushes
19:39maja_zfan LDAP account that has L3 access should have "sufficient" scopes
19:40jmaherKWierso: I am trying to understand where these revisions originally landed at:
19:40KWiersojgraham: yeah, you need to sign in with okta
19:41KWiersoand it should tell you what scopes you're lacking
19:41jmaherKWierso: I think from autoland, but it isn't clear based on the commit messages, usually there is merge from autoland->mozilla-central message
19:41KWiersojmaher: sometimes you don't need to do a merge commit when pushing the changes around
19:41KWiersosomething with multiple heads
19:42KWiersobut those are from autoland
19:42KWiersothe commits about servo/ give it away
19:42KWiersoand checking one of the bugs included in the push should show you the original landings
19:42jmaherKWierso: well sometimes things land directly on m-c, so I want to make sure
19:42jmaherah, yeah- I could have done that- too many bugs today, will do that next time!
19:43KWiersogps's final plan has us get rid of merge commits entirely, I believe
19:43jmaheras long as I can find where things originated from and realistically map it to owners whom I can find in bugzilla for regressions, I have no complaints about that goal
19:47tedjmaher: do you not have enough information in the changeset to figure that out?
19:47tedwhy do you need to map it back to the repo where it landed?
19:48jmaherted: I need to pester patch authors if they cause performance or intermittent regressions
19:49tedisn't the author in the changeset?
19:49jmaherted: ifI have to bisect a 100 changeset commit to m-c, that is going to take a long time and waste resources, so I want to map it to where it originally landed so I can get the most narrow focus for retriggers, etc.
19:49jmaherted: yes, but for stylo that isn't obvious
19:50tedokay, that was the missing information :)
19:50jmaherI specified the minimum requirements, 5 years ago we could assume a lot of things, but today that isn't always the case
19:51jmaherI had to hunt around yesterday for a stylo patch author that had a +bugs in an email address, but that wasn't obvious
19:51tedyeah, i don't think we have a simple way to find out where a changeset originally landed
19:52tedhonestly the simplest way to find that is probably to scrape bugzilla comments for URLs from tbplbot or whatever it is
19:53jmaheryeah, assuming there is bugzilla links
19:53jmaherright now there are plenty to find, so it isn't too hard
19:54tedi guess stylo screws that up too since servo commits aren't going to have bugs
19:56jgrahammaja_zf: Ah, maybe my account is the wrong one
19:58tedchmanchester: can i punt the review on bug 893505 to you?
19:58bugbotBug Application Update, normal, dothayer, ASSIGNED , Simplify the application update UI
19:59tedit's just a single file they want review on but it's installing test support files from a different dir and i can't remember if that will work right with on-demand support file installation
20:00jgrahamI guess the issue here is I have no idea how to sign into ldap with my non-mozilla account
20:00jgrahamAlthough I suppose it must have worked at some point
20:03tedjgraham: i think the okta link will work?
20:06jmahererahm: ok, I am not sure how to land those awsy things via mozreview, the automation->land_commits is greyed out
20:07erahmjmaher: I do not understand mozreview
20:07erahmjmaher: did you r+ it?
20:08jmahererahm: yes, I just did after filing the followup bug
20:09erahmjmaher: the first one shows up as unreviewed...but I r+'d it
20:09erahmjmaher: in this view:
20:10jmahererahm: I am there, I don't see your r+ though
20:10erahmjmaher: it's here:
20:10erahmbut not on the main view
20:10erahmlet's go but #mozreview
20:24jmaherted: it appears that I have introduced a random orange (actually a couple fall here, osx debug and win8 debug wpt tests): , it is odd because I just added BUG_COMPONENTS to files (just validated the actual commits (I know I had to
20:24jmaherted: any idea why this would be the case? is there a clobber build I invoked or something else?
20:36AutomatedTesterrwood: hey
20:37AutomatedTesterabout the supposed marionette talos regression
20:37rwoodAutomatedTester: hello!
20:37AutomatedTesterI am highly dubious about change causing the regression
20:39jmaherAutomatedTester: we hear that often
20:39AutomatedTesterjmaher: it was changing an assert
20:40rwoodAutomatedTester: if you (or ato) could comment in the bug that'd be great
20:40jmaherAutomatedTester: that does sound not likely
20:40AutomatedTesterrwood: will do
20:40rwoodmaybe it was the push before, but the alert only triggered on the push after
20:41AutomatedTesterjmaher: rwood: did you know that perfherder, when you select pushlog takes my push to a merge
20:41AutomatedTester is the alert
20:44AutomatedTesterjmaher: rwood: updated bug 1347675
20:44rwoodAutomatedTester: thanks, yeah my mistake, sorry, I'm going to close the bug
20:44bugbotBug Marionette, normal, ato, NEW , 17.01% tabpaint (windows7-32) regression on push 78b008fb7d12ac75b2ad06d06ab65e6bc4ae8bb5 (Tue Mar 14 2017)
20:45AutomatedTesterrwood: no harm
20:45AutomatedTesterrwood: this is why I thought I would ask about this first here
20:49rwoodthanks for bringing it to my attention :)
20:54AutomatedTesterrwood: no worries
20:55chmanchesterted: yep, no problem
21:09KWiersojmaher|afk: triggered a bunch of jobs on the prior push and some failures popped up :)
21:13jmaher|afkKWierso: ah, that was you
21:14jmaher|afkthanks for the retriggers- I thought ~40 would be ok, but more was needed- and that specific push has now about 8 different intermittents related to it
21:28AutomatedTester 0:01.11 TEST_END: Thread-1 Harness FAIL, expected PASS. Subtests passed 28/28. Unexpected 1
21:28AutomatedTesterI think xpcshell may be drunk
21:35jgrahamAutomatedTester: That makes sense if the overall test has a different status to the subtests
21:36jgraham(it makes sense in wpt for example, not sure about xpcshell)
21:36AutomatedTesterjgraham: I updated the test for the right things and now getting this...
21:36AutomatedTesterI think I know what it is now, thinking about it
21:36AutomatedTesterbut the message could be more useful
21:42wlachjmaher: adding reftest and xpcshell retriggering is looking trivial
21:42wlachjmaher: I am guessing web platform tests will be the same
21:44mcotehuh, my post never made it to planet automation
21:45mcoteI can see it in my feed link on the side
22:12mitchhentgesWithin a taskcluster "test" task, what are "chunks"?
22:15Aryxmitchhentges: if running a test suite as one piece would take very long, the contained tests get put in chunks and these are run in parallel on different machines
22:16wlachdoes the xpcshell harness not support a --run-until-fail? should I add this option? thinking in the context of job retriggering
22:23Aryxwlach: bug 969605
22:23bugbotBug XPCShell Harness, normal, nobody, NEW , Add --run-until-failure to xpcshell-test command
22:25wlachAryx: you rock, thanks :)
22:30mitchhentgesAh, thanks Aryx. How does taskcluster know how to split up the the tasks? Isn't it just running a script (in xpcshell's case ||)?
22:30mitchhentgesOr does the script usually allow some parameter to partition the work?
22:32Aryxmitchhentges: e.g. xpcshell on desktop is one test for windows and osx and several for linux (and are it more than one for win 8 x64?)
22:33Aryxsee here and scroll down for the config scripts
16 Mar 2017
No messages
Last message: 11 days and 18 hours ago