mozilla :: #ateam

16 Feb 2017
08:57whimboojgraham: good morning
08:58whimbooso i think I found the problem from yesterday as you might already have read on bugzilla
08:58whimboothis gfx sanity check window is way nasty :(
10:59jgrahamwhimboo: I see. Do you have a plan?
11:01whimboojgraham: just addig lots of debug lines in marionette.
11:02whimbooso maybe the next try build gives some more details
12:44davehuntAutomatedTester, whimboo: has sphilp shown you his WD dashboard?
12:44davehuntarewewebdriveryet :)
12:46whimboonice one
13:12whimboojgraham: so its indeed the gfx test window which causes this test failure
13:30jmaher|afkjgraham: thank you for the comments about bugzilla components in bug 1338320; I appreciate the help
13:30bugbotBug web-platform-tests, normal, nobody, NEW , annotate testing/web-platform/ with BUG_COMPONENT for tests/ directory
14:31jgrahamMs2ger: Any idea when you will have time to look at the rest of the wpt leak checking review?
14:37jgrahamaka "oh, that one"
14:39jgrahamCan't actually land it until I figure out what's up in
14:39jgrahamMs2ger: ^
14:40Ms2gerjgraham, bad base revision?
14:40jgrahamCould be
14:46Ms2gerI don't suppose there's anyone else who knows this code
14:49tedahal: congrats on landing the mozbase test task!
14:54jgrahamMs2ger: Sorry
14:55jgrahamI mean jugglinmike is making progress, but I'm not sure I can ask him for r?
14:55Ms2gerr+ and let's hope this works, then :)
14:55jgrahamWell it seems to
15:24davehuntahal: thanks for all the reviews.. I believe that's it for now :)
15:26ahaldavehunt: np, thanks for cleaning up pytest-mozlog! I think it'll be useful for |mach python-test|
16:19jgrahamMs2ger: Apparently not
16:20jgrahamI think this is more related to --run-by-dir
16:22jgrahamI mean it's not implausible that shutting down the browser sometimes causes an assert
16:22* jgraham tries again without --run-by-dir in case the earlier leak went away by magic
16:40RyanVMjmaher: are we eventually going to be wanting to turn off linux32 BB jobs on 52 in favor of TC?
16:42jmaherRyanVM: if we get tc configs and tests running on 52
16:42RyanVMis that on anybody's list of priorities?
16:43jmaherwhen we finish higher priority work on osx/windows to meet deadlines regarding datacenter moves and available machines
16:43RyanVMI'm thinking about poking at it a bit too
16:43jmaheroh nice
17:05sgilesping: ahal
17:05ahalsgiles: hey
17:06sgilesahal: Thanks for the review! :) Are you able to land, I don't have access :|
17:06sgilesbug #1329528 for ref
17:06bugbotBug Mozbase, normal, sgiles, NEW , OSError: [Errno 1] Operation not permitted exception when killing a zombie process.
17:06RyanVMjgraham: ping RE: bug 1308969
17:06bugbotBug Document Navigation, normal, nobody, NEW , Intermittent /html/browsers/browsing-the-web/history-traversal/browsing_context_name.html | Retaining on history traversal - assert_equals: expected "test1" but got "test"
17:07ahalsgiles: sure!
17:08ahalsgiles: landed, thanks for the patch
17:09sgilesahal: Perfect, thanks!
17:19jgrahamRyanVM: I feel like the wpt upgrade I'm doing is supposed to help with that, but I'm not sure
17:19RyanVMthis is strictly about 52 at this point
17:19RyanVMcan we backport a fix or should I just disable the test?
17:35jgrahamRyanVM: Oh, right I fixed that locally earlier
17:36jgrahamRyanVM: The fix looks backportable to me
17:36RyanVMi tried and got a bit tangled up in manifest stuff
17:36jgrahamRyanVM: Ignore all the manifest changes and run |mach wpt-manifest-update| to generate a new one
17:37jgraham(unless that isn't working on 52 in which case |mach web-platform-tests --manifest-update NO_TEST| will do it
17:55yzenAutomatedTester hi, our intern, npang is looking into using marionette to run some tests related to a11y (something related to my questions about out of process stuff couple of days ago). Is there a way to have a python module outside of marionette and then be able to import it in test? would it need to be added somewhere for the build steps ?
18:39RyanVMjgraham: remind me again what the fix for is?
18:44jgrahamRyanVM: |mach wpt-manifest-update|
19:02wlachted: btw I don't think we would have caught bug 1339673 with perfherder without your cleanup work to seperate out the build times by aws configuration, so glad you did that :)
19:02bugbotBug Untriaged, normal, cmanchester, NEW , 6.07% build times (linux64) regression on push cff9ca695cca7bf21d6c628f2007c28e778d42f1 (Tue Feb 7 2017)
20:01AutomatedTesteryzen: what module?
20:12yzenAutomatedTester sorry package, not module, we want to be able to test query platform a11y stuff out of firefox (using comtypes package, I believe)
20:20AutomatedTesteryzen: and you want to have this in tree?
20:20tedwlach: yeah, that's good stuff!
20:20tedhaving useful data proves useful, hooray!
20:21tedwlach: and you know, when i first read that bug i was ready to say it wasn't a problem but turns out we are doing dumb things
20:22AutomatedTesteryzen: if you want this in tree you will probably need to speak to releng/build peers about getting the package in tree or at least into our Pypi. And then running tests should simple as getting a test running going picking the tests and making sure they work properly
20:23yzenAutomatedTester and if out of tree , is there a way have those tests included in our CI ?
20:24AutomatedTesteryzen: by in tree, I mean the tests are in tree
20:24AutomatedTesteryzen: if you want it to report to treeherder, thats easy
20:24AutomatedTesterthe treeherder folk have a library that allows you to report test results and upload logs etc
20:29wlachted: yeah I am curious to see how often we catch things like that, but this is a good early sign :)
20:30tednice to see that we can spot a regression even in the noisy build times graph
21:07yzenAutomatedTester yeah the tests would ideally be in tree
21:08AutomatedTesteryzen: so, that sounds like your package needs to be in tree (a build peer will need to ok that) or it will need to be uploaded to our pypi (releng can do that for you) and then you just need to set up a TC job
21:08yzenAutomatedTester sounds good, ill ping someone thanks
21:08yzennpang ^
21:09AutomatedTesteryzen: it should be easy to get that setup
21:09AutomatedTesteryzen: the taskcluster team have great docs for setting up the jobs
21:32RyanVMjmaher: FYI, the M(oth) split is in production
21:32RyanVMif there's TC-related changes that need making, let me know if I can be of help
21:33RyanVMAryx, KWierso: don't be shocked if you get some orphaned jobs from ^
21:33jmaherRyanVM: I notified wcosta about it this morning; tc stuff would be for osx right now, but we need to ensure for win7; if you could look at taskcluster/ci/test/tests.yml and remove the mochitest-other and ensure mochitest-chrome|a11y are set for windows, etc.
21:34jmaherRyanVM: btw, congrats on getting that landed
21:34RyanVMthanks :)
21:34RyanVMone less long pole!
21:34* wcosta took note
21:34RyanVMhope it doesn't make our current OSX issues that much worse, though
21:38RyanVMwcosta: i'm editing test-sets.yml at the moment, should I just make the OSX change too? (mochitest-other is commented out anyway)
21:39jmaherRyanVM: I get the impression non-e10s debug on osx will be disabled soon
21:39RyanVMa11y/chrome better not be since they don't have an e10s equivalent
21:39RyanVMthey're in the xpcshell boat
21:39wcostaRyanVM: feel free to do it, next rebase I test it
21:39RyanVMwcosta: ok
21:42RyanVMjmaher: that proved to be exceedingly easy since win/osx references to mochitest-other were all commented out anyway :P
21:42RyanVMi'm pushing it to try quick just to make sure it doesn't blow up the decision task too
21:53RyanVMwcosta: regarding the BBB comment, do you need to look into that before this patch lands?
22:04wcostaRyanVM: BBB is disabled, so no worries, I fix any mistake when I get back to it
22:05RyanVMgreat, then I'll go ahead and push this to inbound :)
22:05wcostawell, BBB is disabled on osx, but not to Linux talos
22:06RyanVMok, shouldn't matter for that I'd think
17 Feb 2017
No messages
Last message: 5 days and 2 hours ago