mozilla :: #ateam

17 Feb 2017
00:18erahmIs there a way to call the |marionette| cli from a python script? as in something like |marionette --binary firefox/firefox <other args> test_memory_usage.py|
00:28erahmmaja_zf|afk: ^ I think you might know?
02:55allassoHi, on treeherder results, there are orange test failed results that appear as an orange rectangle, and there are test failed results that are only orange text with a star. What is the difference?
03:27mcote|afkallasso: the ones with a star have been classified, that is, there&#39;s some sort of known cause or correlation
03:28mcote|afkallasso: see https://treeherder.mozilla.org/userguide.html (which admittedly isn&#39;t very verbose)
03:29mcote|afkallasso: if you click on a starred one, you&#39;ll see a couple lines of notes on the left side of the bottom panel, something like &quot;intermittent 1338992&quot;.
03:43KWierso|afkallasso/mcote|afk: if the star is hollowed out, the autoclassifier thinks it know the bug to assign it to. A filled in star means a human has looked at it and classified it for real
03:57maja_zf|afkerahm: it looks like you want to call the marionette test runner? Can do it through here: https://dxr.mozilla.org/mozilla-central/source/testing/marionette/harness/marionette_harness/runtests.py
03:57maja_zf|afkerahm: example usage: https://dxr.mozilla.org/mozilla-central/source/testing/marionette/mach_commands.py#33
04:05allassomcotelafk, KWiersolafk: thank you very much
10:19Ms2gerglob, is logs.glob.uno supposed to support https?
10:29jgrahamOK, I don&#39;t understand reviewboard
10:31jgrahamhttps://reviewboard.mozilla.org/r/111308/diff/6/ shows me a diff for a whole set of changes and tells me I last reviewed revision 3
10:31jgrahamBut if I use the slider to view revisions 3-6 I get
10:31jgrahamhttps://reviewboard.mozilla.org/r/111308/diff/3-6/
10:32jgrahamWhich suddenly contains some unrelated changes
10:39whimboojgraham: thats the consequence of a rebase
10:46jgrahamOh, that seems pretty broken
11:00Ms2gerHa
11:00Ms2gerHaha
11:38jmaher|afkMs2ger: hey, thanks for the feedback on my patch in Bug 1339232
11:38bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1339232 DOM, normal, jmaher, ASSIGNED , annotate more dom/* moz.build files with BUG_COMPONENT
11:38Ms2gerNp
11:39jmaher|afkMs2ger: I got mixed signals on the typedarrays, the recent comment from you was core::javascript (which isn&#39;t a real component), so I used Javascript engine, I can make that standard library
11:39Ms2gerhttps://bugzilla.mozilla.org/show_bug.cgi?id=1339232#c9 ?
11:39bugbotBug 1339232: DOM, normal, jmaher, ASSIGNED , annotate more dom/* moz.build files with BUG_COMPONENT
11:40jmaher|afkMs2ger: oh, I got the first comment, not the last one
11:40jmaher|afkgot == saw
11:41jmaher|afkall clear now :)
11:41Ms2gerI realized it just as I hit &quot;submit&quot;, of course :)
11:41jmaher|afkI will update the patch for the web-platform/tests as well
11:41Ms2gerThanks
11:42jmaher|afkthat is what I get for reading the needinfo cancelled messaged and then seeing overholt&#39;s bug work as the other stuff instead of the entire thread
11:44jmaher|afkMs2ger: a lot of this is guess work or looking at bugs that changed code and what components they were in
11:44jmaher|afkso I am always happy to get more opinions :)
11:44Ms2gerI&#39;m glad it&#39;s not my job :)
11:45jmaher|afkwell, I sort of thought it would be good to do and said I would do it this quarter
11:45Ms2gerAnd then people hold you to that, unfortunately :)
11:46jmaher|afkheh
11:46jmaher|afkI think pass1 of 2 or 3 is realistic this quarter
11:46jgrahamMs2ger: FWIW the existing typed array bugs seemed to be in Core::Javascript Engine when I looked
11:46Ms2gerMm
11:47Ms2gerJS: Standard Library is not very old
11:47Ms2gerI do think it&#39;s more appropriate, though
11:47jgrahamOK
11:47jmaher|afkok, lets go with that
11:47jmaher|afkthanks for the input!
11:48Ms2gerNot that any of them really matter, as long as the right people get the bugmail :)
11:49jmaher|afkwell jorendorff is the triage owner for both javascript engine and javascript: standard library
11:50jmaher|afkso today it wouldn&#39;t matter, but maybe in the future it would :)
12:32whimbooato: hello. i assume you feel better!
13:20whimbooTomcat|sheriffduty: hi. do you know something about log entries like &quot;No birth date. Using mtime.&quot;?
13:20whimbooi&#39;m getting spammed with this with a recent build
13:20whimbooits happening for all profile related files
13:22Tomcat|sheriffdutywhimboo: no idea
13:24whimbook, i may hvae to check if that is a regression
13:40whimbooato: i have goBack somewhat working locally but have some questions
14:09whimbooato: if you have some time it would be kinda appreciated given that I won&#39;t be around all the next week. thanks
14:30whimbooTomcat|sheriffduty: its clearly a regression. I will do a regression test
14:30whimbooworks fine with a commit from Monday
14:34Tomcat|sheriffdutyok
15:15whimbooTomcat|sheriffduty: bug 1340549
15:15bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1340549 Location Bar, normal, nobody, NEW , Gecko log spam for all all files inside the profile with: &quot;No birth date. Using mtime&quot;
15:15whimboofzi
15:46atomaja_zf: Oh dammit. I pushed to https://reviewboard.mozilla.org/r/109672/ in the middle of a rebase session and lost the commits you had previously r+ed.
15:46atomaja_zf: Because mozreview doesnt let me reset the r+es, I would be obliged if you could when you have time go through them (again).
15:47maja_zfato: any new changes to look for?
15:48atomaja_zf: No, none.
15:48smacleodato: If you inspect what you&#39;ve pushed before publishing (this is why the draft state exists) you can avoid cases like that
15:50atosmacleod: Thats not clear with the new Always Show All Commits option.
15:50atosmacleod: But I shouldve been more careful.
15:50maja_zfato: r+++
15:50smacleodato: It shows you what you&#39;ve pushed on the command line?
15:51atosmacleod: Right, and normally I use that but in this case I wanted to append a message to the commit (which it apparently doesnt allow you to do, I discovered), so I entered n to look at the UI instead.
15:51atomaja_zf: Thanks!
15:52smacleodato: ah okay, ya, there is a bug for the &quot;Always show all commits&quot; cookie not lasting between sessions
15:53atoAnyway, Im not trying to put blame on mozreview here. I did a stupid thing.
15:53smacleodato: I&#39;m just trying to suggest ways you can avoid it in the future, not defind MozReview
15:54smacleoddefend*
16:07dklALL: requesting feedback - anyone like or not like a more centralized layout of bugzilla? https://bugzilla-dev.allizom.org/show_bug.cgi?id=35
16:07bugbotBug 35: XFE, minor, mcafee, VERIFIED WONTFIX, Navigator does not free preference hash table when exit.
16:10atosmacleod: If you want to talk about safety, the best option would probably be to let `git mozreview push` refuse submitting anything to mozreview if it is currently in a rebase session.
16:10smacleodato: that&#39;s a good idea, can you file a bug please?
16:11wlachahal: awesome work re: the standalone python tests
16:14ahalthanks!
16:15atosmacleod: Sure! https://bugzilla.mozilla.org/show_bug.cgi?id=1340569
16:15bugbotBug 1340569: Integration: Git, normal, nobody, NEW , Make `git mozreview push` error when in rebase session
16:16AutomatedTesteris mozbase out of tree now?
16:16smacleodthanks
16:19wlachAutomatedTester: lol that would bring us full circle
16:19AutomatedTesterwlach: I saw your bug about mozdevice
16:19wlachAutomatedTester: no, ahal just moved the tests out of the build
16:19AutomatedTesterand that was my first thought
16:19AutomatedTesterahhh
16:19ahalAutomatedTester: haha.. please don&#39;t even plan on moving mozbase out of tree :p
16:20wlachAutomatedTester: honestly I would kill that mozdevice code in question with fire if I could
16:20wlachAutomatedTester: but I don&#39;t want to accidentally break something
16:20AutomatedTesterahal: I have no plans like that, wlach put mozdevice was out of tree in a bug
16:20AutomatedTesterand I was
16:22ahalwlach: if there&#39;s code in mozdevice not being used by anything in-tree, I&#39;d be ok with removing it
16:22ahalif there are out of tree consumers using it they can continue using the pypi releases (and if they need to make modifications, they can vendor in the old version that still contains the code)
16:23wlachtrue
16:23atowlach: I did a one-line change to populate an internal property of something in mozprocess once, so that subprocess would report the right value in the event of some special process exit event, and ended up breaking the universe.
16:24wlachato: yeah it&#39;s that sort of thing I&#39;m always afraid of with this sort of thing
16:24atowlach: So I sympathise with your situation )-:
16:25wlachted: is steeplechase still a thing? does it use devicemanager?
16:26wlachoh ffs
16:26wlachhttp://searchfox.org/mozilla-central/search?q=devicemanagerSUT
16:26wlachthere is no way I am touching this with a ten foot pole :)
16:26ahalwlach: I bet you could remove devicemanagerADB
16:26atoHah
16:26ahalit&#39;s probably all for b2g stuff that doesn&#39;t even run anywhere
16:27wlachahal: I believe devicemanagerADB is still used by android mochitest and friends
16:27atoahal: I think the Mn harness uses that for Android.
16:27ahalah
16:27ahalI thout android as all sut
16:27atoBut it also is terrible for lots of different reasons.
16:27ahalI guess the emulator would use adb
16:27wlachideally we would move everything to bc&#39;s new framework
16:28wlachahal: I don&#39;t think anything uses sut anymore, but we never took out the code to execute stuff in the harness
16:28wlachs/execute stuff/uses it/
16:28atoCalling adb is not synchronous, so it happens that it returns before the action has taken place.
16:29atoAlso I wish it had better error messages for when adb does return with an error. Right now it just looks at the exit code of adb and throws an exception. It could be more clever than that.
16:30wlachato: bc wrote a better replacement for devicemanagerADB which (IIRC) solved issues like that
16:30wlachhttp://mozbase.readthedocs.io/en/latest/mozdevice.html#adb
16:31atowlach: At a cursory glance, that looks like a much better API.
16:31wlachato: it totally is
16:32* wlach fighting urge to dive into tree and start ripping stuff out
16:32atoI will cheer you on (-:
16:33ahaldo it! do it!
16:33wlachI will file a bug
16:35AutomatedTesterdavehunt: for the geckodriver firefox lookup via path
16:35AutomatedTesterdo I just need /Applications/FirefoxNightly.app/
16:35AutomatedTesteror full path to the firefox binary?
16:36AutomatedTesterahh got it working
16:44wlachstandups: resisted urge to shave yak and filed bug 1340584 instead
16:44standupsOk, submitted #42590 for https://www.standu.ps/user/wlach/
16:44bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1340584 Mozbase, normal, nobody, NEW , Remove all references to devicemanagerSUT
16:45tedwlach: i&#39;m not sure we&#39;re still running steeplechase anywhere, but you can ask dminor
16:45tedhe&#39;s dealing with all the media automation
16:47ahaldeepasrijit1105: hi, welcome!
16:48deepasrijit1105ahal: Hi, thanks. I will hang around here now onwards :-)
16:50ahalawesome, if you ever have questions feel free to ask in here too
17:04deepasrijit1105ahal: I tried going through https://bugzilla.mozilla.org/show_bug.cgi?id=1306122, I didn&#39;t quite understand though. Is it possible for a small explanation?
17:04bugbotBug 1306122: Lint, normal, nobody, NEW , [mozlint] Create a &quot;compact&quot; formatter similar to the one in eslint
17:12* bc has resisted the temptation to standardize on adb.py and friends... so far.
17:13ahaldeepasrijit1105: of course, so when you run ./mach lint, there is a -f/--formatter option you can pass in that changes how the results are displayed
17:13ahalthat bug is basically about adding a new formatter called &#39;compact&#39;
17:13deepasrijit1105ahal: oh ok, got it
17:13deepasrijit1105I will play around with this
17:14ahalif you&#39;ve ever used eslint, it&#39;s supposed to look similar to eslint&#39;s &quot;compact&quot; formatter
17:14ahalyou could try installing eslint and running it to see what it&#39;s like (or finding an example of it in the eslint docs maybe)
17:14ahaldeepasrijit1105: thanks, do you want to be assigned to the bug?
17:15deepasrijit1105ahal: please, that would be helpful
17:17maja_zfahal: re &quot;It&#39;s safe to ignore because mozinfo.json will still be found via the &#39;dirs&#39; variable which
17:17maja_zfgets passed in from the marionette harness.&quot;
17:18maja_zfahal: are you referring to https://dxr.mozilla.org/mozilla-central/source/testing/marionette/harness/marionette_harness/runner/base.py#34
17:18maja_zf?
17:18ahalit was in the actual MarionetteTestRunner harness that the error was happening
17:19ahale.g when the harness tests instantiate those for their testing purposes
17:20ahalmaja_zf: tbh, I don&#39;t understand the failure super well.. if you like, you can try commenting out that except block and re-running the harness tests to see it
17:21ahalit was hitting some edge case in mozinfo lookup
17:22maja_zfahal: I&#39;ll take a look now. My first thought is that it would be better to modify the failing mnh test rather than mozinfo itself.
17:22ahalmaja_zf: actually, I think it must have been that call
17:22ahalmaja_zf: though, I think this is actually a bug in mozinfo
17:22ahalso I&#39;d still say we should fix it there
17:25ahal(it&#39;s possible both need to be fixed)
17:26ahalbut from a mozinfo point of view, if searching via the build system fails, it should still search via the *dirs variable afterwards
17:26ahalso it shouldn&#39;t be raising
17:28maja_zfahal: after a ./mach clobber i can&#39;t run the tests at all. is that unexpected? do I have to run ./mach configure to get the dependencies??
17:29ahalmaja_zf: have you pulled in the latest central?
17:29maja_zfahal: yes + hg pull your mozreview patches
17:29ahalI&#39;d expect |./mach clobber && ./mach python-test testing/marionette| to work
17:29ahalhm
17:29ahallet me check
17:30maja_zfahal: it&#39;s not recognizing --log-tbpl (the pytest-mozlog arg)
17:30ahalhuh, it seemed to work when I tested
17:31maja_zfi&#39;ll try again, now with extra care
17:31maja_zf:D
17:31ahalit&#39;s probably a bug I missed
17:31maja_zf(it works fine with an obj-dir)
17:32ahalmaja_zf: ok, I see it too
17:32ahalgive me an r- and I&#39;ll upload a new patch when I get a chance
17:33ahal(sorry about that)
17:33maja_zfno problem!
17:40mcoteopinion time
17:41maja_zfuhoh
17:41mcotewhat do people think about a centered layout for BMO? e.g. https://bugzilla-dev.allizom.org/show_bug.cgi?id=78678
17:41bugbotBug 78678: MailNews: Message Display, normal, hyatt, VERIFIED DUPLICATE, File | Attachments | ... will crash
17:41mcoteit&#39;s more in line with modern design
17:42mcotethe fact that the start of the second column in the info tables is centered is a bit weird, but overall I think it&#39;s an improvement
17:42tedmcote: doesn&#39;t really offend me
17:42tedjust feels different
17:42dmosei&#39;m using doing some project work on pine, and trying to cancel jobs using treeherder
17:42dmosefrom what i can tell so far, cancel doesn&#39;t have any effect
17:42dmoseis that known?
17:42tedeither way i have a ton of whitespace on a maximized browser window
17:42mcote&quot;doesn&#39;t really offend&quot;: check
17:42tedheh
17:42mcotehonestly what else could I aim for in BMO ;)
17:42maja_zfmcote: meh, fine either way
17:42tedi mean, i don&#39;t see anything that makes it worse than the existing left-aligned layout
17:43mcotey&#39;all are brimming with enthusiasm
17:43mcoteanyway as long as no one thinks it&#39;s worse
17:43maja_zfmcote: it looks weird and I can&#39;t deal with any sort of change!!
17:44maja_zfsomething something my workflow
17:44mcoteI&#39;m used to clicking within the left 100 pixels when I attach a file!
17:44mcotethis is UNACCEPTABLE
17:48bcit does feel a little weird but it does get us away from the tabs on the left when using tab center. less chance of having them pop out when you mouse left.
17:54mcotegood point
17:55mcotebut I&#39;ll put that down as another &quot;no worse&quot; comment :)
17:56wlachdmose: taskcluster jobs or buildbot jobs?
17:57wlachdmose: if they&#39;re taskcluster jobs I would ask in #taskcluster
17:58wlachmcote: I like it better
17:59wlachI don&#39;t see how anyone could really complain about it
18:00mcoteoh I forgot you&#39;re new at Mozilla
18:00mcote;)
18:01dmosewlach: ok, will do. thanks
18:11allassoHi, trying to cancel a job on treeherder. I am logged in; when clicking the cancel button, am prompted to enter username and password for LDAP account. I am pretty sure I set one up, but login creds aren&#39;t working. Is there a way to retrieve this?
18:11wlachdylan: thanks for moving bug 1331305 along, this has been frustrating me since forever(I was about to write up a patch when I saw that someone else had already done so)
18:11bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1331305 User Interface: Modal, normal, kuoe0, NEW , Auto resize the height of textarea of &quot;add comment&quot;
18:12dylanwlach: thanks for commenting, which made it raise to my awareness :-)
18:14wlachallasso: that sounds like an LDAP accounts issue, I would ask on #moc
18:14allassowlach - thanks !
19:45mystorRunning `mach mochitest -f plain --total-chunks 5 --this-chunk 4 --disable-e10s` on windows for me fails, because it seems to need to fetch and build twisted, which fails complaining that Microsoft Visual C++ 9.0 is required. What should I do / who should I talk to?
19:45mystorRunning individual mochitests works fine
19:51wlachmystor: ahal and/or ted might know
19:56mystorwlach: Thanks :)
20:00ahalhuh, I thought it would only require twisted if you run --subsuite media
20:01ahalmystor: is there a particular reason you&#39;re passing in chunks? What happens if you just pass in the directory you care about?
20:01mystorahal: I was trying to emulate the M(4) task on treeherder
20:01mystorahal: Running just the directory I would normally be interested in isn&#39;t reliably reproducing the leak
20:01mystor(A.K.A it isn&#39;t reproducing it)
20:03ahalfyi, mochitest restarts the browser between directories in automation too
20:03ahalso even if you do get the chunking exact, I&#39;d be surprised if that helped reproduce anything
20:05mystorahal: It does? darn
20:06mystorahal: :&#39;(
20:06mystorI guess I just can&#39;t reproduce the leak locally
20:07ahal(if you&#39;re still interested in solving the twisted issue, the full error would help.. and some googling yielded some potential fixes)
20:08atosmacleod: So apparently, if you commit a file in git that is ignored by .hgignore, it appears in your git commit but is lost when it is pushed to mozreview.
20:08mystorahal: Sure might as well
20:08atosmacleod: Wait, I take that back. It is ignored by .gitignore too, so Im running into another issue.
20:10mystorahal: http://pastebin.com/eJPY4mvA
20:11mystorRunning in a mozilla-build\start-shell-msvc2015 shell
20:11ahalmystor: I guess you have to install http://www.microsoft.com/en-us/download/details.aspx?id=44266 ?
20:12mystorahal: I&#39;ll try that, tnaks
20:12ahalI&#39;ve never tried installing it on Windows before
20:12mystorthanks*
20:12ahalif you get stuck again, maybe someone on the media team has hit this before (as typically, that&#39;s only required to run the media tests)
20:13mystork
20:17AutomatedTesterato: the geckodriver change is r+
20:18tedyuck
20:18tedis there not a wheel package available for twisted?
20:18tedcompiling native bits of python packages is for the birds
20:20tedi guess not for that version that&#39;s like 6 years old: https://pypi.python.org/pypi/Twisted/10.2.0
20:21mystorted: :P
20:22atoAutomatedTester: Thanks! Im going to wait landing it until I can get https://reviewboard.mozilla.org/r/113504/ to compile on the build slaves.
20:22atoAutomatedTester: Might have to make some more modifications depending on how that goes.
20:23tedit looks like that&#39;s code written by a mozilla employee, so presumably we could fix it to work with a newer twisted that has a wheel package
20:23tedso you could avoid this whole mess of compiling it
20:24tedthis code, i mean: https://dxr.mozilla.org/mozilla-central/source/testing/tools/websocketprocessbridge
20:38ahalyeah, bwc on the media team wrote it.. not sure if it specifically needs that version or not
20:38ahalmight just need to update the requirements file to a newer version
21:28wlachhave a great weekend all
18 Feb 2017
No messages
   
Last message: 159 days and 17 hours ago