mozilla :: #ateam

13 Sep 2017
09:18atowhimboo: So regarding, of course there is the potential of a race condition?
09:18bugbotBug 1397007: geckodriver, normal, nobody, NEW , Intermittent /webdriver/tests/ |
09:19atoI dont think you can synchronously guarantee that all state in DOM is updated when a widget layer even fires unless you lock up main thread, which would seem like a terrible idea+
10:21marco1jgraham: ping
10:21jgrahammarco1: Hi
10:22marco1jgraham: regarding the generation of coverage reports per test suite, were you interested in looking at line-level coverage for wpt? or only per-file / per-directory / overall?
10:25jgrahammarco1: line-level seems more useful I would have thought
10:26jgrahamI mean, I'm guessing a bit, but it seems likely that if you're trying to work out how to write more tests that improve coverage, knowing which actual lines of code are covered helps a lot more than knowing which functions are covered
10:32marco1jgraham: so, I'm thinking of generating a weekly report per test suite that shows line-level coverage, overwriting it every week, and only storing per-file / per-directory / overall for seeing the trend week over week
10:33jgrahammarco1: That seems pretty reasonable
10:34jgrahamI feel like in general trends, metrics, etc. are less interesting than details
10:35jgrahamSo I don't object to those being lower resolution (or not existing really, but I guess other people have some use case)
10:51marco1jgraham: oh ok, I thought you were interested in trends to see e.g. if people were adopting wpt more than mochitest, etc.
10:51marco1if that's not the case, I can prioritize generating the reports
10:56jgrahammarco1: I hadn't thought of that :)
10:56jgrahammarco1: But I still don't think that line-level data is needed for trends
10:57jgrahamFor that kind of trend it seems like file-level data would be fine-grained enough
10:57jgrahammarco1: Basically I'm saying your plan is great :)
10:57marco1great ;)
11:38jgrahamI am always surprised at my ability to go "hey this code I've written seems pretty OK, I'm sure it will pass tests first time", despite years of evidence that that never happens, and if it does it probably means I accidentially made all the tests noops
11:46whimbooato: i will have a look at this bug soon
11:52whimboostandups: dev preview of nightly tester tools 4.0 (as webextension) is up on AMO now (
11:52standupsOk, submitted #50858 for
13:38davehuntekyle: are you in the office again today? active data appears to be unresponsive for me :/
13:38ekylejust got in
13:38ekyledavehunt: looking...
13:38davehuntekyle: perhaps we could sit down and go over some of my queries today?
13:39ekyledavehunt: sure, I would like that
13:45ekyledavehunt: back up, sorry about that, I did not automate the fix for this problem yet
13:45davehuntekyle: no problem
13:46davehuntekyle: let's see if we can sit together this afternoon
13:46ekyledavehunt: actually, there was a large number of spot nodes lost as we were speaking
13:47ekyledavehunt: so it will be slow while it rebuilds itself
13:50Mardakrwood|bbl: i extracted and added it and the extracted files but maybe we don't want the zip in that repository?
13:50jmaherMardak: is ok to be publicly available
13:50Mardakwell just that it's a large file ;)
13:51jmaheroh, maybe rwood|bbl has other thoughts
14:26AutomatedTesterwhimboo: is a disabled test?
14:26bugbotBug 1323387: Marionette, normal, nobody, REOPENED , Intermittent TestExpected.test_element_not_displayed_when_stale_element | DMError: Failed to remove connection forwarding.
14:36atojgraham: This is quite a pretty idiom:
14:36ato(No idea what the code does, but the returned () type.)
14:37whimbooAutomatedTester: shouldn't be. but let me see
14:38AutomatedTesterwhimboo: if its not can you see about closing the issue. We havent seen it in a while :)
14:39whimbooAutomatedTester: yes, already doing it
14:39whimbooAutomatedTester: and i can believe once I do, a failure will appear!
14:39AutomatedTesterwhimboo: probably :D
14:40whimbooAutomatedTester: have you also noticed issues with autofill of logins with auth0?
14:40whimbooat least with auth0
14:40AutomatedTesterwhimboo: yes but I also have issues with lastpass not working so...
14:40AutomatedTesterits all crap...
14:41whimbooAutomatedTester: k, i will check the autofill now.
14:43whimboojgraham: hm, i wonder why no tests are getting started
14:43whimbooone of those mystic things again
14:43atowhimboo: --artifact
14:43whimbooato: doesn work with wdspec tests?
14:43atoCorrect, because it needs to build geckodriver, right?
14:44whimboobah. :(
14:44whimbooall again
14:45AutomatedTesterwhimboo: you can do an artifact build and then change to a non-artifact build and do a configure then build ./mach build testing/geckodriver
14:45AutomatedTesterthats how I used to do it
14:46whimbooAutomatedTester: sorry, this was for try
14:46AutomatedTesterthat wont help then :D
14:46whimboousually i build geckodriver myself
14:47atoIt would be lovely if artifact builds would remember geckodriver binaries too.
14:47atoNo reason it couldnt?
14:47atoIts built on CI for Windows and Linux.
14:47AutomatedTesterato: I did speak to chmanchester about this in SF and he said there shouldnt be a reason not to other than we need to special case it
14:48atoAutomatedTester: Well presumably we are for certain other programs, such as certutil.
14:57whimbooato: fyi its better on bugzilla to write &quot;bug 1234&quot; because it gets auto-linkified and the summary shown when hovering
14:57bugbotBug CSS Parsing and Computation, normal, peterl-retired, VERIFIED FIXED, we need default style for the html 4 style tags
14:58whimboowith links like you are forced to click on those to see what&#39;s behind
14:58atowhimboo: Not in email. Besides, the URI is canonical.
14:58whimboowho cares about email. intelligent clients should also show a URL
15:01dylanI think he means mouse-over on bugzilla itself, but we can fix that.
15:04whimboodylan: ++
15:16davehuntekyle: this query keeps returning an error for me, am I doing something wrong?
15:16davehuntI&#39;m wanting a list of suites
15:28davehuntekyle: actually, it looks like it&#39;s not just that query
15:28ekyledavehunt: looking...
15:33dylanato, whimboo:
15:34ekyledavehunt: It looks like too many queries to activedata at once. but here:
15:35davehuntekyle: am I overwhelming activedata?
15:35ekyledavehunt: there is automation using active data too
15:36ekyledavehunt: a combination of long running queries, and many queries is causing a lack of available connections
15:37ekyledavehunt: this happens usually when someone is exploring new queries, like yourself; the new fields take a while to load into memory, so take longer than usual.
15:38ekyledavehunt: so just ping me when it misbehaves again; this is a limitation of old elasticsearch.
15:38davehuntekyle: k, thanks
15:57whimboodylan: any reason why I cannot select firefox 52, 53, or 54 in the target milestone dropdown?
15:57whimboodid we miss to leave them in?
15:57dylanThat sounds odd, I didn&#39;t remove them
15:57bugbotBug 1309231: Marionette, normal, nobody, RESOLVED FIXED, page loading test is failing in selenium
15:58dylanwhat product?
15:58whimbooi have also seen the same for teh firefox product
15:59whimboolooks like firefox is fine
16:00davehuntjmaher: browser_noopener.js takes between 1 and 30 minutes..
16:00davehuntthe 90th percentile is 20 minutes for the last week
16:01dylanwhimboo: while I look at when this changed, it will remain unavail -- but I&#39;ll restore those versions once I understand what happened.
16:01jmaherdavehunt: 30 minutes....whoa
16:02jmahershould we rename it to browser_norunanymore.js ? :)
16:02jmaherdavehunt: I suspect the longer runs are on debug or non opt/pgo platforms
16:05atowhimboo: We fixed, right?
16:05bugbotBug 1279854: Marionette, normal, nobody, UNCONFIRMED , File Upload - File not Found
16:06nalexanderahal: hey, I want to make |mach xpcshell-test| (and friends) output only JSON.
16:07nalexanderahal: |mach xpcshell-test --log-raw /dev/stdout| is really close, but still produces colored output in TBPL format as well. How do I turn that off?
16:07nalexanderahal: also, can I make this thing print actual JSON (i.e., the elements wrapped in an array) rather than fragments?
16:08ahalnalexander: try |--log-raw -|
16:08ahaland no, you&#39;ll need to process the output to get it into actual json
16:09ahalthe output is
16:10nalexanderahal: are strings containing newlines correctly produced as &#39;foo\nbar&#39; by JSON.dumps? I guess they must be for jsonl to work at all.
16:10ahalyeah, they should be
16:13* ahal heads on pto
16:46dylanwhimboo: looks like that happened as part of the normal version uplift.
16:55Mardakjmaher: ha ha.. doesn&#39;t like tp5n file names :( 09:49:27 FATAL - IOError: [Errno 22] invalid mode (&#39;wb&#39;) or filename: &#39;C:\\slave\\test\\build\\tests\\talos\\talos\\tests\\tp5n\\\\\\b\\ss\\uoluolprod,uoltotal\\1\\H.22.1\\?AQB=1&ndh=1&AQE=1&#39;
16:56Mardakmaybe there&#39;s a download without unpack to call some other unzipper
17:08jmaherMardak: was that on your try push?
17:09jmaheroh yeah, I see now
17:09Mardakjmaher: yeah. i&#39;m retrying with a zip that has the &quot;?&quot; file removed
17:09Mardakshould be fine for my purposes of testing rich icons
17:10whimboodylan: hm, any idea why this could have happened?
17:10whimbooato: let me look
17:12whimbooato: yeah. i closed it as WFM
17:12atowhimboo: Thanks!
17:17whimboodylan: is it only Marionette or everything under testing?
17:18dylanall of testing
17:19dylanI can re-enable it there
17:20whimbooato, AutomatedTester: so it&#39;s not artifact related
17:20whimboono tests are getting run
17:20whimboomaybe bad platform selection
17:20whimbooi think i will no longer stress try and just land.
17:20whimboodylan: would be great. thanks
17:21whimbooat least geckodriver can be build fine
17:21atowhimboo: s/linux/linux64/ I think.
17:22whimbooAutomatedTester: we want to do the 0.19 release tomorrow if possible
17:22whimbooato: yeah :(
17:22atowhimboo: Or s/web-platform-tests-e10s/web-platform-tests/
17:22AutomatedTesterwhimboo: ok cool
17:22atowhimboo: Latter is probably more likely.
17:22atowhimboo: Now that everything is e10s.
17:22whimbooato: do all wpt tests use geckodriver? i thougth the wdrunner is using purely marionette
17:23atoYes, thats correct.
17:23whimbooso Wd would be enough as I have selected
17:23atoBut Wds part of web-platform-test.
17:23whimbooits bad that i had to push twice and still no results :(
17:23atoTheres probably a subfilter for just Wd.
17:23whimboowhich i used
17:24whimbooi dont want to run any other tests :)
17:24whimbooonly those which use geckodriver
17:24atojgraham might know, but I suspect s/e10s//.
17:24whimboowould be good to get a warning
17:24atotry syntax sucks. \-:
17:24whimbooso its just wasted time
17:24whimboojmaher: ^
17:25whimbooin case you are interested
17:25atoIn this case you can use the job selector to trigger Wd though.
17:25AutomatedTesterwhimboo: what does ahal|afk&#39;s try fuzzer do?
17:25atoBecause the build already contains the geckodriver binary.
17:25whimbooAutomatedTester: i didnt try that. its just submitting from mozreview
17:25whimbooato: nope. Cannot login right now :(
17:26atoWell that sucks.
17:43dylanwhimboo: done, btw
17:48whimboodylan: 3>
17:48whimbooand it works
17:52whimbooato: your conflicts my patch for logging output on autoland :/
17:52whimboolooks like i have to wait until tomorrow to update and push again
17:52atoWell you can rebase.
17:52whimbooi have the unified repo
17:52whimboothere is no autoland branch
17:54whimbooato: btw. you had a merge conflict
17:54whimbooand missed to correct
17:54whimboo+- [mozrunner crate] upgraded to version 0.4.2
17:54whimbooit should be 0.5.0
17:55atoAh, I missed that during rebase.
17:55whimbooi will correct it when rebasing tomorrow morning
17:55atoBut well fix up the change log anyway (-:
17:55whimboosure, but easy to miss later
17:56whimbooAutomatedTester: you like filling our inboxes today! :D
17:56AutomatedTesterwhimboo: if you spot something that should be closed from your inbox go ahead and close it
17:56AutomatedTesterthat way you fill my inbox too :D
17:57whimbooi correct some of the closings
17:58atowhimboo: You can just pull down the raw file for amending the commit. You dont need to actually check out autoland.
17:59whimbooato: i would have to import it inot my bookmark, reorder via histedit, and push it together with the other commit to mozreview. landing won&#39;t work either
17:59whimbooi think it saves me the time to just wait :)
17:59atoI mean, replace the file from autoland tip, make your changes, commit.
18:00atoBut waiting is OK too (-:
18:00AutomatedTestermaja_zf: hey, the work that you and your contributor have done in actions, is still valid?
18:00bugbotBug 1306259: Marionette, normal, nobody, UNCONFIRMED , Extended key not correctly emitted for send_keys on an element
18:04whimboowow a new sublime release is out!
18:04whimbooafter a full year of inactivity
18:04maja_zfAutomatedTester: That bug is one level up from actions. It&#39;s more about the interpretation of element.send_keys than about actions, no? Doesn&#39;t seem relevant to any recent work or bugs
18:05AutomatedTestermaja_zf: oh right, I thought it might have been with the modifiers
18:09atoI had my first Firefox crash since ~forever.
18:09jmahergps: I recall an interface to HG web where I could get the bugzilla component for a given revision or file?
18:11jmahergps: thanks
18:11gpssome paths may hit server errors because we haven&#39;t upgraded some things in a while
18:11gpsif it is a problem, let me know
18:11gpsit is correctable
18:11jmaheroh cool, I will keep that in mind
18:16AutomatedTesterekyle: jmaher
18:33atowhimboo: I like the elegance of this change:
18:36whimbooato: indeed. so we are not using the TimedPromise in those cases?
18:37atoWell we could do, I suppose.
18:37atoBut are we expecting it to fail?
18:37atoMaximizeWindow we are expecting to fail in certain configurations.
18:39atoOne very nice improvement I came to think of was that we should have a WebElement abstraction in the future so that it could act as an EventTarget, e.g. &quot;let contentElement = this.listener.findElement(); contentElement.addEventListener(&quot;visibilitychange&quot;, )&quot;.
18:39atoA major caveat of how that patch is written is that it only targets the window global.
18:42whimbooato: i&#39;m only afraid of perma hangs
18:43whimboofor those commands we do not define any kind of timeout in teh spec
18:43whimbooi would even be fine with 30s or 60s
18:44whimbooso nothing necessarily <10s
18:44atoI think maybe we should try and fix that at a higher level.
18:44atoPerhaps in server.js somewhere if an execution takes too long?
18:45atoThere are more than just promises failing to resolve that could happen.
18:47atoIn fact we probably need a fix in there in any case: if a promise spends a very long time and we think it has timed out, but then suddenly springs back to life at a random later point, we need to ensure that it isnt allowed to send back another response.
18:47atoCant remember, but we may already have a check for duplicate responses.
18:48atoBut that used to be a real problem in the past (before message sequencing).
18:54whimboogbrown, jmaher: check bug 1399592 and smile!!
18:54bugbotBug Mochitest, normal, nobody, NEW , Mochitest runner is using a too short timeout for starting a Marionette session
18:54whimbooato: better in proxy.js which is using the promise
18:55whimboogbrown, jmaher: maybe one of you can add all the dependencies to the fialing bugs? I lost overview :)
18:56whimbooAutomatedTester: bug 1399592... this will fix all the trouble
18:56whimboowe have with mochitests
18:56whimbooso it&#39;s not sessionstore related i think
18:57whimboowe also need more logging for marionette driver
18:57whimboootherwise those things will be hard to notice
19:02jmaherwhimboo: fix all the things!
19:05whimboomaja_zf: how long will you be around or tomorrow? The patch on bug 1397675 is ready to review. May I ask your or David?
19:05bugbotBug Marionette, normal, hskupin, ASSIGNED , Immediately quit the application if no connection can be made during startup
19:06maja_zfwhimboo: i&#39;ll be here until 4utc tomorrow
19:06whimbooso you are confident with it?
19:07maja_zfwhimboo: yeah, I just might not agree with your changes ;)
19:07whimbooheh, those are workign!
19:07whimboowill do the request now. thanks
19:15jmahergps: I am not able to get info for a specific file:
19:15jmahergps: ./mach file-info bugzilla-compoent works for that one
19:18gpsjmaher: that&#39;s likely bug 1354791
19:18bugbotBug Mercurial:, normal, gps, REOPENED , Unable to find metadata for testing/firefox-ui/tests/functional/security/
19:19gpsif what you are doing is semi-important, i can fix this with not much effort i suspect
19:20* gps lunches
19:25AutomatedTesterout of the triaging i&#39;ve done today is the most interesting
19:25bugbotBug 1368486: Marionette, normal, nobody, UNCONFIRMED , Firefox53 crashes with Marionette and CSS position attribute
19:28jmahergps: not important, just exploring stuff right now
19:31whimbooAutomatedTester: mind attaching the stack of the crashing thread?
19:31AutomatedTesterwhimboo: It doesn&#39;t crash, it hangs
19:31AutomatedTesterI should update that
19:31whimbooso it hangs in sendKeysToElement ?
19:32AutomatedTesterand its not related to any specific character as far as I can see
19:35AutomatedTesterato: is your window work you&#39;re doing going to fix this issue?
19:35bugbotBug 1351940: Marionette, normal, hskupin, ASSIGNED , TabBar.get_handle_for_tab can throw &quot;TypeError: win.outerWindowID is undefined&quot;
19:36whimbooAutomatedTester: no
19:36whimboothis is after a restart
19:36whimbooand it misses the dependency
19:36whimbooi will add it
19:37AutomatedTesterok cool, thank you
19:37whimbooAutomatedTester: well, not really a dependency but same as bug 1387337
19:37bugbotBug Marionette, normal, nobody, NEW , Add support for not yet restored tabs after a restart of Firefox (sessionstore enabled)
19:37whimbooits just that the tab hasnt been restored
19:37whimbooand doesn&#39;t have a content browser
19:37whimbooonce you click on it, the browser gets created
19:38whimbooand a window id is available
19:38whimboothe problem with puppeteer is that it really uses the tabs to check how many of them are open
19:38whimboowhile marionette trusts the window handles
19:38whimboowhich in this case is wrong
19:38whimboothose tabs are just not usable at this time
19:45AutomatedTesterwhimboo: is this a simple case of switching stuff on? can we do this as a good first bug?
19:45bugbotBug 1389103: Marionette, normal, hskupin, ASSIGNED , Unskip temporarily skipped quit/restart tests due to bug 1054740
19:46whimbooAutomatedTester: nope. see the try run i did for my patch
19:46whimbooAutomatedTester: bug 1395504 is the most annoying problem for us right now
19:46bugbotBug Graphics, critical, nobody, NEW , Infinite hang of web content process when parent process crashes [@ CrashStatsLogForwarder::Log() | mozilla::dom::PContentChild::SendGraphicsError()]
19:47whimboowhereby it seems to not happen that often the last days
19:47whimboomaybe we I could rebase and try again
19:47whimbooplease put ni? on me
19:48AutomatedTesterof course
19:48AutomatedTesterI am not expecting you to do it, just thought if its reenabling we could do a good first bug and move on
19:49atoAutomatedTester, whimboo: Well if were no longer using outerWindowID it might help.
19:54whimbooAutomatedTester: those are important tests which got skipped due to a silly change which got even reverted
19:55whimbooato: we wont have a browser anyway in such a case, so maybe we an id but no underlying browser in marionette
19:55whimboofor puppeteer i can handle that differently
20:15whimbooi give up on bug mail today
20:15whimbooto be continued tomorrow! :
20:21AutomatedTesterwhimboo: sorry, nearly done
20:21AutomatedTester15 bugs to go
20:22AutomatedTesterato: Is this another one for your window work since the element map is tracked against the window ID?
20:22bugbotBug 1318843: Marionette, normal, nobody, UNCONFIRMED , Using SwitchTo results in StaleElement exception
20:27AutomatedTesterato: also... is still an issue for you?
20:27bugbotBug 1326227: Marionette, normal, nobody, NEW , TestScreenCaptureChrome.test_capture_flags fails on Linux
20:46emceeaichjmaher: I&#39;m now able to close all those intermittents
20:46emceeaichdylan landed a change that allows for a specified set of users to make silent closes
20:49dylantho&#39; careful with how much load you put on the server. :)
20:49emceeaichI have a script now that I&#39;m using.
20:49emceeaichit batches the changes
20:56jmaheremceeaich: oh great!
20:57jmaherthis is wonderful news
22:39atoAutomatedTester: I would need to spend some time investigating.
22:40atoAutomatedTester: Set a needinfo for me if its a priority.
14 Sep 2017
No messages
Last message: 7 days and 6 hours ago