mozilla :: #automation

16 Feb 2017
09:23AutomatedTesterwhimboo: if we move ondemand update tests to TC, what we do we need from releng again?
09:24whimboodepends on where should those be integrated
09:24whimboowe could keep it by triggering manually
09:24whimbooor integrate it into the release promotion code
09:24whimboothe latter might be more difficult
09:25whimboowe would need the logic to determine which builds to tests
09:25whimboowhich means current build - 3 releases, or latest esr
09:26AutomatedTesterok
10:55AutomatedTesterwhimboo: are you still going to be able to finish back/forward if you're investigating bug 1340099
10:55firebothttps://bugzil.la/1340099 NEW, nobody@mozilla.org GeckoDriver gets instantiated multiple times during startup
10:57whimbooAutomatedTester: i have the feeling that this might be causing bug 1315611
10:57firebothttps://bugzil.la/1315611 NEW, nobody@mozilla.org Ignore GFX sanity check window during startup of Firefox
10:57AutomatedTesterok, its not a priority though
10:57whimboowhich blocks the huge patch on bug 1322383 from landing
10:57firebothttps://bugzil.la/1322383 ASSIGNED, hskupin@gmail.com Chrome context methods (driver.js) miss checks for valid window - no NoSuchWindow failure thrown
10:58whimbooi just added some debug lines so i can trigger a new try build
10:58AutomatedTesterok
10:58whimboothis all takes a while
10:58whimbooand i cannot repro it on the builders at all
10:58whimboomanually
10:58whimboothis was just a side-effect what I have seen
11:47mhaouechwhimboo: Hello, how are you?
11:54whimboomhaouech: good day! good. just came back from lunch :)
11:54whimboowhat bout you?
12:03mhaouechwhimboo: I'm fine thanks, going for lunch in a moment :)
12:04mhaouechhow's work
12:09whimbootrying to get bug 1330348 fixed before the weekend
12:09firebothttps://bugzil.la/1330348 ASSIGNED, hskupin@gmail.com Make forward- and backward commands synchronous
12:11whimboomhaouech: have you been lucky with the first tests?
12:27mhaouechwhimboo: I had a problem with the initial tests
12:27mhaouechAt first I thought it's something in my code
12:27mhaouechbut then I cloned the mozdownload project and launched tox and there were the same errors
12:28mhaouechI think something is not working properly either in my tox or in the mozdownload project
12:28whimbooAutomatedTester: re goBack() / goForward() I wonder if we should add a check if that action is possible at all before we call history.back().
12:29whimbooAutomatedTester: waiting for a not thrown pagehide is funky
12:29whimboobecause we would need a timeout
12:29whimbooand it delays the execution
12:45whimboomhaouech: what sspecific failure do you see?
12:46mhaouechwhimboo: Can I send you the output of tox?
12:47whimboosimply do a pastebin
12:47mhaouechhttps://pastebin.mozilla.org/8979459
12:51whimboomhaouech: which version of Python are you using
12:52whimboodata_string = '20131001030204\r', format = '%Y%m%d%H%M%S'
12:52mhaouechwhimboo: Ah yes, I recently worked on a machine learning project so I switched to 3.5
12:52whimboowe do not support 3.x
12:53whimboonot sure where this \r is coming from
12:53mhaouechI'll fix that, I spent all day yesterday to understand the problem
12:53mhaouechthanks
12:53whimboobut well you run with Python 2.7.10
12:54whimboowe run our tests via appveyor on Windows
12:54mhaouechyes, I see that in the log. I can't understand where that \r comes from
12:54whimbooso they do not fail
12:54mhaouechso I need to install it or something?
12:55whimboodoes your git translate \n to \r?
12:57mhaouechI think when the scraper downloads from the specified url, it gets both \r and \n. For example: '20131001030204\r\n'
12:58mhaouechbecause I switched \n to \r in get_latest_build_date, and there's still the same problem but data_string became '20131001030204\r\n'
13:02whimboocan you access https://ci.appveyor.com/project/whimboo/mozdownload ?
13:02whimboonothing fails on appveyour side
13:04mhaouechI see that the changes that I recently made passed successfully
13:04mhaouechwhat should I do in order to use the appveyour locally?
13:17whimbooyou cannot use it locally
13:17whimbooits done automatically when you push something to your remote branch
13:17whimbooand the PR gets updated
13:18whimboostandups: started working on getting goBack/goForward synchronized (bug 1330348)
13:18standupsOk, submitted #42532 for https://www.standu.ps/user/whimboo/
13:18firebothttps://bugzil.la/1330348 ASSIGNED, hskupin@gmail.com Make forward- and backward commands synchronous
13:20mhaouechwhimboo: ok I got it. There was an error with TravisCI, it couldn't find a certain build, but I think now it's there, I've checked the directory containing the builds
13:20mhaouechSo what should I do now?
13:21whimboowas it not there for the changeset you had locally, or is it still not there on the remote repository
13:23mhaouechthe same test passed locally, so it could find the build
13:23mhaouechHow to check for the remote repository?
13:30whimboowell, you said it is even failing for you when you freshly cloned the repo
13:30mhaouechyes but it's not the same failures
13:32mhaouechTravis CI failed the test_firefox.py
13:32mhaouechbut it passed locally after cloning
13:42whimbooAutomatedTester: for step 3 i only see: Traverse the history by a delta 1 for the current browsing context.
13:43whimboothere is no abort mentioned
13:43AutomatedTesterwhimboo: click the link that takes you to the html spec
13:43whimbooi assume you mean an abort as mentinoed https://html.spec.whatwg.org/#traverse-the-history-by-a-delta
13:43whimbook
13:43AutomatedTesterso, in my comment I linked to the html spec
13:43AutomatedTesterwhich is also linked in step 3
13:44whimbooAutomatedTester: so what should it return? nothing?
13:44AutomatedTesterwhimboo: return that the command has executed successfully
13:46whimbooAutomatedTester: so it should be a boolean? i dont see anything in our implementation regarding a success value
13:47whimboousually we throw an exception
13:48AutomatedTesterwhimboo: it looks like in listener, just ending the command will automatically return sucess
13:48AutomatedTestersuccess*
13:48whimbooall the code will be in driver.js
13:49whimbooso we dont have to care about remoteness switching
13:49AutomatedTesterhttps://dxr.mozilla.org/mozilla-central/source/testing/marionette/driver.js#1030-1034
13:49AutomatedTesterthat will return success when we do the processing of commands
13:51whimbooAutomatedTester: the question here is what to do in case of eg no history items avaiable.
13:52whimboosorry for asking those questions but its not that clear
13:52whimbooi assume andreas won't be around today?
13:52AutomatedTesterwhimboo: we abort and return
13:53AutomatedTester"If the index of the current entry of the joint session history plus delta is less than zero or greater than or equal to the number of items in the joint session history, then abort these steps."
13:53whimbooAutomatedTester: abort and return == return statement and no exception thrown?
13:53AutomatedTesterno exception thrown
13:53whimbooi can read the spec but its unclear what Marionette server should call
13:53whimbooso we silently ignore this request
13:53whimboointereting
13:54whimboogiven that trying to click a disabled element we throw an exception
13:56AutomatedTesterit looks like we need to make sure that if pageHide is emitted in https://dxr.mozilla.org/mozilla-central/source/testing/marionette/listener.js#1128 we wait
13:56AutomatedTesterotherwise we dont wait and return
13:57AutomatedTestermake sense?
13:58whimboonope :)
13:58whimbooi still feel that we loose information with this logic
13:58AutomatedTesterok :)
13:58AutomatedTesterlose what info?
13:58whimboothat a navigation did not happen
13:58whimboothe consumer will not see that
13:59whimbooand a follow up command will fail
13:59whimboowhich eg. relies on an element of the expectd page
13:59whimbooand its hard to see why its failing
13:59AutomatedTesterwhimboo: not everything will have the page change
13:59AutomatedTesterclicking back could move you within a document
14:00whimboobut it should cuase dead objects for already retrieved elements, or? i think the document gets replaced
14:01AutomatedTesterI would be surprised if we replace a document for changing the hash via forward/back
14:02whimbooanything else beside hash changes
14:03AutomatedTesterif the page changes then we will return staleElement when people try use elements already found
14:04AutomatedTesterwe will either throw that if people try use elements or the element references wont be found
14:43whimbooAutomatedTester: so re loading from cache. we get DOMContentLoaded and pageshow. depending on the page load strategy we would have to return for one or the other
14:44whimbooif we always wait for pageshow we would not be able to handle eager
14:51AutomatedTesterwhimboo: we wouldnt be able to handle eager because its not spec'ed or?
15:05whimbooAutomatedTester: you mean the html spec?
15:09AutomatedTesterwhimboo: I am trying understand what you mean by 14:44 (whimboo) if we always wait for pageshow we would not be able to handle eager
15:11whimbooAutomatedTester: waiting for pageshow the readystate will always be complete
15:11whimboofor eager we would have to return already for DOMDocumentLoaded
15:11whimbooin the webdriver spec you wrote pagehide and pageshow
15:12AutomatedTesterperhaps we need to change it to DOMContentLoaded instead of pageShow
15:13whimbooi think we need both. it depends on whch page load strategy is set
15:14AutomatedTesterwhimboo: I suggest coding it for both and then we can see about updating the spec to match what you have coded
15:15whimbooAutomatedTester: i don't think we support page load strategies yet
15:15whimbooin pollForReadyState we always wait for complete
15:15whimboounless its an error page
15:15AutomatedTesterwhimboo: ok
15:16whimbooi only ask regarding the spec so we might be able to get this in before RC
15:16whimbooif it is necesary
15:16AutomatedTesterin that case, let's not worry about the page load strategy for your patch but leave a TODO that we need to add pageLoadStrategy changes here
15:16AutomatedTesterwe can add this into the spec before Rec
15:17whimbooo shall i file an issue now?
15:18whimbooAutomatedTester: which prio would bug 937659 be?
15:18firebothttps://bugzil.la/937659 NEW, nobody@mozilla.org Implement page loading strategies
15:20AutomatedTesterwhimboo: not that high, will set it now
15:26whimbooAutomatedTester: i also filed the spec issue
15:26AutomatedTester<3
15:26AutomatedTesterthank you
15:42whimbooAutomatedTester: what if pagehide is not firing? which timeout shoudl be used then
15:42whimbooalso page load timeout_
15:42whimboo?
15:43AutomatedTesterwhimboo: when would it not fire pageHide?
15:43whimbooi don&#39;t know
15:46AutomatedTesterwhimboo: the only way I can see us not firing is if its a hash change that occurs
15:47whimbooAutomatedTester: looks like, yes
15:47whimboo*** attaching event listeners
15:47whimboo******** Caught event: hashchange
15:47whimboo******** readystate: complete
15:49whimbooAutomatedTester: should this be a separate issue or should we combine all findings in the recently filed one? i would do the latter
16:12whimbooAutomatedTester: heh. given that about:blank is complex...
16:13whimboofor about:blank#34 => about:blank we get a pagehide
16:13whimboono hashchange
16:13whimboonice
16:13whimboomconley: hi. are you already around?
16:24mconleywhimboo: hi
16:24mconleywhimboo: what&#39;s up?
16:25whimboomconley: quick question regarding remoteness changes. is there a specific time when those happen? when I trigger a page load can I expect the pagehide, unload events always happening before a remoteness change?
16:26mconleywhimboo: I believe yes, you get pagehide and unload.
16:26whimboosounds not 100% sure :)
16:27whimboobut that would be great
16:27whimboomaybe i should check with some pages which trigger those changes
16:27mconleywhimboo: best way to know for sure is to try it. Been a while since I&#39;ve touched that stuff, which is why I&#39;m hesitant. :)
16:27whimbooheh
16:27mconleyyou should absolutely test and check
16:27whimboono worries
16:27whimboolet me try that and i will report back
20:49whimbooAutomatedTester: i have a wip for goBack()
20:49whimboowould you mind having a look at it jsut so I know it&#39;s the right direction?
20:50whimbooit generally works for both hash changes and real location changes
20:50AutomatedTestersure
20:54whimbooAutomatedTester: will upload the current state
20:54whimbookeep in mind lots of debug lines
20:54whimbooand not all implemented
20:54whimboolike timeout etc
20:56AutomatedTestersure
20:59whimbooAutomatedTester: https://bugzilla.mozilla.org/show_bug.cgi?id=1330348#c14
20:59firebotBug 1330348 ASSIGNED, hskupin@gmail.com Make forward- and backward commands synchronous
21:03AutomatedTesterwhimboo: looks like its definitely going in the right direction
21:03whimboogreat
21:04whimboowith the callback it can be any command
21:04whimbooalso your click and wait for page load
21:04AutomatedTesterooooo
21:05whimbooAutomatedTester: maybe I cannot finish it completely but by end of tomorrow we should have a close to be ready for review patch
21:05AutomatedTesterawesome
21:06whimboomaybe someone else has to finish it
21:06whimbooif we have to get it in by next week
21:06whimbooAutomatedTester: have you had a chance to talk with relman about uplifts to esr52?
21:06AutomatedTesterno, I forgot
21:07whimbooAutomatedTester: k
21:07* AutomatedTester sets a reminder to speak to Sylvestre tomorrow
21:07whimbooAutomatedTester: btw. page load strategy will also be easy to implement
21:19whimbooAutomatedTester: so back and forth are actually more complex!
21:19whimbooit depends on which page was loaded before
21:19whimbooso we can get pageshow for normal pages
21:20whimbooor interactive + erorr page
21:20whimbooand maybe others
21:20whimbooeverythign what could happen for a normal page load too
21:29whimbook, be back tomorrow
17 Feb 2017
No messages
   
Last message: 183 days and 23 hours ago