mozilla :: #ateam

19 Apr 2017
08:20whimbooTomcat|sheriffduty: good morning.. will we have another merge from autoland to mc before the next Nightly?
08:24Tomcat|sheriffdutylet me check
08:24whimbooTomcat|sheriffduty: i have a patch for update tests I kinda would like to see in the next nightly
08:24whimbooif we dont have maybe we can cherry pick?
08:25whimbooits for bug 1304656
08:25bugbotBug Firefox UI Tests, normal, hskupin, ASSIGNED , Change update tests to use the restart button in the about and software update window to restart Firefox
08:27Tomcat|sheriffdutyfound a changeset whimboo
08:27Tomcat|sheriffdutypushign now
08:28Tomcat|sheriffdutythe last possible changeset before a set of bustages
08:33whimbooTomcat|sheriffduty: oh wow! were I lucky! thanks a lot!
08:57Tomcat|sheriffdutywhimboo: np :) all bugs should be marked now too
09:37pmooreemorley: jonasfj: do you see any reason why we shouldn't use gzip content encoding for *all* artifacts (not just text/plain ones)?
09:39pmoore.... we could have a sanity check that if the compressed version is larger than the uncompressed version, we don't - but maybe that never even happens
09:41pmoore.... i also can't imagine that compression / decompression costs a lot in wall-time - and if it does, probably it is because of enormous artifacts which would then be served more quickly as a result ...
09:41emorleypmoore: for a point of reference, whitenoise takes the &quot;blacklist rather than whitelist&quot; and &quot;check compression ratio <95%&quot; approach
09:42emorleyand logs cases where files were skipped due to compression ratio not being good enough
09:42emorley(where those logs then allow for figuring out which cases to add to the blacklist)
09:42pmooreah, nice
09:42pmoorethat seems like a sane approach to me
10:03pmooreemorley: in the absence of any objections, i&#39;ll implement like this in generic-worker - the same blacklist, and the same <95% compression ratio requirement
10:12bugbotBug 1348097: Telemetry, normal, nobody, NEW , Intermittent browser/components/search/test/browser_yahoo_behavior.js | A promise chain failed to handle a rejection: - at resource://app/modules/experiments/Experiments.jsm:840 - Error: experiments are disabled
10:12Standard8that was interesting
10:13Standard8Ill get a patch up in a bit
12:16jmaherStandard8: yeah, I am wondering if we can take a lesson learned here about other services/prefs/etc.
12:16jmaherStandard8: thanks for digging into that :)
12:17Standard8jmaher: certainly an interesting one
12:58maja_zfhm, this looks useful, e.g.
14:20maja_zfdylan: what kind of 2fa mechanisms does bmo support?
14:20dylanmaja_zf: Duo if you already have a duo account through mozilla, or TOTP using Google Authenticator, Authy, etc
14:21maja_zfdylan: do you know if I can make a yubikey work with totp on bmo?
14:21dylanfor a yubikey you need duo
14:22dylanOR use the yubikey authenticator, which stores the secrets on the yubikey
14:22dylanduo is probably the nicest experience provided you use glob&#39;s firefox extensions
14:23globdylan: fwiw with auth0 that addon shouldn&#39;t be needed anymore
14:24dylanglob: I still hit it on a few things
14:38wcosta|brbjmaher: is a showstopper for migration? The task succeeds but it fails to log cpu time
14:38bugbotBug 1357344: Generic-Worker, normal, wcosta, ASSIGNED , Some tests fail to collect resource info in macosx
14:47jmaherwcosta|brb: is this a problem with BBB? I assume we are not running on tcworker, right?
15:03whimbooato: will you join?
15:15wcosta|brbjmaher: nop, running tests on generic-worker
15:18jmaherwcosta|brb: so we are running non BBB osx normally now?
15:18jmaherwcosta|brb: oh, you are testing it
15:31wcostajmaher: yep, testing it
15:32jmaherwcosta: commented in the bug, doesn&#39;t block migration :)
15:32wcostaI just need to know if this will block migration, although I am investigating it soon
15:32wcostajmaher: thanks
16:22hwinesmacleod: are you happy with mozlogging? As in any reason I shouldn&#39;t use it? (lambda function on AWS)
16:25smacleodhwine: The way it&#39;s designed to take the arguments in logging.debug(...) et al, doesn&#39;t gracefully degrade when the mozlogging handler isn&#39;t the logging handler
16:25smacleodI was planning to fix that, which would mean some adjustment to how the argument passing works
16:35hwineah, so a breaking change -- I think I can live with that :) that&#39;s what pip version pinning is for!
16:58jgrahamato: ^
16:58atojgraham: Yes, simonstewart just confirmed to me on privchat that it works fine on Linux.
17:12mrbkapIs there a way to make a beta build (with --enable-update-channel=beta) with on try?
17:17tedmrbkap: ryanvm has done those before
17:18mrbkapted: do you know how? :)
17:19mrbkapted: I think he&#39;s traveling today.
17:19tedi do not, sorry!
17:19jgrahammrbkap: Ask a sheriff
17:19jgrahamAryx: perhaps?
17:20* mrbkap hopes Aryx knows. All of the other sheriffs are afk.
17:20Aryxyes, let me get the links: and
17:20bugbotBug 1280663: General, normal, cbook, NEW , Uplift Simulation configs
17:20bugbotBug 1343779: General, normal, ihsiao, NEW , Uplift Simulation configs of ihsiao
17:21mrbkapAryx: you rock. thanks.
17:21Aryxif you have questions, you might also highlight phi.lor who does uplifts more often than i do
17:23Aryxthere is also but that&#39;s 6 months old, better to try the files from one of those bugs
17:23bugbotBug 1343779: General, normal, ihsiao, NEW , Uplift Simulation configs of ihsiao
17:26jgrahamato: Put that up as a review request. Tested on linux and macOS
18:27Manishearthwhoever added [log...] links to taskcluster is my hero
18:27Manishearththat is all
18:55jgrahamManishearth: Worship accepted
18:55jgrahamWell assuming you mean treeherder
18:56Manishearthyes that
18:56Manishearththey both start with a t
18:56Manishearthhash collision
18:56* jgraham renames them both to Persona
19:11dylantreecluster and taskherder?
19:14klrmnso, (on OSX) i installed rust, cloned geckodriver, ran &#39;cargo build release&#39;, dropped the file in /usr/bin of my trusty VM, ran chmod +x on it, and tried to run my tests. i got &quot;OSError: [Errno 8] Exec format error&quot;, which is similar to what i got when i downloaded the release and forgot to untar it. ideas?
19:41atoklrmn: Are you trying to run a macOS binary on Linux?
19:42klrmni was. now i&#39;ve built it on trusty, and just have to figure out how to get it to do insecure certs
19:43atoYou need to pass the acceptInsecureCerts capability set to true.
19:46klrmnato: which means i need to figure out what kind of data structure that needs to be wrapped in in order to pass that to webdriver.Firefox()
19:46atoIts hard to tell without examining your language binding.
19:48atoGenerally, however, you want to send {&quot;capabilities&quot;: {&quot;alwaysMatch&quot;: {&quot;acceptInsecureCerts&quot;: true}}} if you want it to reject the session creation if its unable to meet the insecure cert requirement.
19:48atoI dont know how good Seleniums W3C WebDriver support is, but geckodriver masters HEAD also supports legacy desiredCapabilities/requiredCapabilities.
19:49atoIn that case you want {&quot;desiredCapabilities&quot;: {&quot;acceptInsecureCerts&quot;: true}}.
19:58AutomatedTesterato: talking of that, where are we with acceptInsecureCerts fix?
19:58atoIts fixed (-:
19:58AutomatedTesterawesome, we ready to release then?
19:58atoThe problem was that we werent reading the Firefox version number correctly on macOS.
20:00atoAutomatedTester: There are two open PRs on webdriver-rust and one on geckodriver that I think it would be good to get in.
20:00klrmnAutomatedTester: just after i build my own, you&#39;re ready to release =]
20:00AutomatedTesterklrmn: its good for you to always be able to build it :D
20:00AutomatedTesterato: I will look at the PRs after this meeting
20:01atoI dont think any of it is controversial or requires extra testing FWIW.
20:01atoBut we are currently in a position to do
20:02klrmnAutomatedTester: how if only i can figure out how to get it to do the insecure certs part (i can&#39;t find docs for webdriver.Firefox(), and the arguments for selenium.webdriver.firefox.webdriver
20:02klrmn don&#39;t appear to be the right answer
20:02atoklrmn: In Python your best bet is to use FirefoxOptions.
20:03atoYou construct that and pass it as an argument to webdriver.Firefox.
20:03atoAs firefox_options=instance
20:04atoAutomatedTester: In other news, who knew DOM resize events didnt fire when you didnt resize the window.
20:05klrmni know that lots of companies test against hosted test installations that have real certs, but i find it strange that more people haven&#39;t been crippled by the lack of insecure certs between FF47 and now. i managed to pin things in a VM while they were still available, but i haven&#39;t been abel to set up a new environment in months.
20:07klrmnato: that page isn&#39;t saying much about how to use it
20:07atoklrmn: To be quite honest, I find it bizarre people are relying on ignoring security certificates.
20:08atoklrmn: What information are you looking for? You have to be more specific.
20:08klrmnato: when you&#39;re spinning up dozens of servers to test in parallel, getting real certificates on them is a pain
20:09klrmnato: my previous code said &quot;wd = webdriver.Firefox()&quot; and worked with FF47. i&#39;m trying to figure out what it needs to say in order to work with FF53
20:09atoI think I fail to comprehend what parallelism has to do with that.
20:09klrmnthe reason we&#39;re not using real certs is parallelism.
20:11klrmnato: it would be helpful if you could tell me where the docs for webdriver.Firefox() are
20:11atoDid you try webdriver.Firefox(capabilities={&quot;desiredCapabilities&quot;: {&quot;acceptInsecureCerts&quot;: True}})?
20:11atoTheyre in
20:11ato(I searched for &quot;Firefox.&quot;)
20:27klrmnato: that finally workedit is one of 4 different ways i&#39;ve seen to do capabilities today.
20:27atomaja_zf: My suspicion about sandbox.window.onload was correct:
20:27atoklrmn: \o/
20:28atoklrmn: Im here if you have questions about capabilities, but the summary of things right now is that we decided to change the way capabilities worked with the forthcoming W3C WebDriver specification that geckodriver implements, precisely because people were having a hard time understanding how they work.
20:29atoklrmn: So to not make geckodriver completely useless with the existing corpus of Selenium clients, we made the decision to support the old-style capabilities to not break peoples existing tests.
20:30klrmn*nod* and i just happened to run into differing defaults along the way.
20:30maja_zfato: on the one hand, yay, on the other, ugh :)
20:31atomaja_zf: Yeah, something is seriously wrong with addEventListener and sandbox. That is worrying.
20:31atomaja_zf: I should probably add a comment before I push this in the morning.
20:32atomaja_zf: I feel like Im retracing my own steps three years ago.
20:34maja_zfor a follow-up bug?
20:36atoYes, thanks for reminding me.
20:36maja_zfvaguely reminds me of this:
20:37atomaja_zf: Hah!
20:38atomaja_zf: What is worse than that, is when you google something and the only thing you find is a bug on your own code.
20:38atoThat has happened so many times with geckodriver and Rust in general.
20:38atoWhat you get for a young ecosystem I guess!
20:52AutomatedTesterato: I would prefer jgraham to review
20:53AutomatedTesterI _think_ its fine
20:53atoGood with an extra pair of eyes.
20:53atoId quite like it in 0.16.0 because it gives us better logging, so I think its worth holding off for.
20:56AutomatedTesterato: I will build with that PR, run tests and then see if we can get everything ready to ship
22:06Silne30maja_zf: I have a test suite that I would like to get up and running.
22:06Silne30On task cluster.
22:07Silne30i was told that I could either either add it to a current job or try to get a job created for it.
22:07Silne30Do you have any suggestions?
22:10maja_zfSilne30: this tutorial explains how tasks are scheduled and how you can create a new job -
22:12Silne30maja_zf: Sweet! Is it better to create a new job than to add my tests to current jobs?
22:14maja_zfSilne30: depends what your tests are. how do you run them locally? do you use an existing harness? does it make sense to add them to an existing manifest?
22:15Silne30They are telemetry tests. Right now I just install the test harness and then run them through the marionette harness.
22:15Silne30I am also working on a mach command for this suite.
22:15Silne30I don&#39;t know if it would make sense for my suite to run with any other firefox UI tests.
22:16maja_zfSilne30: &quot;install the test harness&quot;? like a custom harness just for telemetry tests? in that case maybe follow the precedent set by firefox-ui-tests -- you can dig around in m-c to see how taskcluster tasks are set up for fx-ui
22:18Silne30maja_zf: Yes. It&#39;s basically a custom runtest and adds a telemetry-harness command.
22:18Silne30I will do that. Thanks.
22:24Silne30maja_zf: You have a path that I can look in?
22:24Silne30I am on dxr
22:29maja_zfsearch for firefox-ui on treeherder on mozilla-central, pick any job, click on &quot;inspect task&quot; in job details, then look at &quot;task details&quot;. all kinds of useful info there, like &quot;task name&quot; -- you can search substrings of that on dxr to find what you need
20 Apr 2017
No messages
Last message: 4 days and 7 hours ago