mozilla :: #ateam

10 Jan 2017
10:33AutomatedTestervkats: thanks for all the webdriver PRs
10:33AutomatedTestervkats: could I have your full name so I can add it to the end of the webdriver spec?
12:49vkatsAutomatedTester: your welcome, it's "Vangelis Katsikaros"
12:51vkatsAutomatedTester: I started from the easier ones, and will continue with some more uncertain (for me). Hope, I don't create too much noise with all these small changes
13:02AutomatedTestervkats: I dont see it as noise
13:02AutomatedTestervkats: there is still lots to do so go ahead and do as many as you want
13:09vkatsAutomatedTester: thanks!
14:15Tomcat|sheriffdutywlach|afk: also i'm here :)
14:27ahalekyle: in your end-to-end times dashboard, what are the white gaps between the grey wait time and the start of the tests?
14:33ekyleahal: that is the side effect of sum-of-medians - the full length is the full median-end-to-end time, while each color bar is the median of the individual times; sometimes the sum-of-medians is less than the overall median, making gaps.
14:33ekyleahal: if you have suggestions, please give them.
14:34ahalekyle: no, that's fine.. was just curious
14:34ekyleahal: I agree it s confusing and ugly
14:35ekylebut averages are out of the question because of outlines, and even if we only average the Nth percentile, we will still get unusally large numbers, occasionally.
14:36ekyleoutlines = outliers
14:41wlachTomcat|sheriffduty: ok, I'm gonna kick off the data ingestion changes now
14:41ahalekyle: interesting, thanks
14:42ekyleahal: well, it is try
14:46Tomcat|sheriffdutywlach: ok closing all trees
14:48Tomcat|sheriffdutywlach: all closed
14:56wlachTomcat|sheriffduty: thanks!
15:07whimbooato: thanks for the quick review
15:09atowhimboo: No thank you, its an excellent patch.
15:09ato(-;
15:09whimbooato: thanks
15:10whimbooit looks like we have a couple of files with the logger included
15:10whimbooand no usage
15:10whimbooi left it in for a quicker way to add debug messages
15:10whimboobut i can indeed remove it
15:10atoI find myself using dump() more and more these days.
15:10whimbooi had a hard time to get this working.
15:10atoLets remove it (them, eventually) so it loads faster.
15:10whimboodoes it work?
15:10atoI didnt build or run it.
15:11whimbooi mean dump()
15:11whimbooi had issues with it especially in listener.js
15:12atoAt some point that got fixed.
15:12atoI dont know what was the cause or what fixed it.
15:13atoI like problems that go away on their own.
15:13atoExcept when the problems are mine. (-:
15:14whimboointeresting. i will try it the next time
15:45armenzgmcote: AutomatedTester jmaher|afk FYI there's a reserve from the people's team to get a temporary person while an employee is out on parental leave
15:46armenzgjust contact our HRBP
15:46mcoteyup
15:46RyanVM|mtgorly
15:46armenzgjust passing the info around
15:46RyanVM|mtgthat's....good to know
15:46armenzgyep :)
15:51Ms2gerAnd by the time we find a qualified temp, the parental leave is over?
15:52mcotehopefully we have some advance warning that there will be a new baby ;)
15:52Ms2gerAlso, England won a game of cricket in India today
15:56whimboojmaher|afk, armenzg_mtg: is add new jobs broken on try?
15:56whimboowhen I select the menu entry nothing happens
16:24whimboowlach: ^ maybe you hvae an idea?
16:24whimbooi tried it with https://treeherder.mozilla.org/#/jobs?repo=try&revision=0a67b0b6eef1&filter-tier=1&filter-tier=2&filter-tier=3
16:24wlachwhimboo: looking
16:25whimbooand wanted to add the Android 15 Mn jobs
16:28jmaherwhimboo: I don't see a 'sch' task, but maybe this is TC only and bstack has the new work for taskcluster
16:28whimboojmaher: i cannot even get the jobs to display
16:29whimbooso i cannot click them
16:30jmaheroh
16:33jmaherrunnable jobs is blank or taking too long: https://treeherder.mozilla.org/api/project/try/runnable_jobs/?decision_task_id=NSncVMw5SKqJSvaY8FQwnQ
16:33jmaherwlach: armenzg ^ that is whimboo's problem
16:34whimboojmaher: so as a workaround i should just wiat until it appears if it does?
16:34whimboolemme see
16:34wlachjmaher: not blank for me
16:34armenzglet me check
16:34jmahersounds like a problem with treeherder
16:34jgrahamdylan, dkl: Is /rest/prod_comp_search/ a bmo-specific API?
16:34* jgraham doesn't see it documented
16:34wlachjmaher: there is a problem, but that's not it
16:34wlachwhimboo: there is no workaround right now, so
16:34wlachsorry
16:35dylanjgraham: yes, it's an internal thing we use for the product/component search
16:35jmaherwlach: oh, on reload there is some ata
16:35jmaherbut it is hanging my browser
16:35armenzgjmaher: use &format=json
16:36armenzgit is the CSS that makes this terribly slow
16:36jmaherok, still waiting
16:37armenzgafter clicking on "add new jobs" nothing happens
16:37armenzglet me look at the dev tools to see what's going on
16:37jmaheryes, because we are stuck waiting for runnable jobs
16:37jmaheryou can see that in the network requests in devtools, that is where I got the link ^^^^
16:37whimbooi assume I should file abug on treeherder?
16:38jgrahamdylan: OK, thanks. Seems like treeherder is using it :/
16:38armenzgyes, please
16:38jgraham(which makes it hard to test changes)
16:38armenzgI will look into it
16:38whimbooarmenzg: which component is it
16:38whimbooand thanks
16:38armenzgtree management - treeherder
16:38armenzgI believe
16:39whimbooso the general compoentn
16:39whimboowill do
16:39wlachjmaher: again, that's not the problem
16:39wlachjmaher: the data is coming through fine
16:39wlachI've figured it out, it's some kind of frontend issue
16:39wlachfor some reason we are always marking the runnable jobs as "should not be visible"
16:40armenzgoh phew
16:40armenzgI thought it might have been some of my changes
16:41whimboojmaher, armenzg_mtg: bug 1330015
16:41bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1330015 Treeherder, normal, nobody, NEW , Add new jobs broken (does not show runnable jobs)
16:41jmaherwlach: while I believe you, when I debug and my browser hangs and devtools shows a long load for runnablejobs, it is easy to come to that conclusion
16:41armenzgthank you whimboo
16:41armenzgjmaher: runnable jobs shows as 200 for me in the network logging
16:42wlachjmaher: the browser just isn't designed to render that much html/css
16:42jmaheryeah
16:43wlachyou could argue about whether we should bother with that fancy view or do something simpler, but that's a bit of a bikeshed
16:44jmahertrue
16:45jgrahamThe browser isn't designed for what?
16:46wlachrendering 1MB+ of json with rest framework's html and css styling
16:47armenzgI'm going to try locally
16:47jgrahamrest framework?
16:48wlachdjango rest framework
16:48jgrahamAh
16:49jgrahamGiven the number of multimegabyte web pages these days it seems like it shouldn't be a browser issue unless it's hitting an especially bad case
16:49jgraham(with TH in general the architecture is not really designed for speed, but obviously that doesn't apply to drf)
16:50wlachchrome does a slightly better job of this case than firefox last I checked
16:50wlachfeel free to file a platform bug I guess
16:51armenzgyes, I have the same experience
16:55armenzgI can reproduce the issue locally even though it hits my localhost runnable jobs
16:55armenzgthanks Will for looking into it; let me know if you want to test anything
17:04whimbooato: so for the return value of close() I have to update what?
17:04whimboogeckodriver only?
17:05atowhimboo: webdriver-rust first, then geckodriver.
17:06whimbooato: i cannot see anything specific to return values of commands in webdriver-rust
17:07whimbooi assume it forwards whatever comes from Marioentte server
17:08armenzgfirst RFC! (/me thinks)
17:08whimbooato: https://github.com/mozilla/webdriver-rust/blob/master/src/command.rs#L131
17:09armenzgwell, I believe there was one already but this is the first since the announcement
17:24jmaherarmenzg++ on the rfc process :)
17:25jgrahamFWIW this increases my conviction that we should focus on "what" not "how"
17:26jgraham(and it should be called something closer to "intent to implement")
17:26jmaheryeah, I think iti is more what we are doing here
17:28armenzg_lunchthanks
17:28armenzg_lunchthat is what the platform people had been doing in their mailing list for a while
17:31jgrahamRight. Although the platform one is specifically for implementing part of the platform it's a closer fit to what we seem to want, which is a lightweight ping reaching the community so that relevant people are aware of what work is going on
17:35jmahergbrown: looks like the l32 tc tests stuck this time
17:35jmaherjgraham: I think some of the problems are that we have so many different mailing lists and team specific places to post
17:36jmaherif we agree to all post to mozilla.tools, then any intent to implement will work fine
17:36gbrownjmaher: yes. I'm still working on plain reftests, but otherwise, looking good
17:36jmahergbrown: above and beyond :)
17:38jgrahamjmaher: Yes, but one concern is that a middleweight process (which I think is what we have at the moment) won't be followed for long. I'm in favour of making things as lightweight as possible so that people actually bother.
17:39wlacharmenzg_lunch: yay! thanks armen
17:42jmaherjgraham: that is valid; too often we have often done half the coding or fully committed to the work before announcing it; then the annoucement is just a formality vs a chance for people to complain
17:42jgrahamjmaher: Seems like the Stockwell email would benefit from some links
17:45jmaherjgraham: oh?
17:45jmaherjgraham: like to orangefactor?
17:45jmaherI tried to keep it simple and put the more verbose information in my blog post
17:46jgrahamjmaher: To the list of top intermittents in particular, but also to OF, yes
17:47jgrahamI think links in the footer are not harmful to brevity
17:47jmaherjgraham: ok, thanks for the feedback
17:47jgrahamAt the moment if you think "oh I wonder what the top intermittents are" you have to guess "oh maybe it says in the blog post" and then click through a link, vs just clicking a link that clearly will give you that information
17:57jmaherack, I am not able to do a text search on orangefactor (as in ctrl+F)
17:58jmaherseems to be a firefox only thing :(
18:10imadueme667422
18:10imadueme921362
18:11dylanjgraham: what's treeherder's dev env look like? docker? vagrant?
18:11armenzg_lunchwlach: yw :)
18:11imaduemeigore those numbers above sry :)
18:11armenzgimadueme: no worries; I've done that many times
18:11jgrahamdylan: Vagrant
18:13dylanjgraham: cool. I would say use our vagrant setup to get a bmo running, but currently getting a database with a working list of products requires a per-person signoff... as soon as I fix that I'll let you know.
18:13jgrahamdylan: OK, thanks
18:51RyanVMjgraham: ping
18:52RyanVMjgraham: to be clear - https://hg.mozilla.org/mozilla-central/diff/f863d5050b3f/testing/web-platform/tests/referrer-policy/generic/subresource/image.py is the only part of the patch from bug 1257249 that "matters" right?
18:52bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1257249 web-platform-tests, normal, james, RESOLVED FIXED, "TypeError: must be string or pinned buffer, not bytearray" log spam from various referrer-policy tests
19:02jgrahamRyanVM: Ish. I think the referer policy ini changes might be a result of that change, and so related
19:02RyanVMok
19:11RyanVMjgraham: we'll see how this goes - I'm a bit worried about some of the INI changes RE: 12.04 vs. 16.04
19:11jgrahamRyanVM: Yeah :/ In the worst case I can regenerate you an ini file from the bustage
19:12jgrahams/an/some/ +s
19:12RyanVMok, I'll let you know
19:16ahaljmaher: I just r+ed the removal of the last whitelisted leak
19:17jmaherahal: oh great
19:17jmaherahal: so we can remove that entire block of code from runtests.py, eh?
19:17ahalyeah, I'll do that once the whitelist removal lands
19:18ahalit's still blocking on something else landing aiui
19:18jmaherahal: a 3 week cycle to fix all those leaks, big props to the devs for that
19:18jmaherahal: did we fix both of the manifest .ini hacks?
19:18ahaljmaher: I'm not sure
19:18jmaherI think that was gamepad (fixed) and the shield_heartbeat stuff
19:18ahalI think one of them was re-enabled at least, not sure about the other
19:21jmaheryeah, I think the shield stuff is DOA
19:22jmaherKWierso: I had emailed sheriffs@moz with a question about linux32-debug as tier-1 for taskcluster; I know that was 22 hours ago, when would be a realistic time to get a response from one+ on the sheriff team?
19:23KWiersojmaher: hopefully sometime after I realize that I didn't actually remember to star that email so I can see it in my "look at this when you have time" list...
19:24jmaherKWierso: cool :) possibly by tomorrow this time you and others will have responded with thoughts/questions :)
20:30RyanVMjgraham: things are looking good :)
20:31RyanVMjgraham: except OSX is still having the same problems :( https://treeherder.mozilla.org/logviewer.html#?job_id=67735640&repo=mozilla-aurora
20:32RyanVMjgraham: BTW, looking at OrangeFactor, looks like things got better on m-c around 18-NOv
20:36RyanVMnot that anything stands out from around that time...
20:45RyanVMjgraham: the only thing I see that looks possibly-relevant is a wpt upstream update on 21-Nov
20:45RyanVMguess I could try running that through Try
20:53RyanVMyeah, that hit inbound on the 18th, sounds like a good place to start
22:34RyanVMjgraham: looks like the updated wptrunner makes OSX on Aurora much happier
11 Jan 2017
No messages
   
Last message: 6 days and 19 hours ago