mozilla :: #ateam

13 Jul 2017
08:05whimboojmaher|bbiab, wlach|afk loaners on osx should work
11:17leni1whimboo: ping
12:12whimbooleni1: hi
12:12leni1Hello. Was wondering if you'd be able to look at my needinfo for Bug 1315704
12:12bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1315704 Marionette, normal, lenikmutungi, NEW , [marionette harness tests] Test that BaseMarionetteArguments constructs arguments as expected
12:13leni1Not sure what timezone you're in or how available you are apart from what your handle on Bugzilla says.
12:13leni1Wanted some advice regarding how to handle the arguments for the bug.
12:14leni1Comment is here: https://bugzilla.mozilla.org/show_bug.cgi?id=1315704#c24
12:14bugbotBug 1315704: Marionette, normal, lenikmutungi, NEW , [marionette harness tests] Test that BaseMarionetteArguments constructs arguments as expected
12:14leni1If you can respond here, great. I'll write it up and cancel the needinfo myself.
12:14leni1If you want to respond on Bugzilla, that's fine too. Whatever works for you :)
12:19whimbooleni1: i'm only partly avaialble this week
12:19whimbooand i have to catch up on lots of things
12:19leni1True. I noticed and I understand
12:19whimboothere are still some other things I have to look into first
12:20leni1Cool. Np :-)
12:45whimboostandups: workign half-day only makes it kinda hard to keep up with bugs at all. when can I code myself again? :D
12:45standupsOk, submitted #48164 for https://www.standu.ps/user/whimboo/
13:28marcoekyle: gmierz: do we actually need "shutdown" coverage? I thought we only cared about startup
13:28ekylemarco: not in the short term
13:29ekylemarco: yes, we only care about startup right now;
13:30marcoekyle: gmierz: and do we need to run it periodically, or is it a one-time ask?
13:32ekylemarco: good question. I had assumed we would run it periodically. but I also assumed this is a short term request
13:32ekylemarco: what are you thinking?
13:33ekylemarco: run it locally?
13:33marcoyes, I was thinking of running it locally
13:48ekylethat may be best, because they will want a human-readable coverage report too (humans!)
13:48ekylemarco: ^
13:58gbrownemorley: got a minute for an orangefactor question?
13:58emorleygbrown: hi! sure :-)
13:58gbrowngreat
13:59gbrownI am trying to run woo_commenter.py --test locally for the first time in months
13:59gbrowninstead of getting json from brasstacks, I'm seeing it return an auth0 login page
14:00gbrowndo you know how I can get around that?
14:01jmahergbrown: ack, I have had that problem, I couldn't get the python ssl stuff to work on windows, so I just load the url and save->as the json
14:02gbrownjmaher: hmm. but I want to test woo_commenter, so...inconvenient
14:03jmahergbrown: possibly on linux you can get the pyssl stuff to work
14:03ekylegmierz: so, your job, https://treeherder.mozilla.org/#/jobs?repo=try&revision=d4b1a77e96d9a0ef3266713bc206a35194b35e8a&selectedJob=113893030 can we generate human coverage artifact from it now using the local coverage instructions/
14:03ekyle?
14:04jmaherekyle: human coverage- like applying enough sunscreen?
14:04ekylejmaher: oh joel! :/
14:05marcoekyle: if we do, we won't get "startup" coverage information
14:05marcoekyle: but shutdown too
14:06marcotogether
14:06ekylemarco: yes, I know it includes shutdown, but mconley mentioned it may be fine in the short term because it is close enough.
14:07gmierzekyle: I think so, but I'll need to test that to be sure.
14:07ekylemarco: then you can solve the problem of emitting multiple coverage artifacts per task, this ts_paint suite just being one specific case where it is useful
14:08marcoah, ok, if that's fine then yes, it should be possible
14:08marcogmierz: you can use the firefox-code-coverage csript
14:08jmahersolving this problem (even as a hacky one off) would end up opening the door for many others to get coverage and related data
14:09marcogmierz: but you will need to change it a bit to only use the ts_paint coverage
14:09ekylejmaher: yes, marco solved this problem a while ago, this is the first time a third party will use it (in think)
14:09gmierzmarco: Do you have a link to those scripts? I'm having trouble finding a link
14:09marcogmierz: https://github.com/marco-c/firefox-code-coverage/blob/master/codecoverage.py
14:09gmierzthanks!
14:10ekylegmierz: if you can change the script to accept a TH job number, that would be best
14:10marcogmierz: basically this line: https://github.com/marco-c/firefox-code-coverage/blob/master/codecoverage.py#L98
14:10marcoyou can use any property of the task
14:10marcoI was using the name
14:11marcommmh, actually you won't need to modify the script
14:11jmaher:)
14:11marcoyou can pass "talos-scov" as the --suite parameter
14:12marcogmierz: ^
14:12gmierzmarco: Sounds good, thanks!
14:12marcolet me know if it works
14:13gmierzmarco: does the script work on windows?
14:13marcogmierz: I don't think so
14:13gmierzmarco: ok
14:14marcojmaher: how can I run ts_paint locally?
14:15marco./mach talos-test I guess
14:15marco:)
14:15jmaher./mach talos-test -a ts_paint
14:16gbrownemorley: any better suggestions than pyssl? ^^ would it be easy enough to test on brasstacks1?
14:16emorleygbrown: sorry for the delay, tabbed away waiting for the next nick highlight :-)
14:16gbrownemorley: I figured. I do that all the time.
14:17marcojmaher: thanks!
14:18emorleygbrown: so the options I can think of are (a) test on brasstacks (if accessed via localhost it skips auth), (b) modify the various scripts to be able to handle passing a cookie taken from your local profile, (c) use an SSH tunnel to access the ES instance locally and run the OrangeFactor API locally too
14:19emorleygbrown: if it were me, I'd try (c) - in my profile I use: `alias orangefactor-es-tunnel='ssh brasstacks -L 9200:of-elasticsearch-zlb.webapp.scl3.mozilla.com:9200 -N'`
14:20emorleygbrown: and then modify the local API config to point at localhost not the remote ES
14:20emorleygbrown: ah actually it's in the readme: https://hg.mozilla.org/automation/orangefactor/file/tip/README.txt#l8
14:21gbrownemorley: thanks - I'll try that! one other thing...
14:21gbrownI can ssh to brasstacks1, but I've never done a deployment or anything there
14:22gbrownemorley: I notice you usually "sudo su webtools -"
14:22gbrownwhen I try that sudo prompts "password for gbrown"...which I cannot seem to determine
14:23gbrownis it asking for my ldap pwd?
14:23emorleygbrown: yeah there are a few pieces: the source directory is owned by webtools (hg pulling as root will mess the permissions up) - see webtools history for example pull command (I've used them as one-liners to make re-using history easier)
14:23marcojmaher: is talos going to parse the output of the test after Firefox is closed?
14:23emorleygbrown: ah sounds like you need adding to the sudoers list
14:24gbrownah, yes, that sounds like a good idea
14:24jmahermarco: yes/no
14:24jmahermarco: talos does process stdout of the process it creates (subprocess.popen, etc.), and mozharness does process the output of the talos harness as well
14:25emorleygbrown: it's currently https://emorley.pastebin.mozilla.org/9027003
14:25marcojmaher: okok, thanks
14:25jmahermarco: but we don't necessarily do perf work based on the output, for ccov we don't expect perf data in the output and it will not be uploaded
14:25marcoI need to find a reviewer for bug 1380659
14:25bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1380659 Code Coverage, normal, nobody, NEW , Support dumping/resetting coverage counters from JavaScript
14:26gbrownemorley: will you add me to the venerable list?
14:27emorleygbrown: it's controlled via puppet to which I don't have write access - filing a bug now
14:28gbrownah, thanks
14:34marcogmierz: can you set "checkin-needed" on bug 1378533? Or were you waiting for something?
14:34bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1378533 Code Coverage, normal, nobody, NEW , linked-files-map.json keys should contain the build directory
14:34emorleygbrown: filed bug 1380666 - could I leave it to you to chase someone to do it?
14:34bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1380666 WebOps: Other, normal, server-ops-webops, NEW , Update sudoers list for brasstacks1.dmz.scl3.mozilla.com
14:34gbrownemorley: absolutely. thanks very much!
14:34emorleynp!
14:35gmierzmarco: Oh thanks for reminding me :) will do
14:36ekylegmierz: eventually, I would like to see CoCo let the user choose a revision, and a "test" (aka suite chunk) to see the coverage directly. But that is a long way off
14:37gmierzekyle: Ok, I'll file an issue for that in the main repo
14:37ekylegmierz: oh! yes! we can do that sort of thing.
14:37wlach|afkwhimboo: jmaher: ok, if loaners work on os x, then retrigger actions should be easy to implement if they don't work already
14:38jmahermarco: I am looking at your review, I am unfamiliar with the moz.configure changes- is there a try run to validate?
14:38gmierzekyle: Do you want to file it or do you want me to?
14:38marcojmaher: will do
14:38ekylegmierz: i will file the bug
14:38gmierzekyle: ok, ty
14:38jmahermarco: cool
14:41gmierzekyle: marco: jmaher: The script worked on the talos-scov build without any mods, here's the report I got from it: https://www.dropbox.com/s/0w1slb57uhxrkm7/report_talos_scov.tar.gz?dl=0
14:42ekylegmierz: looking....
14:42marcogmierz: cool, I'm downloading it
14:44ekylegmierz: love it: a gigabyte
14:45marcoekyle: it contains the entire firefox source code I think
14:48gmierzekyle: lol I think marcos right about that, I couldn't think of a better way to send it without breaking it
14:48ekylegmierz: I think this is right, it is just big
14:49ekylewe can update the bug with instruction pointers to generate this, rather than providing a link to what you produced.
14:51marcoekyle: gmierz: it would be good to wait for the fix to the linked-files-map.json
14:51marcoas most of the header files are in the wrong position
14:51gmierzmarco: Ok, we should be able to do that tomorrow
14:52gmierzekyle: I'll write it in there
14:52ekylemarco: the coverage report still seems useful
14:55ekylemarco: maybe I am wrong; the css files are referenced wrong
14:55marcoekyle: yes, definitely, but the header files can't be easily found
14:56marcogmierz: you could actually workaround the problem by removing the linked-files-map.json from the gcno zip
15:04gmierzmarco: Ok, I'm also going to just apply the patch manually and get a run with the good mappings.
15:04jmahermarco: looking at your MOZ_CODE_COVERAGE patch, does "ac_add_options --enable-coverage" set a value for code_coverage=1 ? I don't follow how that gets set
15:05marcogmierz: makes sense, although you will also need an update to grcov to make use of the new linked-files-map.json format
15:05emorleygbrown: I've added a new deployment section to the wiki to make the process clearer (I've also tweaked the workflow slightly to simplify it) - https://wiki.mozilla.org/Auto-tools/Projects/OrangeFactor#Deploying - also above that is an Architecture section I added for the RRA not long ago, if it's of any use
15:05marcojmaher: I think it will only set CONFIG['MOZ_CODE_COVERAGE'] and DEFINES['MOZ_CODE_COVERAGE'] to 1
15:05marcowhich is what we need, right?
15:06gmierzmarco: Ah, have you posted an updated version to tooltool? We could get it updated today.
15:06jmahermarco: yeah, I just don't see how we get a value of 1 from "set_config('MOZ_CODE_COVERAGE', code_coverage)"
15:07jmaherthis could be my lack of experience in moz.build files, I seem to hack on them once every 2-3 years
15:07gbrownemorley: that looks great. very clear.
15:08jmaheremorley++
15:08marcojmaher: I *think* it will be set to 1 by the code_coverage function returning True
15:09marcogmierz: not yet, I will do it shortly
15:09gmierzmarco: ok, thank you
15:09jmahermarco: ok, lets see if it works on the try run; I think it does that as well
15:14marcogmierz: can you send me the try run with the new linked-files-map again?
15:15marcogmierz|linux: ^
15:15marcoI want to verify the grcov's fix
15:16gmierzmarco: https://treeherder.mozilla.org/#/jobs?repo=try&revision=e822e3811fab66f8d23d1e4905f1ff49e65fc630
15:17marcogmierz: ty
15:18marcogmierz: ah, there are no tests, so I can't verify it :P
15:19gmierz|linuxmarco: hah, want me to start a new run for you?
15:21marcogmierz|linux: no, it's fine, I can take another test run and use the linked-files-map from that run
15:22gmierz|linuxmarco: cool
15:55marcogmierz|linux: I think it's working
15:55marcogmierz|linux: I will upload it to tooltool and ping you when it's done
15:56gmierz|linuxmarco: Sounds good
15:58jmahermarco: I was looking for a try push with the build system changes for enable_coverage and didn't see one
16:02marcojmaher: I haven't sent it to try yet, will do shortly :P
16:05jmahergreat
16:16marcojmaher: I added linux64 code coverage on trychooser
16:17jmahermarco: oh, cool
16:22marcojmaher: here's the try build https://treeherder.mozilla.org/#/jobs?repo=try&revision=1c4a6de39727684bd3774469ab4c11f1180d202a, let's see if it works
16:25jmahermarco: thanks!
17:03marcojmaher: can you r? 1380712? I'd like to merge https://bugzilla.mozilla.org/show_bug.cgi?id=1378533 and https://bugzilla.mozilla.org/show_bug.cgi?id=1380712 together
17:03bugbotBug 1378533: Code Coverage, normal, nobody, NEW , linked-files-map.json keys should contain the build directory
17:03bugbotBug 1380712: Code Coverage, normal, mcastelluccio, ASSIGNED , Update grcov after modification of the linked-files-map.json format
17:13jmahermarco: I will look
17:13marcoty
17:13marcojmaher: the try build with the MOZ_CODE_COVERAGE change failed, because it was being set twice
17:14jmaherok
17:14marcoI've removed it from old-configure.in, let's see if it works now
17:14jmahercool
17:15jmahermarco: r+
17:18marcojmaher: thanks!
17:26leni1ahal: I have some bad news about Bug 939755
17:26bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=939755 httpd.js, normal, nobody, NEW , With httpd.js we sometimes don't get the most recent version of the file
17:27ahalleni1: what happened?
17:27leni1I have made all the modifications to the files but as separate commits.
17:27leni1ahal: ^^
17:28ahalleni1: that's ok, are you using hg?
17:28leni1Yes I am. :-)
17:28ahalok with or without mq?
17:28ahal(if you don't know what that is, then you probably don't have it on)
17:29leni1I use bookmarks. So I have one covering the bug. mq is enabled but I never really got the hang of it.
17:29ahalok great
17:29ahalmq isn't recommended anyway, so that's good
17:29ahalleni1: first make sure you have the 'histedit' extension enabled
17:29ahalthen run `hg histedit`
17:29ahalthere's an option in there to fold the commits all into one
17:31leni1How do you enable it? My hgrc file shows `histedit=`
17:32ahalleni1: great, that means it's already enabled
17:34leni1Cool
17:34leni1ahal: Let me do that. What would I be aiming for?
17:35ahalleni1: update to the last commit and run hg histedit and it'll open up an editor
17:35ahalfor all the commits below the top one, change the 'pick' to 'roll' and then save and exit
17:36ahalthat will 'roll' all the bottom commits into the top one
17:38leni1Done. Thanks. :-). So let me go and work on the last bit you mentioned in the bug and we'll be good to go.
17:54ahalleni1: awesome, thanks!
18:02gmierz|linuxekyle: marco: jmaher: The start of the diff viewer can be found here for future reference, tomorrow I'll get an example of it running on my github.io page: https://github.com/gmierz/moz-coco/tree/dev_gmierz_patchdifftool
18:06jmahergmierz|linux: nice!
19:29whimbooato: hey. it got a bit silent those last days
19:29whimbooato: which progress did you make so far for teh window / browser handling?
19:56atowhimboo: Im making progress and the window tracking is working and reliable, but Im not banging my head against the wall on the peculiarities of how switchToWindow and window handles work in Marionette.
19:56atos/not/now/
19:56atohow*
19:56atoIts a bit of a mess at the moment, and certainly not spec conforming.
19:57atoThere seems to be a very blurred line between chrome and content when it comes to window switch, what should get focus when, and so on.
19:58atoSo at the moment Im trying to encode the existing behaviour, but Im having a headache trying to synchronise the outerWindowIDs that it currently uses for frame script registration (in GeckoDriver#browers). I suspect maybe I should rip it all out, but its hard to know where to stop.
19:59whimboooh, yes. I can imagine
19:59whimboothat's nasty code
20:00atoYes, I have a creeping feeling that Ive solved the easy bit with sorting out the window tracking, and now the hard bit of making Marionette use it remains.
20:00whimbooato: david mentioned that you also want to fix the message handling part in that patch? I kinda wondered about because afair we decided to do that later
20:00atoI told him it _enables_ us to no longer broadcast IPC messages, but I dont intend to fix that in the same changeset.
20:01atoIts already monstrously big.
20:01whimboogood! i thought that
20:01whimbooso he misunderstood it then
20:01whimbooato: whenever you have something you need help/feedback with please let me know
20:02whimbooeven a WIP patch uploaded
20:02atoSomewhat surprising to myself, I have the window handle tests working for content and chrome. Now working on window switching focus.
20:02whimboonice
20:03atoBut I suspect we are facing an equally big patch to make us spec conforming after this.
20:03atoThe biggest part there will fixing tests.
20:03atobe*
20:04atoThough maybe a good chance to write some WPT tests from scratch.
20:04atoWill be back in the morning, got to head home now.
20:10whimboosure
20:11whimbooenjoy the rest of the day
14 Jul 2017
No messages
   
Last message: 10 days and 21 hours ago