mozilla :: #ateam

14 Sep 2017
09:24whimboojgraham: good morning
09:24whimbooregarding http://searchfox.org/mozilla-central/source/testing/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py#79
09:24whimbooare we ok with 120s of timeout?
09:24whimboothat's what it will be by default now, and not 60s
09:24whimboootherwise we would have to pass in the timeout
09:25whimboooh wait
09:25whimbooyou set the startup_timeout via __init__
09:25whimbooso taht is fine then
13:55bkellywhat is the fd limit on our linux test servers?
13:55jgrahambkelly: You can get a one-click loaner to find out
13:58bkellyjgraham: I guess that would assume they are all configured the same way?
13:58bkellydo we have a config file that applies sysctl settings to linux automation machines, etc?
13:59whimboobkelly: they all use docker
14:00whimboohttps://hg.mozilla.org/integration/autoland/file/tip/taskcluster/docker/desktop1604-test should it be
14:01bkellywhimboo: this suggests that docker guests inherit the fd limit from the host? https://stackoverflow.com/questions/24318543/how-to-set-ulimit-file-descriptor-on-docker-container-the-image-tag-is-phusion
14:02whimboosorry, no idea. i haven't used docker myself yet
14:02bkellyI guess it just sounds like its determined by whatever server the docker image is run on
14:02bkellyso the docker config doesn't describe this
14:05jgrahambkelly: It seems like just checking on a one click loaner for the test type you care about will answer the question? Certainly all the machines for a specific test type should be the same AWS image and the same docker container
14:06bkellyok
15:14whimboojgraham: for the geckodriver release.. do we usually add all version bumps of external crates or only our own ones?
15:14whimbooin the CHANGES.md I only see ours
15:15jgrahamwhimboo: Add where? I think version bumping crates that have other users in m-c is hard for some reason that I forget
15:15jgrahamI don't think we need to document such version bumps at all
15:15whimboofor the release chagnes in the above file
15:16whimbooright
16:11jgrahamahal: I already said that, but maybe only on one list :) But there was a followup asking if we run by dir for mochitest-browser, which I don't know but you might
16:14ahaljgraham: yeah, I think I took too long writing out my reply and didn't check before hitting send :)
16:14ahalI replied to the follow-up
16:15camdekyle: Hey Kyle-- Where can I get more information about Active Data? And/or are you available for a quick chat to answer some questions I have? :)
16:15jgrahamahal: Thanks! :)
16:17jgrahamahal: Incidentally, where is that specified? Is it just hardcoded somewhere? The options seem to be default False in the mochitest_options.py file, and I couldn't see which flags for the -bc tests would cause chunking
16:18ekylecamd: there is https://github.com/klahnakoski/ActiveData/tree/dev/docs, and specifically https://github.com/klahnakoski/ActiveData/blob/dev/docs/GettingStarted.md
16:18ahaljgraham: yeah, I think it's hardcoded in mochitest :/.. we should remove the option for it
16:19ahalI think it's at least hidden by default
16:19camdekyle: thanks! I'll start there.
16:20ekylesure, we can setup a vidyo meeting. I am eating lunch, right now, so in 1/2 hour? Ping me when you are ready, and I can jump into vidyo
16:20jgrahamahal: OK, just checking why my code reading was failing :)
16:21camdekyle: Thanks! Yeah, that'd be great. I'll try to educate myself a bit in the mean time... :)
17:17camdekyle: that query you sent in email is a bit help to the direction I was going to ask about. :) I would love to make a link for that from the Test-based UI to find the chunk by specific result.test. I'll experiment with the query a bit... :)
17:17camds/bit/big/
17:19ekylecamd, I saw you bug and showing how it can be done with curl; and as a consequence a python requests call: https://github.com/klahnakoski/ActiveData/blob/dev/docs/GettingStarted.md#7-further-processing
17:25camdekyle: awesome! I take it that ``build.revision`` would be the correct field for adding push revision to the where clause?
17:25ekylecamd: yes
18:18whimbooAutomatedTester, jgraham: so we cannot release geckodriver 0.19 today
18:19whimboosafebrowsing devs have removed some recommended preferences from our code
18:19whimbooi wonder if that usually should require review from testing peers
18:22whimbooAutomatedTester: I assume you and Andreas can build and release it?
18:22jgrahamwhimboo: Those prefs don't seem to exist any more
18:23jgrahamwhimboo: AFAICT the only question is whether to add browser.safebrowsing.downloads.enabled: false
18:24whimboofrancois_: ^ any reason why that hasn't been done for Marionette/geckodriver?
18:24whimbooat least Marionette has it
18:25whimboooh, also for geckodriver
18:25jgrahamwhimboo: I (hope) it's not that important anyway? I guess the issue is that something could be blocked by safebrowsing that shouldn't be in a test environment
18:25whimbooit's all there
18:25whimbooso we are fine
18:25whimboo"browser.safebrowsing.downloads.enabled", Pref::new(false)
18:26jgrahamOh, maybe I missed some context
18:26jgrahamAnyway it seems like it's fine
18:30RyanVMwho would know about the mochitest harness self-tests? ahal?
18:31ahalRyanVM: yeah
18:32RyanVMhttps://public-artifacts.taskcluster.net/B2d_TxsvSbe09tyV0dwOmQ/0/public/logs/live_backing.log happened on the m-b -> m-r merge
18:32RyanVMnot sure how much to care
18:32RyanVMI've retriggered it for good measure too
18:35ahalRyanVM: those tests are only useful for catching regressions in the mochitest harness, and people hopefully aren't modifying that too much on release.. so I guess you don't need to be overly worried
18:35ahalbut would be good to figure out anyway
18:35RyanVMthat's reassuring :)
18:36ahalmaybe things are configured differently on release
18:49RyanVMahal: retrigger succeeded
18:49RyanVMI officially care less :)
18:49ahalRyanVM: weird, me too :p
18:49ahalyou won't even see those tasks unless mochitest or taskcluster/taskgraph is modified
18:50RyanVMmakes sense
19:35erahmHi folks, I've got a marionette question
19:35whimbooerahm: hi! just ask
19:35erahmI'm updating an old script that uses pypi marionette and am seeing "InvalidArgumentException: Unrecognised timeout: page load"
19:35erahmwhimboo: ^
19:36AutomatedTestererahm: ooo.... we might have a really old version up there
19:36AutomatedTesterI thought you were using the in tree version
19:36erahmhmmm actually I think I pinned to 4.0
19:36erahmAutomatedTester: this is for atsy (not awsy)
19:36erahmlet me unpin and see what happens
19:36whimbooi released the driver on 2017-06-29
19:36whimboo2.3.0
19:37whimbooso that should work with it
19:37whimbooyepp
19:37whimbooBug 1342162 - Make timeout API in client backwards compatible; r=whimboo
19:37bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1342162 Marionette, normal, ato, RESOLVED FIXED, Align page load duration key with spec
19:39erahmwhimboo, AutomatedTester: sorry worked it out, I needed marionette-harness 4.1.0
19:39AutomatedTestererahm: \o/
19:39whimboogood
19:55whimbooAutomatedTester: could you have a look at https://github.com/mozilla/geckodriver/issues/918 ?
19:56whimboonot sure if that is something we should make clear in the spec
20:00whimbooAutomatedTester: or would the be the change simon did via https://github.com/w3c/webdriver/pull/932/files
20:21nalexanderahal: when I run |mach xpcshell-test --log-raw=- ...| from within Node.js, I see only a few messages -- namely "test_start" and "test_end".
20:22nalexanderahal: all the others don't seem to make it through. This happens with Node.js's subprocess and spawn, but not from a regular shell.
20:22nalexanderAny thoughts?
20:30ahalnalexander: node must be intercepting the subprocess stdout?
20:31ahalI'd bet test_start/test_end are logged from the parent process
20:31nalexanderahal: it's supposed to be forwarding everything; and yes, it looks like a parent/child process isue.
20:31nalexanderahal: I just fixed it by setting the cwd of the |mach xpcshell-test|. That's... odd.
20:32nalexanderThat is, with cwd=$topsrcdir, success. Without, failure.
20:32ahalyeah, that doesn't make a ton of sense
20:32nalexanderahal: maybe the subprocess silently fails? It said 1 passed, even with a different cwd.
20:33ahalseems plausible
20:36AutomatedTesterwhimboo: I think you're right.
15 Sep 2017
No messages
   
Last message: 7 days and 13 hours ago