mozilla :: #ateam

9 Aug 2017
01:12ahal|afkRyanVM|bbl: late reply, but !jsreftest works
01:13ahal|afk(^, $ and | ) also all work
07:46whimboostone:
07:46whimboohttps://treeherder.mozilla.org/#/jobs?repo=mozilla-central&filter-searchStr=firefox%20ui%20update&filter-tier=1&filter-tier=2&filter-tier=3&bugfiler&fromchange=32d9d1e81cc607320a36391845917f645f7a7f72
07:46whimbooi don't see at least failures in our CI update tests
07:47stonethem are backout in 7/28
07:49stonethe problem only happened on 7/23~27 nightly
07:50whimbooyes, the results above go back to those days
07:51whimboohere a reduced view
07:51whimboohttps://treeherder.mozilla.org/#/jobs?repo=mozilla-central&filter-searchStr=firefox%20ui%20update&filter-tier=1&filter-tier=2&filter-tier=3&bugfiler&fromchange=057d626fc5e1afbf4086eb9400ebb3718a4cabca&tochange=09a412e3cc3b
07:51whimboono specific failures for OS X
07:51whimbooyou may want to run those update tests locally with the old builds
07:51whimbooif you can see it you have a base to ensure chagnes to your patch doesn't cause it again
07:55stoneLet me tried if I can reproduce it with the old builds.
07:55stoneThanks for your help
08:02whimboosure. let me know if you have further questions regarding the tests
09:45stonewhimboo: Can I use ./mach firefox-ui-update to test my local build? Where should I specify the binary path?
09:50whimboostone: yes
09:50whimboojust download a nightly
09:50whimboostone: and as mentioned in our DM use the --binary argument
09:51stoneSo it will run the local binary and update it to the --binary one?
09:52stoneOr it runs the --binary one and perform tests on it?
09:56whimboothe latter
09:56whimbooyou always specify the pre-build
09:56whimbooand it updates to the latest available build
09:56whimbooas returned by AUS
09:57whimbooif you want a custom partial, you can override the update url
12:05valentin1is anyone else having trouble pushing to try?
12:07jgrahamvalentin1: What kind of trouble?
12:16Aryxvalentin1: yes
12:17Aryxjgraham: try doesn't respond
12:17valentin1jgraham: Aryx: same here. pushing works, but they don't show up on threeherder
12:17Aryxguess the same issue like last week, filing bug atm
12:18jgrahamvalentin1: "they don't show up on treeherder" sounds like bug 1387407 again
12:18bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1387407 Mercurial: hg.mozilla.org, normal, nobody, NEW , hg.mozilla.org pushlog for Try isn't showing pushes for recent commits
12:18jgrahamHappened three times in the last week, so I guess there's some bigger unfixed problem here
12:21Aryxfor the Try issue, needinfoed the op in bug 1387407
12:21bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1387407 Mercurial: hg.mozilla.org, normal, nobody, NEW , hg.mozilla.org pushlog for Try isn't showing pushes for recent commits
13:16armenzg_afkKWierso|afk: you're welcome
14:15AutomatedTesterthat face where your network can't speak to BMO
14:25ahalRyanVM: not sure if you saw my reply last night, but to select reftest but not jsreftest, use: !jsreftest 'reftest
14:26RyanVMthx
16:10maja_zfAutomatedTester: questions about firefox-driver extension. It only works with older versions of Firefox because it's unsigned, right? Is the extension still supported by the current version of Selenium?
16:12maja_zfAutomatedTester: also, I'm guessing firefox-driver only knows about legacy actions, correct?
16:23AutomatedTestermaja_zf: its only supported to Fx47 and doesnt know the new style actions
16:24AutomatedTestermaja_zf: and I made the python bindings know which type of action chain to create depending on which firefox driver (legacy or geckodriver) it is speaking to
17:00jgrahamSo, mercurial fans. I have a repository with two commits A - B. And I want to create a merge commit like the one you would get from git checkout A; git merge --no-ff B. Is that a possible thing?
17:01jgrahamSo far the internet suggested something called debugsetparents which sounds terrifying
17:01RyanVMheh, I'm pretty sure that's how we generate our merge commits when we do merge day uplifts
17:01jgrahamAnd is also on a post from 6 years ago
17:02jgrahamRyanVM: Oh goodie
17:27KWierso|afkjgraham: you could also try asking #vcs
17:28jgrahamKWierso|afk: True, although this works for now
17:28jgrahamMaybe one day I will understand why
17:28jgraham:)
18:22whimboojgraham, AutomatedTester: if one of you have the time please check bug 1384969. I strongly support an immediate backout due to the major regression it caused
18:23bugbotBug https://bugzilla.mozilla.org/show_bug.cgi?id=1384969 geckodriver, normal, ato, RESOLVED FIXED, Remove browser.dom.window.dump.enabled from geckodriver recommended prefs
18:25AutomatedTesterwhimboo: yea, ok, ask the sheriffs to back out
18:25jgrahamwhimboo: What output does it stop? I was under the impression that logging should not be affected and dump() in chrome would still work
18:26whimboojgraham: all of our Marionette logging is done via log.jsm and not dump
18:26whimbooso everything we log in Marionette is not coming through
18:26whimboomeans any DEBUG and TRACE messages
18:27whimboowhich leaves tracing output useless
18:27whimboowe should better filter the output
18:27whimbooto only let the Marionette log entries pass-through to geckodriver
18:27whimbooand add an option/capability to add really everything
18:28jgrahamwhimboo: Hmm, OK. I don't entirely understand your explaination
18:28whimboojgraham: eg. 1502303146988 Marionette DEBUG Received DOM event "DOMContentLoaded" for "https://www.google.de/?gfe_rd=cr&ei=qlOLWbq6Gu3M8gfGkYHoAg"
18:28jgrahamWhy are only DEBUG and TRACE affected?
18:28whimboo1502303147216 Marionette TRACE 0 <- [1,4,null,{&quot;cause&quot;:&quot;shutdown&quot;}]
18:29whimbooand INFO too for sure
18:30AutomatedTesterwhimboo: so does log.jsm just use dump under the hood?
18:30whimbooi dislike that we write patches for platforms and test on those
18:30whimbooAutomatedTester: that&#39;s the module every JS code is using for logging
18:31jgrahamAutomatedTester: Looks like the log.jsm code has several output options, but dump() is used for stdout
18:31AutomatedTesterthank you jgraham for answering my question
18:31jgrahamBut ato was under the impression that dump() in chrome code was unaffected; is that wrong?
18:31whimbooi can try but i would believe so, yes
18:31whimboolet me test
18:34whimboojgraham: yeah. dump doesn&#39;t work either
18:35jgrahamOK, if that pref stops dump() from chrome space working it&#39;s a bad idea
18:37whimbooso there is DumpAppender and ConsoleAppender
18:37whimboothe latter is only for the Js console
18:37whimbooso DumpAppender is really what we have to continue to use
18:55whimbooAutomatedTester: hm, not sure if I understand your comment here https://bugzilla.mozilla.org/show_bug.cgi?id=1386977#c16
18:55bugbotBug 1386977: Marionette, normal, hskupin, ASSIGNED , No page load detected for pages manipulating the browser history (pushstate/popstate)
18:56whimboothere is basically no page loaded in this case
18:56whimbooso going back and forward we have the same page
18:58whimboowow fx crrashed when i tried to load the local test page
19:01whimbooi really hate the MOZ_RELEASE_ASSERT(parentBuildID == childBuildID)
19:02whimbooa second fx process replaces the installation underneath and causes my main process to die at certain times
20:32felipedoes anyone know if the all-tests.json file (that was generated by the build) is gone?
20:34ahalfelipe: it&#39;s called all-tests.pkl now
20:35ahalyou can parse it with python&#39;s pickle module
20:35felipeah
20:35felipeokay thanks
21:00erahmDo we have some sort of mach command to run an xpcshell script?
21:28ahalerahm: I don&#39;t think so
21:37erahm|obj-x86_64-pc-linux-gnu-clang-dmd/dist/bin/run-mozilla.sh obj-x86_64-pc-linux-gnu-clang-dmd/dist/bin/xpcshell| is pretty awful
21:40jgrahamerahm: Seems easy enough to add if that&#39;s all it is (I&#39;m not volunteering though)
10 Aug 2017
No messages
   
Last message: 14 days ago