mozilla :: #fx-team

10 Aug 2017
00:27Callek_cloud9RyanVM: if you ever get pinged about M(cl) again, they should go green on a retrigger e.g: https://treeherder.mozilla.org/#/jobs?repo=try&revision=ae8cd4ad217ede707b2229bc42ff0890a56ba7c6
00:28* Callek_cloud9 will be landing said patch to *actually* enable on m-c in just a bit
09:46Gijson 57, is it still possible to get "old" about:newtab and not activity stream?
09:47Gijsto be clear, not because I want to, but because it's security-relevant to the code I'm touching because AS is doing better at some stuff...
09:48Gijslooks like there's a pref...
09:49Standard8theres a pref to turn off activity stream, I dont know what that gets you though
09:49Standard8I think its a blank new tab
10:07daleharveylol what magic makes the url bar different colours (to highlight domain), in xul it just looks like <div class=&quot;anonymous-div&quot;>https://example.org/path</div>
10:07Gijsdaleharvey: selections
10:08Gijsspecifically, selection ranges
10:08Gijsdaleharvey: see e.g. https://dxr.mozilla.org/mozilla-central/source/browser/base/content/urlbarBindings.xml#441-450
10:08Gijsand then https://dxr.mozilla.org/mozilla-central/source/browser/base/content/urlbarBindings.xml#468-482 does the domain highlighting vs. the &quot;everything else&quot; highlighting
10:09daleharveyah ok, thats pretty neat
10:10* Gijs forgets where the CSS lives :s
10:10Gijsmaybe you&#39;re looking for https://dxr.mozilla.org/mozilla-central/source/browser/themes/shared/compacttheme.inc.css#62 ?
10:11daleharveythat looks very likely, definitely have enough to go on now though, cheers
10:15Gijshm, well, maybe...
10:15Gijshttps://bugzilla.mozilla.org/page.cgi?id=splinter.html&bug=451833&attachment=529440 looks like the secondary color might be fixed :s
10:27daleharveyyeh its just calculating the primary color + some opacity http://searchfox.org/mozilla-central/source/layout/generic/nsTextFrame.cpp#3952, awkward
10:30johannhany hg command on my Windows 7 VM takes approximately 1 million seconds
10:31johannhI didn&#39;t measure it, but I&#39;m sure it&#39;s close
10:50Gijsjohannh: you installed watchman, right?
10:50Gijsalso, why are you running it in a VM?
10:51johannhGijs: I have a Win 10 machine (in fact, 3 right now) but none of them is running Windows 7, I&#39;m just running that in a VM on my Mac
10:52johannhGijs: no, I don&#39;t have watchman, it&#39;s been causing problems on Windows for me unfortunately :(
11:16nhnt11anyone else seeing the top left corner pixels in the title bar turning white on mac, if you leave Nightly idle for a second?
11:18johannhnot yet
11:18daleharveyI dont think I have seen that, just the super flashy title bars
11:18nhnt11daleharvey: do you have an external screen?
11:18daleharveyAssuming there is a bug open for that, anyone point me to it?
11:18nhnt11and if so, is it a &quot;retina&quot; screen?
11:18daleharveynhnt11: yup
11:19daleharveyerr not sure, its like a 2011 cinema
11:19nhnt11is it 2dppx
11:19daleharveynope
11:19nhnt11either way, check if the flashing isn&#39;t as severe (or nonexistent) when the external screen isn&#39;t connected
11:20* nhnt11 is curious
11:20nhnt11(and also I vaguely recall hearing about flicker issues with external screens)
11:21daleharveySo on my macbooks (retina) screen the flashing is still fairly obvious, but a lot less severe than on my non retina external display
11:21nhnt11hmm
11:22daleharveyit happens wether or not the external is plugged in
11:22nhnt11I haven&#39;t seen any flashing :-/
11:22daleharveyOnce I finish up this bug I will make a video, its very obvious here
11:23daleharveynot even just the url bar, the whole ui
11:29nhnt11:(
11:57Sylvestredao, Gijs, is https://bugzilla.mozilla.org/show_bug.cgi?id=1388584 something you could help with ?
11:57firebotBug 1388584 NEW, nobody@mozilla.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
11:57Sylvestrephilipp found the regression which caused it
11:59daoMarco is in pto? *sigh*
12:01daoSylvestre: I know very little about this low-level places stuff. adw reviewed it and isn&#39;t on pto
12:01Sylvestreok, I ni him already, thanks :)
12:05GijsSylvestre: that report is all over the place :s
12:06GijsSylvestre: but the symptoms sound like bug 1356597
12:06firebothttps://bugzil.la/1356597 WONTFIX, nobody@mozilla.org Bookmarks and History are completely lost after downgrade version from 55.0a1 to 53, and completely
12:06GijsSylvestre: what makes less sense is that this happens on a clean profile
12:23markhGijs: was chatting with philipp in #sync about this, then we moved to #places :)
12:23markhI&#39;ll point him here
12:24Standard8markh: huh, mak never told me about #places
12:24markhso yeah, multiple reports of it happening with a clean profile
12:24markhStandard8: heh - yeah
12:25markhnot sure why it exists tbh
12:25Standard8twas probably created in the depths of time
12:25philipphi
12:27Standard8philipp: so clean profile, no extensions?
12:27philippyap
12:27philippsame with a protable version of firefox, newly downloaded and extracted
12:28markhphilipp seems to have a handy user able to repro and run mozregression etc
12:28philippit&#39;s logged at https://mozilla.creutziger.net/logs/firefox.de/latest.log.html#t2017-08-10T10:53:17 in case anybody speaks german
12:30Standard8from looking at the pastebin, it looks like it just blows up
12:31Standard8well, detecting the components
12:31Standard8philipp: Windows?
12:32Standard8oh yes
12:32Standard8comment 8
12:32markhStandard8: yeah, looks like it&#39;s failing to init @mozilla.org/browser/nav-bookmarks-service;1
12:34markhIt must be something about the environment
12:34philippwe&#39;re trying to capture the error console output right from the start now
12:35Standard8markh:https://bugzilla.mozilla.org/show_bug.cgi?id=1388940 is also interesting
12:35firebotBug 1388940 UNCONFIRMED, nobody@mozilla.org Places Database > Task: checkIntegrity - Unable to check database integrity
12:35Standard8philipp: is this person the one on a windows machine with limited rights?
12:35Standard8or if not, do they have limited rights?
12:36daleharveySo can anyone point me to someone who has done work on the awesomebar layout?
12:36philippStandard8: no they have an admin account
12:38Gijsdaleharvey: adw, mak, dao
12:39daleharveyI am trying to debug layout but cant as it hides when I try to inspect it, the no autohide in browser toolbox is enabled, I tried adding noautohide=&quot;true&quot; to PopupAutoCompleteRichResult
12:40markhStandard8: yes - which might even point at bug 1036440
12:40firebothttps://bugzil.la/1036440 FIXED, nobody@mozilla.org Rewriting PlacesDBUtils using async/await and promiseDBConnection
12:40markhthat seems to be where &quot;The task queue was cleared by an error in another task.&quot; comes from
12:41Standard8though the regression range pointed to in the first bug was https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=b01f44c896b2472b154a48c68cf343ae2289d6bd&tochange=d80c517658f5048491b7bf27da25246a297ab50b
12:41markhyeah
12:42markhbut still maybe intertwined
12:42Standard8possilbe
12:42philipphi Gijs - https://pastebin.mozilla.org/9029331 was the ful error console output from the user right from launching firefox
12:42Gijsphilipp: reading along in #firefox.de but I don&#39;t trust my german to reply - after you try the crash thingy, can you ask if there&#39;s a favicons.sqlite in the profile dir?
12:43philippGijs: should i check in his existing profile or a clean one?
12:43Gijsboth would be interesting
12:43Gijsbut maybe the latter first
12:43Standard8While creating services from category &#39;profile-after-change&#39;, service for entry &#39;Notification Telemetry Service&#39;, contract ID &#39;@mozilla.org/notificationTelemetryService;1&#39; does not implement nsIObserver.
12:43Standard8hmm
12:46Standard8well that appears to be true
12:46Standard8although doesnt seem to affect anything
12:47Gijsyeah, I think that&#39;s a red herring
12:47Standard8Ill file it anyway, since I believe it isnt filed already
12:52Standard8ok, filed https://bugzilla.mozilla.org/show_bug.cgi?id=1389057 on that red herring
12:52firebotBug 1389057 NEW, nobody@mozilla.org Console log on startup: &quot;While creating services from category &#39;profile-after-change&#39;, service for e
12:52Gijsphilipp: they&#39;re using avast, judging by the DLLs loaded. Does the problem go away if they disable it (temporarily) ?
12:53Standard8oooh anti-virus
12:53Gijsyeah :|
12:54Standard8that could certainly account for missing services
12:54Gijsbut various other people on SUMO already said they were using kaspersky, so my hopes aren&#39;t very high
12:54Gijsmaybe.
12:54Gijsthen again, third parties just dumping DLLs in our process is... ugh
12:55Standard8and Im not sure why it would be that bug would be affecting it
12:55philippi&#39;ll try but other affected users were telling us they were using a bunch of different av software
12:55philippso no commonalities there so far
12:56Gijsyeah :(
12:56GijsI&#39;m a bit stumped to be honest
12:59GijsStandard8: markh: so... logically speaking... here&#39;s my current hypothesis
12:59Gijs1) for service to be null, initialization must have failed
13:00markhof @mozilla.org/browser/nav-history-service;1 the most recent pastebin implies, but yeah :)
13:00Gijs2) looking at the history service, constructor is basically empty, but Init() isn&#39;t and returns an nsresult. The only way that can be non-NS_OK is if NS_ENSURE_STATE(mDB) fails
13:00Gijshttps://dxr.mozilla.org/mozilla-release/source/toolkit/components/places/nsNavHistory.cpp#315
13:01Gijspresumably that&#39;s only going to happen if the GetDatabase() call fails
13:01Gijsthat calls GetSingleton() which is a macro (ugh, took me a while to figure out) here: https://dxr.mozilla.org/mozilla-central/source/toolkit/components/places/nsPlacesMacros.h#51
13:01Gijswhich will call Database::Init()
13:02Gijshttps://dxr.mozilla.org/mozilla-release/source/toolkit/components/places/Database.cpp#512
13:02Gijsnow there are more options...
13:03markhindeed :)
13:03markh(but so far ;)
13:04Standard8oh
13:04Standard8so https://hg.mozilla.org/releases/mozilla-release/rev/604166699051
13:05Standard8comes out with https://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2017-08-03&keys=__none__!__none__!__none__&max_channel_version=release%252F55&measure=PLACES_DATABASE_CORRUPTION_HANDLING_STAGE&min_channel_version=null&processType=*&product=Firefox&sanitize=0&sort_keys=submissions&start_date=2017-08-03&table=0&trim=1&use_submission_date=0
13:05Standard8if I read all of that right
13:06Standard8then theres been 47k instances reported
13:06Standard8where it gets to stage_removing and cant reopen it
13:06Gijsseems odd for there to be a non-fixable corrupt database on startup?
13:06GijsI think I missed a few steps here :)
13:07Gijs*on startup on a clean profile
13:07GijsStandard8: but if that&#39;s submitting it should show up in about:telemetry
13:08* Gijs pokes at the regression window
13:08Standard8as a result I think BackupAndReplaceDatabaseFile could return a failure which would then get passed back up to Database::Init ?
13:11GijsStandard8: sorry, I&#39;m not sure how you&#39;re getting to the backupandreplacedatabasefile thing
13:11Standard8Gijs: https://dxr.mozilla.org/mozilla-release/source/toolkit/components/places/Database.cpp#539
13:11Standard8Gijs: though admittedly that shouldnt happen on fresh profile
13:13Gijshm
13:13Standard8especially as NS_ERROR_FILE_CORRUPTED would opnly be set if places.database.replaceOnStartup was true
13:13Gijsbut the locked thing would also fire an event
13:13Standard8s/set/returned/
13:13Gijsfor which we listen in nsBrowserGlue
13:14Gijsand show a user-visible notification
13:14Gijsah, but we call the backup-and-restore thing a few more times later on
13:14Gijs*replace
13:15Gijsalso, if the file exists on a clean profile, we must have gotten at least to EnsureFaviconsDatabaseFile
13:15Standard8philipp: did we have a response about turning off avast yet?
13:16Gijsyeah, didn&#39;t help
13:16philippStandard8: yep, it made no difference unfortunately
13:16Standard8ok, so its unlikely to be AV blocking files then
13:16Standard8disk space?
13:17Gijsseems unlikely that many people suddenly have disk space issues
13:17Gijsplus it happened both on upgrades and for new profiles
13:17Standard8true
13:17markhtelemetry tells us that one of the &quot;most popular&quot; sync failures is disk space :/
13:17Gijsso this telemetry thing
13:17markhalong with &quot;corrupt database&quot; :\
13:17Gijsmarkh: yeah, but this user also used mozregression which easily downloads several 100s mb
13:18Gijsin theory could be a different disk
13:18Gijsor quotas
13:18Gijsbut seems unlikely
13:18markhyeah, soory, cynical tangent ;)
13:18Gijs:D
13:18* markh hits bed - good luck!
13:31GijsStandard8: so is it worth checking this stuff in about:telemetry, or something?
13:32Standard8Gijs: it might be worth asking if the user is seeing it there
13:32Standard8I guess it might point to a potential location for the issue
13:32Standard8this is relatively new stuff, and today is the first time Ive seen it
13:33Gijsmm
13:34Gijsphilipp: ^^ could you ask about this? In about:telemetry, under &quot;Histogramme&quot;, filter for &quot;CORRUPT&quot;, see if anything with &quot;PLACES&quot; in it comes up.
13:34philippwill do - meanwhile they&#39;v etested in a different (limited) win account and the issue is not present there
13:36Standard8oh that is strange
13:39philippGijs: no CORRUPT places entries there
13:40GijsWait, so it doesn&#39;t happen on a new user account? :|
13:40Gijsthat&#39;s... what.
13:47Gijsphilipp: can you explain to the user that effectively, the bookmarks/history db portion of Firefox fails to initialize in all the profiles where it&#39;s broken, and it works in the new user account, and, uh, it&#39;s not really clear why it would work in one user account but not another? :)
13:47Standard8I gotta go out for a bit and pick up my lawn mower, bbiab
13:47Gijsmaybe they have some ideas about something in their regular account that would interfere or something?
13:47* Gijs is kind of stumped
14:29philippGijs: he also gave me his mail address in case we want to reach out to him later on
14:29Gijsthat might be helpful, yeah
14:29GijsI posted an update to the bug
14:29Gijshttps://bugzilla.mozilla.org/show_bug.cgi?id=1388584#c18
14:29firebotBug 1388584 NEW, nobody@mozilla.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
14:29Gijsmaybe adw will have more ideas when they get here
14:30* Gijs has run out of most of them :(
14:30Gijsin part because the sqlite initialization is a bit greek to me
14:30SylvestreGijs, interesting update, do you see that as a blocking issue ?
14:30SylvestreI am working on the go live of 55.0.1 right now
14:30Gijsor rather, I don&#39;t see why it would work in one user account and not another
14:30Gijsnone of it makes a lot of sense to me right now
14:30GijsSylvestre: I honestly don&#39;t know.
14:31Sylvestreok, do you know around what time adw comes?
14:33GijsI&#39;m not sure, he&#39;s west coast US, so in the next hour or 2 I expect?
14:33Sylvestredolske, wdyt about https://bugzilla.mozilla.org/show_bug.cgi?id=1388584 ?
14:34SylvestreI can enable updates only for 55.0 users
15:32mconleyGijs: for bug 1388584, do we know if the user&#39;s are entering the &quot;database is corrupt and needs backup and replacement&quot; codepath?
15:32firebothttps://bugzil.la/1388584 NEW, nobody@mozilla.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
15:37Gijsmconley: we don&#39;t know, but the telemetry absence suggests not, I guess?
15:37Gijsmconley: equally, a bunch of the other failure modes there throw up UI in the form of the places locked notification
15:37Gijs(at least, that&#39;s what nsBrowserGlue.js looked like to me, but maybe I missed something)
15:37mconleyit&#39;s pretty suspicious that across all channels, we never reported that (opt-out) Telemetry
15:38Gijswe have?
15:38Gijshttps://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2017-08-03&keys=__none__!__none__!__none__&max_channel_version=beta%252F55&measure=PLACES_DATABASE_CORRUPTION_HANDLING_STAGE&min_channel_version=null&processType=*&product=Firefox&sanitize=0&sort_keys=submissions&start_date=2017-07-13&table=0&trim=1&use_submission_date=0
15:38Gijslooks non-empty to me
15:38mconleyhuh
15:38Gijsis that the wrong histogram or something?
15:38mconleywtf - sec
15:38* Gijs pre-emptively blames the histogram viewer for being wonky
15:39mconleyah, was looking at Release 55
15:39mconleystill interesting that there&#39;s nothing for Release 55 though.
15:39Gijshm, I could have sworn that there were hits there too?
15:39Standard8mconley: maybe wait a long time, or turn sanitzation off
15:39Gijshttps://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2017-08-03&keys=__none__!__none__!__none__&max_channel_version=release%252F55&measure=PLACES_DATABASE_CORRUPTION_HANDLING_STAGE&min_channel_version=null&processType=*&product=Firefox&sanitize=0&sort_keys=submissions&start_date=2017-08-03&table=0&trim=1&use_submission_date=0
15:40Gijsmconley: again, I suspect the histogram viewer is playing tricks on one of us
15:40mconleygrrrr, sanitize was doing it
15:40mconleythanks
15:41mconleyhttps://mzl.la/2vSFjyo
15:42mconleyMight be a red-herring, but those are 100% at stage_closing
15:42mconleywhich means failing in this block: https://hg.mozilla.org/releases/mozilla-release/annotate/tip/toolkit/components/places/Database.cpp#l807
15:45Gijscan we end up there before mMainConn has been opened ?
15:47mconleylet&#39;s see
15:52mconleyGijs: pretty lost in these twisty storage passages. Have we roped smaug into this yet?
15:52GijsI don&#39;t think so
15:53Gijsdoes smaug know about storage?
15:53Gijsmak is away, and I think adw reviewed a bunch of this stuff
15:53Gijss/storage/places/, I gues
15:53Gijs*guess
15:53mconleyI think so, yes
15:53mconleyfailing that, asuth also knows storage
15:56Gijsmconley: tbh, based on the mozregression range in https://bugzilla.mozilla.org/show_bug.cgi?id=1388584#c13 it feels like this must be related to the favicon db
15:57firebotBug 1388584 NEW, nobody@mozilla.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
15:57Standard8mconley: sample value is 1
15:57Standard8so doesnt that mean they got to stage_removing ?
15:58mconleyStandard8: oh wow, yeah, you&#39;re right. Histogram viewer burns me again - where&#39;s the 0 column!
15:58Gijstrim is on
15:58* mconley flips table
15:58GijsI wonder if it&#39;s possible there&#39;s some kind of clownshoes &quot;open this file, try to do stuff, think it&#39;s broken, try to close it, remove it, reopen it&quot; cycle here
15:59Gijswhere the remove step fails because we have the file in use
15:59Gijsor something
15:59Gijsbut tbh I don&#39;t see how that&#39;d actually happen in this code :s
15:59mconleySo if we&#39;ve reached stage_removing, I guess we&#39;re failing to remove the file, and it&#39;s not because the file does not exist
16:00Gijson the other side, it&#39;s odd that people manage to close a db but then not remove it
16:00mconleyperhaps something is holding a lock on it?
16:00Gijsmm
16:18mconleyGijs: last stab at this until someone more qualified can take a look - but it looks like the favicon migration is async
16:19mconleyI wonder if we:
16:19mconley1) Connect to the database, and kick off the migration in InitSchema
16:20mconley2) InitSchema, however, fails later on (while the migration is still ongoing?) and so Database::EnsureConnection considers it corrupted
16:21mconley3) Meanwhile, migration is still occurring off main thread, but main thread decides that because the database is corrupted, it attempts to back up the corrupted database, blow the original away, and start fresh
16:21mconleySo we end up attempting to remove a file that&#39;s still being accessed off of the main thread to do the migration? *sigh*, we need someone who knows storage.
16:22Gijsmconley: I guess it&#39;s possible, though I don&#39;t see why that would happen on a clean profile
16:24philippconstant disk usage when trying to load pages is also one of the described symptoms: https://support.mozilla.org/en-US/questions/1170881
16:25_6a68Could someone with knowledge of the JS interpreter code take a look at bug 1387598? I got needinfo&#39;d because of a Screenshots change, but I think the bug might lie deeper
16:25firebothttps://bugzil.la/1387598 NEW, nobody@mozilla.org Intermittent sessionrestore_no_auto_restore | application crashed [@ GetNameOperation]
16:25_6a68I&#39;m not quite sure how to proceed
16:34Gijs_6a68: try asking in #jsapi
16:34_6a68Gijs: thank you
16:34Gijs_6a68: if that fails, check phonebook for JS folks and needinfo some :)
16:34* Gijs agrees that this shouldn&#39;t be crashing no matter what screenshots is doing
16:36_6a68heh, thanks
16:38Gijsmconley: Standard8: has one of you pinged smaug? If not, I guess I can add to the stack of needinfos in the bug...
16:38Standard8nope
16:48mconleyGijs: I have not - let&#39;s do it
16:57mconleyGijs: are you doing it? If I don&#39;t hear back in 5 mins, I&#39;ll do it.
17:04mconleyOkay, doing it.
17:07RyanVM_6a68: interesting, I had to backout bug 1386333 from Beta today for blowing up ts_paint on Windows (I&#39;ll comment in the bug with links after I finish eating lunch)
17:07firebothttps://bugzil.la/1386333 FIXED, jhirsch@mozilla.com Remove Screenshots rollout pref
17:08MossopGijs: Do you know about PanelMultiView
17:09GijsMossop: yes
17:09mconleyfor the folks following along with bug 1388584, I&#39;ve opened up comms with asuth in #developers
17:09Gijsmconley: sorry, got recruited into cooking
17:09firebothttps://bugzil.la/1388584 NEW, nobody@mozilla.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
17:10MossopGijs: So I&#39;m doing this weird thing where shortly after startup I iterate over all elements in the window. This is causing lots of tests to fail with &quot;TypeError: this._identityPopupMultiView._mainView is undefined at get _identityPopupContentHosts@chrome://browser/content/browser.js:7134:7&quot;. I&#39;m trying to figure out why
17:11GijsMossop: iterate and do what?
17:11MossopGijs: I can trigger the failure by doing nothing but iterating over the elements
17:11GijsMossop: so the high-level why is probably &quot;getting a reference to this node will instantiate the XBL binding&quot;
17:11MossopYeah that was my suspicion also
17:11Gijscombined with when you&#39;re doing the iterating
17:12GijsI expect... something... is unhappy about when it&#39;s being instantiated and what is/isn&#39;t available at that point
17:12Gijsfor the identity popup, we&#39;ve had an issue here before...
17:12* Gijs tries to remember useful keywords for a search :s
17:12MossopI&#39;ve tried delaying the iterating (and tests) by 20 seconds and still not got good results
17:13GijsMossop: I bet the identity popup wouldn&#39;t normally be instantiated at all unless you open it though
17:14* Gijs gulps and opens blame/annotate on browser.js
17:14GijsMossop: bug 1281493
17:14firebothttps://bugzil.la/1281493 FIXED, jkt@mozilla.com Identity popup sometimes doesn&#39;t display EV/DV site identity above &quot;Secure connection&quot; label
17:15Gijsespecially my last sentence in https://bugzilla.mozilla.org/show_bug.cgi?id=1281493#c2 seems to match what you&#39;re seeing
17:15Gijshopefully that can help you figure out what&#39;s going on?
17:16MossopGijs: Wonderful, I was hoping this would ring a bell. Thanks
17:21jktMossop: feel free to cc me on any bug, was one of my first patches in central so sorry about any breakage there.
17:28jktMossop: I guess it could also have become more lazy recently too? Is there a bug?
17:29Mossopjkt: I don&#39;t really have a bug for this failure specifically at the moment. I&#39;m working on bug 1376546 though
17:29firebothttps://bugzil.la/1376546 NEW, dtownsend@mozilla.com Add instrumentation to browser-chrome tests to measure the usage of XUL, XBL, HTML and SVG
17:30jktAh I see interesting, is there other similar bugs you are seeing with similar symptoms if you skip over those elements?
17:30Mardakmconley: is document.QueryInterface(Ci.nsIDOMDocument).hidden the same as document.visibilityState === &quot;hidden&quot; ?
17:31mconleyMardak: hah, it might be!
17:31mconleythere might be some nuance there
17:31mconleylooking
17:32mconleyMardak: looks like you&#39;re right:
17:32mconleyhttp://searchfox.org/mozilla-central/source/dom/base/nsDocument.cpp#12354-12358
17:33mconleyhttp://searchfox.org/mozilla-central/source/dom/base/nsIDocument.h#2774-2781 <-- rather, hidden is true if the state is anything other than &quot;visible&quot;.
17:33mconleybut that&#39;s probably sufficient for this test.
17:33Mardakok thanks
17:33Mossopjkt: Not sure how to do that. Until recently this was passing all tests but now I&#39;ve updated to post-photon work it is failing a lot of tests
17:37_6a68RyanVM: that&#39;s very interesting! I wonder if it&#39;s the same root cause. we&#39;ve seen some really weird process crashes with Screenshots since OOP webextensions landed
17:37RyanVMi&#39;m adding the bug comment now
17:37RyanVMkinda scary that the stacks are a bit all over the place
17:37RyanVMtook me forever to track down because it looked so innocent amongst everything in that push :(
17:38RyanVMhad to do a bunch of Try backouts to finally track it down
17:39RyanVM_6a68: you definitely seem to be hitting some kind of ugly Gecko bug under the surface :(
17:39sfosterwith a new profile I have to add the library button to the toolbar. Is that how its going to be?
17:40_6a68RyanVM: sure has been fun flushing out all these bugs in Firefox via Screenshots ;-P
17:40RyanVMheh, lucky you!
17:44sfosterdao: sorry about that bad review push, its now corrected. Did you want to double-check before I land it?
17:44sfosterhttps://reviewboard.mozilla.org/r/166822/diff/4/#3
17:44_6a68;-)
17:45daosfoster: looks good
17:47_6a68RyanVM: now that I think about it, this was an awfully weird bug, too https://bugzilla.mozilla.org/show_bug.cgi?id=1380267
17:47firebotBug 1380267 FIXED, kmaglione+bmo@mozilla.com Crash in AsyncShutdownTimeout | profile-change-teardown | Extension Shutdown: jid0-GXjLLfbCoAx0LcltE
17:48RyanVM_6a68: well, you&#39;ve got a hell of a reproducible testcase now at least
17:48_6a68lol
17:48_6a68RyanVM: were you seeing crashes in regular builds, or just the pgo builds?
17:51Gijssfoster: no, I landed a change today to move it and replace the bookmarks menu button
17:52Gijssfoster: but that&#39;s on autoland now, I guess it&#39;ll be in tomorrow&#39;s nightly
17:53sfosterGijs: sweet.
17:53Gijsit&#39;ll also make that change for existing profiles, fwiw
17:55RyanVM_6a68: we only run talos on the PGO builds on Beta
17:55_6a68oh, ok
18:06Mossopjkt: Is there some set of steps I could take at startup that would ensure the identity popup is initialised correctly?
18:13Mardakmconley: did you want the info(visibilityState) removed? i figured it might be useful if something goes wrong in the future and someone is looking at logs
18:13mconleyMardak: sure, let&#39;s leave it in.
18:16jktMossop: from what I can see it&#39;s always initialised with the MainViewId which I was depending on
18:18jktMossop: I notice this changed from xml to jsm: http://searchfox.org/mozilla-central/source/browser/components/customizableui/PanelMultiView.jsm#170
18:18MossopLooks like if I open the identity panel before doing the iteration it starts working
18:20jktHmm maybe we should check if _mainView exists then, Gijs?
20:33Sylvestreadw, could you please help with bug 1388584 ? this is blocking the 55 release currently
20:33firebothttps://bugzil.la/1388584 ASSIGNED, bugmail@asutherland.org New Version 55, constantly freezes, not responding, bookmarks lost, all bookmark commands do not fun
20:34adwlooking
20:41philippasuth is asking for a logfile there - the user able to reproduce it has left for today & told me he would come back tomorrow
21:04jawsmconley: so... is that memory leak something that we can write an eslint rule for?
21:05mconleyjaws: might be - gonna add it to the TIL for the next meeting
21:05mconleyand we can maybe talk it out
21:05jawssounds good
21:06jawsmconley: i bet florian and standard8 will have ideas (i don&#39;t see them online right now)
21:06Mossopjkt: I solved my problem by opening about:home and the identity panel before running tests. We&#39;ll see if anything else trips up now
22:10jktMossop: ah cool, let me know. I wonder if other things will come out of the woodwork.
22:12Mossopjkt: Only a couple of remaining failures that look to be of a different nature. Not bad enough to care about really
23:10jawsam i the only one who is always confused when using the Computed view in devtools, and not knowing if the size reported is WxH or HxW ?
23:10jawsi mean, WxH is what i would expect
23:10jawsbut i wish it was explicit
23:10jawsit seems odd that we require users to know the convention here
23:12jawsi&#39;ll file a bug about this if someone +1s it :)
23:12jawsat least the tooltip could explain it
23:13MossopI can&#39;t see myself being confused by that
23:13sfosternever had this confusion, but doesnt sound like a terrible idea to make it explicit
23:16jawsfair enough
11 Aug 2017
No messages
   
Last message: 11 days and 6 hours ago