Closed Bug 1240227 Opened 10 years ago Closed 8 years ago

Intermittent test_clearHistory_shutdown.js | application crashed [@ mozalloc_abort(char const * const)] | after Assertion failure: mPendingEventCount (Mismatched calls to observer methods!), at LazyIdleThread.cpp:550

Categories

(Toolkit :: Places, defect, P3)

Product:
Toolkit
Component:
Places
Type:
defect

Tracking

()

Status:
RESOLVED WORKSFORME
Tracking Flags:
Tracking Status
firefox46 --- affected

People

(Reporter: KWierso, Unassigned)

Details

(Keywords: intermittent-failure)

This is either a bug in IDB or a bug in LazyIdleThread. ni myself to remember to look at it.
Component: Places → XPCOM
Flags: needinfo?(khuey)
Product: Toolkit → Core
This log is actually really interesting. LazyIdleThread is being shutdown by nsThreadManager::Shutdown because it's still around, but it only accepts events from the thread that created it, which is not the main thread.
Flags: needinfo?(khuey)
I assume this is the QuotaManager's mIOThread. Jan, does QuotaManager block xpcom-shutdown-threads to shutdown that thread? It looks like it does profile-before-change, but maybe that doesn't fire in xpcshell?
Flags: needinfo?(jvarga)
(In reply to Kyle Huey [:khuey] (khuey@mozilla.com) from comment #4) > I assume this is the QuotaManager's mIOThread. Jan, does QuotaManager block > xpcom-shutdown-threads to shutdown that thread? It looks like it does > profile-before-change, but maybe that doesn't fire in xpcshell? Yeah, QuotaManager registers as an observer for profile-before-change that should always come before xpcom-shutdown-threads. xpcshell only fires profile-before-change if do_get_profile() was called. See http://mxr.mozilla.org/mozilla-central/source/testing/xpcshell/head.js It seems the test here doesn't call do_get_profile().
Flags: needinfo?(jvarga)
Marco, can we change this test to get a profile?
Component: XPCOM → Places
Flags: needinfo?(mak77)
Product: Core → Toolkit
Hm, I remember a problem with an xpcshell test didn't get profile-before-change on time because it registered a cleanup function via do_register_cleanup(). The main xpcshell head.js uses do_register_cleanup() for firing profile-before-change, so profile-before-change may be fired too late.
what do you mean by too late? functions registered through do_register_cleanup are executed before proceeding with shutdown in a while loop in head.js, so everything should be well alive at that point, we should be well before xpcom-shutdown. Though if the function returns a promise, head.js spins the events loop (http://mxr.mozilla.org/mozilla-central/source/testing/xpcshell/head.js#605) and that is not exactly safe, maybe it is confusing other threads.
xpcom-shutdown is something else, quota manager shuts down before that when it gets profile-before-change and AFAIK, profile-before-change is fired via a cleanup function for xpcshell tests. This is the patch where I had to workaround it: https://hg.mozilla.org/mozilla-central/rev/8ff7083db0ea The test used a cleanup function to remove data for a domain, but that actually needs quota manager, so it tried to recreate quota manager after profile-before-change was already fired via a cleanup function that was registered sooner than the cleanup function in the test. Maybe it's unrelated to this bug, though.
(In reply to Jan Varga [:janv] from comment #11) > This is the patch where I had to workaround it: > https://hg.mozilla.org/mozilla-central/rev/8ff7083db0ea Ah, that's a bug in xpcshell head.js, it should run its cleanup function after all the others, not before them! I will file a separate bug to fix it, since it may be causing other intermittents. Not sure it's our culprit though, this test runs a cleanup earlier.
(In reply to Marco Bonardo [::mak] from comment #12) > (In reply to Jan Varga [:janv] from comment #11) > > This is the patch where I had to workaround it: > > https://hg.mozilla.org/mozilla-central/rev/8ff7083db0ea > > Ah, that's a bug in xpcshell head.js, it should run its cleanup function > after all the others, not before them! > I will file a separate bug to fix it, since it may be causing other > intermittents. Not sure it's our culprit though, this test runs a cleanup > earlier. Did this get filed? Fixed?
Flags: needinfo?(mak77)
(In reply to Kyle Huey [:khuey] (khuey@mozilla.com) from comment #15) > > Ah, that's a bug in xpcshell head.js, it should run its cleanup function > > after all the others, not before them! > > I will file a separate bug to fix it, since it may be causing other > > intermittents. Not sure it's our culprit though, this test runs a cleanup > > earlier. > > Did this get filed? Fixed? Oh sorry, I completely forgot, just filed Bug 1244723, should be easy to fix, but I'm not sure it's going to help this bug.
Flags: needinfo?(mak77)
Bulk assigning P3 to all open intermittent bugs without a priority set in Firefox components per bug 1298978.
Priority: -- → P3
Bug not seen for >1 year, WFM.
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.