Closed Bug 1108777 Opened 10 years ago Closed 9 years ago

There's some kind of racey corruption in test_bug600307-DBOps.html

Categories

(Core :: DOM: Core & HTML, defect)

Product:
Core
Component:
DOM: Core & HTML
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED DUPLICATE of bug 1165214

People

(Reporter: mccr8, Assigned: mayhemer)

References

Details

As you can see from the very long list of blocking bugs, there's something seriously wrong with test_bug600307-DBOps.html. Each of these failures hasn't happened more than a few times, generally speaking, but they are all happening in this one tests, and in such scattered areas of the code there must be some kind of race condition in local storage that is messing up memory, or something.
I bet there are even more of these failures that don't get filed, because they only happen once.
Am I misreading this test, or is the indentation hiding how nested all the tests are within each other?
In every single log I've looked at, the crashes appear to happen after the same line: 3788 INFO TEST-PASS | /tests/dom/tests/mochitest/localstorage/test_bug600307-DBOps.html | Item deleted in case 1 Linux32 debug mochitest-e10s is far and away the most common suite to hit these crashes, but there are some x64 and opt/asan crashes mixed in for good measure. I'm working on a Try stress test right now to see if we can reliably crash it.
This is a great candidate for rr.
You need to have the specialpowers extension installed to be able to test it: http://people.mozilla.org/~mwargers/extensions/specialpowers/specialpowers.htm
Thanks for this umbrella bug. Will take a look.
Assignee: nobody → honzab.moz
I'm probably going to disable this test on Linux soon. Hitting these crashes way too frequently.
Ryan, did you disable this? Say on Dec 17th? Because the failures seem to have mostly stopped then.
Flags: needinfo?(ryanvm)
Interestingly, no. Bug 1108565 does show one failure since then, but you're right that something definitely dropped the failure rate dramatically.
Flags: needinfo?(ryanvm)
Around when it went away, Terrence landed bug 1110931, which would eliminate one source of races. But one crash did happen on revision 41a8532f3fc3, which is after that landed. Ben also landed bug 701634 in there, maybe that could have had an effect? https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?startdate=12%2F16%2F14&enddate=12%2F17%2F14
localStorage has nothing to do with IndexDB, maybe except the backend (sqlite). Bad I didn't get to this sooner.. will be harder to catch now. Still, I can return to an older revision and try to find out.
Ryan, what is the status here? Is the test disabled? Are the crashes still occurring in such high and random rate?
Flags: needinfo?(ryanvm)
The most recent report I can see is https://bugzilla.mozilla.org/show_bug.cgi?id=1108565#c17 and it's from Jan 5 2015. More than a month old.
No, this was never disabled. The crashes all seem to have disappeared over the last couple months. Would be cool to know why, but who knows.
Flags: needinfo?(ryanvm)
During work on bug 1165214 I found out we were touching a certain dom storage specific hashtable on multiple threads. The original cause of this bug could be that. I'll duplicate it to that bug.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → DUPLICATE
Component: DOM → DOM: Core & HTML
You need to log in before you can comment on or make changes to this bug.