Bug 31461 – REGRESSION(??? - 50919): various layout tests crashing on build and commit bots during JSC::Heap::collect()

• Status: NEW
• Component: New Bugs

[Eric Seidel (no email)]

http/tests/xmlhttprequest/access-control-and-redirects.html crashed on Leopard Commit Bot

https://bugs.webkit.org/show_bug.cgi?id=31366#c5
I'll attach a crash report tomorrow.

I expect this is related to, if not the same as bug 31453.

[Eric Seidel (no email)]

Another crash, same test:
https://bugs.webkit.org/show_bug.cgi?id=31406#c7

[Eric Seidel (no email)]

Created attachment 43180 [details]
Crash report from bug 31366

[Eric Seidel (no email)]

Created attachment 43181 [details]
crash report from bug 31406

[Eric Seidel (no email)]

Created attachment 43187 [details]
crash report from trying to land bug 31213

https://bugs.webkit.org/show_bug.cgi?id=31213#c5 also crashed on the same test!

[Eric Seidel (no email)]

The first recorded crash of this type was from 11/12/09 at 17:12 PST.  r50919 was the last change before that date.

[Eric Seidel (no email)]

Created attachment 43193 [details]
crash report from access-control-basic-allow-access-control-origin-header-data-url.html while trying to land bug 31468

Another failure.  Different test.  I dont' think we need a 7th but on the subject though, so adding it to this one.
https://bugs.webkit.org/show_bug.cgi?id=31468#c6
http/tests/xmlhttprequest/access-control-basic-allow-access-control-origin-header-data-url.html

[Eric Seidel (no email)]

Created attachment 43194 [details]
crash report from fast/workers/worker-terminate.html while trying to land bug 

fast/workers/worker-terminate.html
https://bugs.webkit.org/show_bug.cgi?id=31454#c6

This looks to be the same bug. :(

[Eric Seidel (no email)]

*** Bug 31453 has been marked as a duplicate of this bug. ***

[Eric Seidel (no email)]

*** Bug 31460 has been marked as a duplicate of this bug. ***

[Eric Seidel (no email)]

*** Bug 31480 has been marked as a duplicate of this bug. ***

[Eric Seidel (no email)]

*** Bug 31481 has been marked as a duplicate of this bug. ***

[Eric Seidel (no email)]

Created attachment 43208 [details]
Another crash, seen while trying to land bug 31468

[Eric Seidel (no email)]

This has exhibited similar symptoms to bug 30835.  Defining COLLECT_ON_EVERY_ALLOCATION did not cause this to crash more reliably.  The crash point is not the same as bug 30835, but it's similarly difficult to reproduce.

[Eric Seidel (no email)]

The bots look less crashy today.  I'll try re-enabling the commit-queue tomorrow and see if the crashiness has stopped.

[Eric Seidel (no email)]

Created attachment 43376 [details]
crash report from http/tests/xmlhttprequest/access-control-basic-allow-access-control-origin-header.html trying to land bug 31504

I tried to turn on the commit-queue again this morning.  Looks like this is still with us:
https://bugs.webkit.org/show_bug.cgi?id=31504#c8

[Darin Adler]

Oliver Hunt told me he is investigating this and thinks it may have something to do with JavaScript wrappers for worker thread objects.

[Oliver Hunt]

(In reply to comment #16)
> Oliver Hunt told me he is investigating this and thinks it may have something
> to do with JavaScript wrappers for worker thread objects.

I was looking at the older random crash.

[Eric Seidel (no email)]

Created attachment 43385 [details]
Crash report for repeated http/tests/xmlhttprequest/workers/shared-worker-methods.html runs

Actually, the reduced command I posted above seems to be crashing at a different place.  We may be chasing a red herring.

[Eric Seidel (no email)]

Created attachment 43386 [details]
different crash report from repeated http/tests/xmlhttprequest/workers/shared-worker-methods.html runs

Nevermind.  I'm seeing multiple crashes by running http/tests/xmlhttprequest/workers/shared-worker-methods.html repeatedly.  One of which is a crash under Heap::collect() as expected.

[Eric Seidel (no email)]

Comment on attachment 43385 [details]
Crash report for repeated http/tests/xmlhttprequest/workers/shared-worker-methods.html runs

Sorry, both of those crash reports were destined for bug 30835.