Bug 48996 - http/tests/websocket/tests/workers/close-in-onmessage-crash.html sometimes crashes on Windows
Summary: http/tests/websocket/tests/workers/close-in-onmessage-crash.html sometimes cr...
Status: RESOLVED WONTFIX
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: PC Windows XP
: P2 Normal
Assignee: Per Arne Vollan
URL: http://build.webkit.org/results/Windo...
Keywords: InRadar, LayoutTestFailure, PlatformOnly
Depends on:
Blocks: 55579
  Show dependency treegraph
 
Reported: 2010-11-04 07:44 PDT by Adam Roben (:aroben)
Modified: 2022-02-12 21:57 PST (History)
7 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Adam Roben (:aroben) 2010-11-04 07:44:39 PDT
http/tests/websocket/tests/workers/close-in-onmessage-crash.html sometimes crashes on Windows. See the URL for one such example.

I've added the test to the Skipped file.
Comment 1 Adam Roben (:aroben) 2010-11-04 07:55:48 PDT
<rdar://problem/8630448>
Comment 2 Yuta Kitamura 2010-11-04 19:52:43 PDT
To understand the situation better, I want to ask:

- Is the stack trace available?
- How frequent did the crash happen?
- Does this test crash on debug builds (not only release builds)?
Comment 3 Adam Roben (:aroben) 2010-11-05 05:26:10 PDT
(In reply to comment #2)
> - Is the stack trace available?

Unfortunately, the crash has only happened on our Windows 7 bot, and we don't have a way to get crash logs on Windows 7 currently. So: no. :-(

> - How frequent did the crash happen?

I looked back in the buildbot logs and recorded every time I saw the crash, until I got bored:

http://build.webkit.org/results/Windows%20Release%20(Tests)/r71321%20(5993)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71311%20(5985)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71306%20(5981)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71303%20(5979)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71233%20(5941)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71227%20(5938)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71206%20(5922)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71163%20(5901)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71113%20(5887)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71107%20(5881)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71099%20(5874)/results.html
http://build.webkit.org/results/Windows%20Release%20(Tests)/r71093%20(5869)/results.html

That's an average of 12 builds between crashes, but since we have two different bots running Windows Release (Tests) and only one of them is seeing the crash, that means it's happening once every 6 builds (on average).

> - Does this test crash on debug builds (not only release builds)?

Unfortunately the logs for the Debug testers seem to be screwed up (see <http://build.webkit.org/waterfall?last_time=1288749756&show=Windows%20Debug%20%28Tests%29>), so I can't say.
Comment 4 Adam Roben (:aroben) 2010-11-05 05:27:03 PDT
(In reply to comment #3)
> the crash has only happened on our Windows 7 bot

We have 2 Windows XP bots and 2 Windows 7 bots. The Windows XP bots have Cygwin 1.5/Python 2.5.2 installed, while the Windows 7 bots have Cygwin 1.7/Python 2.6.5. Maybe that's a clue?
Comment 5 Yuta Kitamura 2010-11-05 07:23:44 PDT
Thank you for providing information. I'm not sure about the root cause yet, and I will try to reproduce in my local environment later (but it may take some time to set up).
Comment 6 Yuta Kitamura 2010-11-19 05:49:18 PST
OK, I set up a new Windows 7 environment with cygwin 1.7, and I think I have reproduced the crash. I will take a close look.
Comment 7 Yuta Kitamura 2010-11-22 01:33:52 PST
Today I tried to track this down but failed. What I got was a bogus (corrupt) stack trace which only contained CoreFoundation.dll functions and a few functions from OS DLLs.

I'm not sure but my guess is that there is some sort of race condition behind this crash (because this test employs a worker).

This needs further investigation. Probably I should take a close look at WebSocket+worker implementation.
Comment 8 Yuta Kitamura 2010-11-22 01:35:48 PST
FYI, this crash occurs about 6% of the time on my local environment.