Bug 57048 - http/tests/websocket/tests/workers/worker-handshake-challenge-randomness.html crashed once on Windows XP Debug (Tests)
Summary: http/tests/websocket/tests/workers/worker-handshake-challenge-randomness.html...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Page Loading (show other bugs)
Version: 528+ (Nightly build)
Hardware: PC Windows XP
: P2 Normal
Assignee: Fumitoshi Ukai
URL: http://build.webkit.org/results/Windo...
Keywords: InRadar, LayoutTestFailure, PlatformOnly
Depends on:
Blocks: 55579
  Show dependency tree / graph
 
Reported: 2011-03-24 13:12 PDT by Adam Roben (:aroben)
Modified: 2016-05-18 21:14 PDT (History)
8 users (show)

See Also:

Attachments
Patch (1.78 KB, patch)
2011-05-18 19:14 PDT, Fumitoshi Ukai 		no flags Details | Formatted Diff | Diff
Patch (1.78 KB, patch)
2011-05-18 19:21 PDT, Fumitoshi Ukai 		ap: review+
Details | Formatted Diff | Diff
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Adam Roben (:aroben) 2011-03-24 13:12:16 PDT 
http/tests/websocket/tests/workers/worker-handshake-challenge-randomness.html crashed once on Windows XP Debug (Tests). See the URL for the crash log. It looks like the same CFSocket is being modified on multiple threads concurrently.
Comment 1 Adam Roben (:aroben) 2011-03-24 13:13:06 PDT 
Seems related to bug 56625.
Comment 2 Adam Roben (:aroben) 2011-03-24 13:14:52 PDT 
<rdar://problem/9183661>
Comment 3 David Levin 2011-03-24 13:15:25 PDT 
+ ukai due to websocket issue.
Comment 4 Adam Roben (:aroben) 2011-03-25 12:36:35 PDT 
This test is now being skipped (see bug 55579).
Comment 5 Alexey Proskuryakov 2011-05-18 17:19:22 PDT 
This is a fairly obvious object lifetime bug, and it's probably reasonably easy to fix, but I couldn't figure it out in the time I had for this so far.
Comment 6 Fumitoshi Ukai 2011-05-18 18:21:39 PDT 
(In reply to comment #5)
> This is a fairly obvious object lifetime bug, and it's probably reasonably easy to fix, but I couldn't figure it out in the time I had for this so far.

I don't know much about CoreFoundation, but it seems it schedules read/write stream on loaderRunLoop() on PLATFORM(WIN) in scheduleStreams(), but it tries to unschedule stream on CFRunLoopGetCurrent(), which is used in scheduling stream on other platform than WIN.
So, I think this can be fixed to use loaderRunLoop() on PLATFORM(WIN) in platformClose().
I'll create a patch.
Comment 7 Fumitoshi Ukai 2011-05-18 19:14:28 PDT 
Created attachment 94029 [details]
Patch
Comment 8 Fumitoshi Ukai 2011-05-18 19:21:32 PDT 
Created attachment 94030 [details]
Patch
Comment 9 Yuta Kitamura 2011-05-18 20:01:02 PDT 
Comment on attachment 94030 [details]
Patch

Looks good to me. (I'm not a reviewer, so I can't r+ this.)
Comment 10 Alexey Proskuryakov 2011-05-18 23:18:02 PDT 
Comment on attachment 94030 [details]
Patch

Heh, this is a good catch! I'm not sure if it takes care of the whole problem - do you have a Windows setup to run the tests?
Comment 11 Alexey Proskuryakov 2011-05-18 23:23:06 PDT 
Comment on attachment 94030 [details]
Patch

I'm going to say r+, since this is clearly an improvement, but the bug may need to remain open until someone can test on Windows, and until the skipped tests are re-enabled.

See also: <http://trac.webkit.org/changeset/81111> (where I unsuccessfully tried to fix the crashes).
Comment 12 Fumitoshi Ukai 2011-05-18 23:40:31 PDT 
Committed r86822: <http://trac.webkit.org/changeset/86822>
Comment 13 Fumitoshi Ukai 2011-05-18 23:43:20 PDT 
I just landed the patch (id=94030).
I don't have windows, so please someone test whether it fixes problems.
I leave this bug open for now.
Comment 14 Brent Fulgham 2016-05-18 21:14:13 PDT 
This code was landed five years ago. I think we can close it.