Bug 57030 - REGRESSION (r81782): http/tests/inspector/extensions-resources-redirect.html sometimes crashes WebKit2's web process while handling a WebIconDatabaseProxy::ReceivedIconLoadDecision message
Summary: REGRESSION (r81782): http/tests/inspector/extensions-resources-redirect.html ...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: WebKit2 (show other bugs)
Version: 528+ (Nightly build)
Hardware: Mac OS X 10.6
: P2 Normal
Assignee: Brady Eidson
URL: http://build.webkit.org/results/SnowL...
Keywords: InRadar, LayoutTestFailure
Depends on:
Blocks:
 
Reported: 2011-03-24 09:55 PDT by Adam Roben (:aroben)
Modified: 2011-03-24 13:23 PDT (History)
13 users (show)

See Also:


Attachments
Fix WebCore and unskip the test (3.56 KB, patch)
2011-03-24 13:17 PDT, Brady Eidson
aroben: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Adam Roben (:aroben) 2011-03-24 09:55:54 PDT
http/tests/inspector/extensions-resources-redirect.html sometimes crashes WebKit2's web process. We know this because the web process crashing then causes the UI process to crash (bug 57029). We don't know what's causing the web process to crash, though, because the UI process crash is masking it.
Comment 1 Adam Roben (:aroben) 2011-03-24 10:00:10 PDT
Ooh, I found a crash log from the web process. See the URL.
Comment 2 Adam Roben (:aroben) 2011-03-24 10:04:47 PDT
Looks like the first time this happened was in r81782, which seems like a related change.

http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28WebKit2%20Tests%29/builds/9955
Comment 3 Adam Roben (:aroben) 2011-03-24 10:05:09 PDT
Added the test to the mac-wk2 Skipped file in r81874.
Comment 4 Adam Roben (:aroben) 2011-03-24 10:07:38 PDT
<rdar://problem/9182074>
Comment 5 Brady Eidson 2011-03-24 13:08:18 PDT
Figured this out.

A serious of events have to happen with both the network layer and the WebProcess -> UIProcess layer to reproduce this.  There's probably no practical way to reproduce the exact conditions necessary in layout-test land, but we obviously already had coverage that caught this.

In fact, the single test Adam deactivated isn't the only one that can cause this - I've seen it locally with other https tests (though less frequently)

Patch is coming soon.
Comment 6 Brady Eidson 2011-03-24 13:17:20 PDT
Created attachment 86819 [details]
Fix WebCore and unskip the test
Comment 7 Brady Eidson 2011-03-24 13:23:20 PDT
r81892