Bug 50649 - REGRESSION(r73378-r73395): Leopard layout-test failures after build 24324
Summary: REGRESSION(r73378-r73395): Leopard layout-test failures after build 24324
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Mac (Intel) OS X 10.5
: P2 Normal
Assignee: Nobody
URL:
Keywords: LayoutTestFailure, PlatformOnly, Regression
Depends on:
Blocks:
 
Reported: 2010-12-07 15:42 PST by Deirdre Saoirse Moen
Modified: 2010-12-14 17:20 PST (History)
5 users (show)

See Also:


Attachments
Crash Log (69.28 KB, application/octet-stream)
2010-12-08 14:42 PST, Stephanie Lewis
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Deirdre Saoirse Moen 2010-12-07 15:42:35 PST
1. http://build.webkit.org/builders/Leopard%20Intel%20Debug%20%28Tests%29/builds/24324 was clean for Leopard layout tests
2. http://build.webkit.org/builders/Leopard%20Intel%20Debug%20%28Tests%29/builds/24325 had two build failures and one crash:

fast/blockflow/broken-ideograph-small-caps.html -> failed
fast/events/tabindex-focus-chain.html -> crashed
http/tests/misc/acid3.html -> failed

3. The only reported change affecting that set of tests was r73395, an unreviewed build fix.
Comment 1 Ryosuke Niwa 2010-12-07 15:59:50 PST
(In reply to comment #0)
> 3. The only reported change affecting that set of tests was r73395, an unreviewed build fix.

You mean http://trac.webkit.org/changeset/73395 ?  I don't think it can ever cause that kind of failures.  How does explicitly casting double to float that has already been implicitly casted before will cause a test failure?

As the build bot pages indicate, it's a regression between r73378 and r73395.
Comment 2 Ryosuke Niwa 2010-12-07 16:07:05 PST
tabindex-focus-chain.html crash might be due to http://trac.webkit.org/changeset/73388.

The other two tests seem to be just being flaky.  They're passing on release builds but failing on debug builds.
Comment 3 Deirdre Saoirse Moen 2010-12-07 16:13:47 PST
I'm new at looking through the test failures (and thus also to flaky tests), so bear with me, but it looked like r73378 was included in build 24324, which is why it didn't seem like it was the cause.
Comment 4 chris fleizach 2010-12-07 16:57:52 PST
(In reply to comment #2)
> tabindex-focus-chain.html crash might be due to http://trac.webkit.org/changeset/73388.
> 
> The other two tests seem to be just being flaky.  They're passing on release builds but failing on debug builds.

what's the crash log for this one. i can't find it
Comment 5 Ryosuke Niwa 2010-12-07 16:58:16 PST
(In reply to comment #3)
> I'm new at looking through the test failures (and thus also to flaky tests), so bear with me, but it looked like r73378 was included in build 24324, which is why it didn't seem like it was the cause.

Right, 78 was but not 88.
Comment 6 Ryosuke Niwa 2010-12-07 16:59:14 PST
(In reply to comment #4)
> what's the crash log for this one. i can't find it

I don't think we can grab the crash log from bots.
Comment 7 Stephanie Lewis 2010-12-08 14:42:49 PST
Created attachment 75971 [details]
Crash Log

Adding the last DumpRenderTree crash log from the bot.  This test is the only one crashing at the moment so they should match up.
Comment 8 chris fleizach 2010-12-08 14:44:46 PST
Exception Type:  EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x00000000667f3d15
Crashed Thread:  0

Thread 0 Crashed:
0   com.apple.WebCore             	0x04bc6939 WebCore::GraphicsLayer::contentsOrientation() const + 9 (GraphicsLayer.h:328)
1   com.apple.WebCore             	0x04bc63b0 drawLayerContents + 82 (WebLayer.mm:51)
2   com.apple.WebCore             	0x04bd0853 -[WebTiledLayer drawInContext:] + 59 (WebTiledLayer.mm:87)
3   com.apple.QuartzCore          	0x94b8336c tiled_layer_render + 452
4   com.apple.QuartzCore          	0x94b3d267 source_perform + 259
5   com.apple.CoreFoundation      	0x96d823c5 CFRunLoopRunSpecific + 3141
6   com.apple.CoreFoundation      	0x96d82aa8 CFRunLoopRunInMode + 88
7   com.apple.Foundation          	0x942d63d5 -[NSRunLoop(NSRunLoop) runMode:beforeDate:] + 213
8   DumpRenderTree                	0x00016270 __ZL7runTestRKSs + 1844
9   DumpRenderTree                	0x00016725 __ZL20runTestingServerLoopv + 119
10  DumpRenderTree                	0x000169b8 dumpRenderTree(int, char const**) + 340
11  DumpRenderTree                	0x00016be4 main + 94 (DumpRenderTree.mm:708)
12  DumpRenderTree                	0x00002f72 start + 54
Comment 9 chris fleizach 2010-12-08 14:45:36 PST
(In reply to comment #8)
> Exception Type:  EXC_BAD_ACCESS (SIGSEGV)
> Exception Codes: KERN_INVALID_ADDRESS at 0x00000000667f3d15
> Crashed Thread:  0
> 
> Thread 0 Crashed:
> 0   com.apple.WebCore                 0x04bc6939 WebCore::GraphicsLayer::contentsOrientation() const + 9 (GraphicsLayer.h:328)
> 1   com.apple.WebCore                 0x04bc63b0 drawLayerContents + 82 (WebLayer.mm:51)
> 2   com.apple.WebCore                 0x04bd0853 -[WebTiledLayer drawInContext:] + 59 (WebTiledLayer.mm:87)
> 3   com.apple.QuartzCore              0x94b8336c tiled_layer_render + 452
> 4   com.apple.QuartzCore              0x94b3d267 source_perform + 259
> 5   com.apple.CoreFoundation          0x96d823c5 CFRunLoopRunSpecific + 3141
> 6   com.apple.CoreFoundation          0x96d82aa8 CFRunLoopRunInMode + 88
> 7   com.apple.Foundation              0x942d63d5 -[NSRunLoop(NSRunLoop) runMode:beforeDate:] + 213
> 8   DumpRenderTree                    0x00016270 __ZL7runTestRKSs + 1844
> 9   DumpRenderTree                    0x00016725 __ZL20runTestingServerLoopv + 119
> 10  DumpRenderTree                    0x000169b8 dumpRenderTree(int, char const**) + 340
> 11  DumpRenderTree                    0x00016be4 main + 94 (DumpRenderTree.mm:708)
> 12  DumpRenderTree                    0x00002f72 start + 54

unlikely this was caused
http://trac.webkit.org/changeset/73388
Comment 10 Stephanie Lewis 2010-12-08 14:56:54 PST
> The other two tests seem to be just being flaky.  They're passing on release builds but failing on debug builds.

I don't think its accurate to call these test flaky.  They've been consistently failing on two different machines for two days.  There are enough differences between debug builds and release builds that a failure could only occur on one type of build.   This wouldn't be the first time.

There several different graphics patches that went into the tree during that range.  I don't have access to a leopard machine at the moment, but if nobody else makes progress I'll try and pin down the range tomorrow.
Comment 11 Stephanie Lewis 2010-12-14 16:08:10 PST
The acid3 failure was caused by http://trac.webkit.org/projects/webkit/changeset/73385.  

The other two failures I haven't been able to repro yet.
Comment 12 Darin Adler 2010-12-14 17:20:52 PST
(In reply to comment #11)
> The acid3 failure was caused by http://trac.webkit.org/projects/webkit/changeset/73385.  
> 
> The other two failures I haven't been able to repro yet.

Are we going to use a separate bug for the Acid3 failure, or track it with this bug?