Bug 100408 – REGRESSION (r132328): /WebKit2APITests/TestWebKitAccessibility unit test is failing

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Zan Dobersek]

The TestWebKitAccessibility unit test has been failing after the run-gtk-tests script was modified in r132328.
http://trac.webkit.org/changeset/132328

I'll skip the test for now.

[Zan Dobersek]

Of course, I should give some more information about the failing test.
Here's the output:

TEST: ./Tools/gtk/../Scripts/../../WebKitBuild/Release/Programs/WebKit2APITests/TestWebKitAccessibility... (pid=12069)

failed to create drawable

  /webkit2/WebKitAccessibility/atspi-basic-hierarchy:                  **

ERROR:../../Source/WebKit2/UIProcess/API/gtk/tests/TestWebKitAccessibility.cpp:136:void checkAtspiAccessible(AtspiAccessible*, const char*, AtspiRole): assertion failed (targetRole == atspi_accessible_get_role(accessible, 0)): (23 == 0)

FAIL

[Mario Sanchez Prada]

(In reply to comment #1)
> Of course, I should give some more information about the failing test.
> Here's the output:
> 
> TEST: ./Tools/gtk/../Scripts/../../WebKitBuild/Release/Programs/WebKit2APITests/TestWebKitAccessibility... (pid=12069)
> 
> failed to create drawable

This doesn't seem to me like a fail in the test but in the environment in the buildbot. Actually, I have just logged in the buildbot, run the test and it works fine, so maybe we should unskip it again and give it another shot.

What do you think?

>   /webkit2/WebKitAccessibility/atspi-basic-hierarchy:                  **
> 
>ERROR:../../Source/WebKit2/UIProcess/API/gtk/tests/TestWebKitAccessibility.cpp:1
> 36:void checkAtspiAccessible(AtspiAccessible*, const char*, AtspiRole): 
> assertion failed (targetRole == atspi_accessible_get_role(accessible, 0)): 
> (23 == 0)
> 

Couldn't reproduce this, neither in the bots neither locally (and I have tried both with Release and Debug builds, right after updating my local repo this morning)

[Mario Sanchez Prada]

(In reply to comment #2)
> (In reply to comment #1)
> > Of course, I should give some more information about the failing test.
> > Here's the output:
> > 
> > TEST: ./Tools/gtk/../Scripts/../../WebKitBuild/Release/Programs/WebKit2APITests/TestWebKitAccessibility... (pid=12069)
> > 
> > failed to create drawable
> 
> This doesn't seem to me like a fail in the test but in the environment in the 
> buildbot. Actually, I have just logged in the buildbot, run the test and it 
> works fine, so maybe we should unskip it again and give it another shot.

Hmmm.. one clarification: I tried running this in the normal buildbots, not in the WebKit2 one (no idea how to try it out there). Maybe Sergio (added to CC) could help us with that? 

JFTR, I'd just run this command:

$ Tools/Scripts/run-gtk-tests \
     WebKitBuild/Release/Programs/WebKit2APITests/TestWebKitAccessibility

[Sergio Villar Senin]

(In reply to comment #3)
> (In reply to comment #2)
> > (In reply to comment #1)
> > > Of course, I should give some more information about the failing test.
> > > Here's the output:
> > > 
> > > TEST: ./Tools/gtk/../Scripts/../../WebKitBuild/Release/Programs/WebKit2APITests/TestWebKitAccessibility... (pid=12069)
> > > 
> > > failed to create drawable
> > 
> > This doesn't seem to me like a fail in the test but in the environment in the 
> > buildbot. Actually, I have just logged in the buildbot, run the test and it 
> > works fine, so maybe we should unskip it again and give it another shot.
> 
> Hmmm.. one clarification: I tried running this in the normal buildbots, not in the WebKit2 one (no idea how to try it out there). Maybe Sergio (added to CC) could help us with that? 

Not sure what you mean Mario, the WK2 behaves exactly the same as any other bot, the difference is that we just copy the WebKitBuild/Release folder there instead of generating it in the bot. There should be no difference.

[Mario Sanchez Prada]

(In reply to comment #4)
> [...]
> Not sure what you mean Mario, the WK2 behaves exactly the same as any other 
> bot, the difference is that we just copy the WebKitBuild/Release folder there 
> instead of generating it in the bot. There should be no difference.

I just meant that I was not able to try it in the WK2 and so that I could not be sure whether there could be any (subtle?) difference explaining why it worked in the non-WK2 bots but not in the other one.

If you (or anyone else with access to the WK2 bot) could try this out that would be of great help I think.

[Mario Sanchez Prada]

(In reply to comment #5)
> (In reply to comment #4)
> > [...]
> > Not sure what you mean Mario, the WK2 behaves exactly the same as any other 
> > bot, the difference is that we just copy the WebKitBuild/Release folder there 
> > instead of generating it in the bot. There should be no difference.
> 
> I just meant that I was not able to try it in the WK2 and so that I could not be sure whether there could be any (subtle?) difference explaining why it worked in the non-WK2 bots but not in the other one.
> 
> If you (or anyone else with access to the WK2 bot) could try this out that would be of great help I think.

I've just talked to Sergio again now and he confirms that running this test directly in the WK2 is working fine now, so we believe we can give it a try and unskip it.

[Mario Sanchez Prada]

Committed r155038: <http://trac.webkit.org/changeset/155038>