Bug 46870 – fast/canvas/webgl/draw-elements-out-of-bounds.html is crashing on Snow Leopard

• Status: RESOLVED
• Resolution: FIXED
• Component: WebGL

[Eric Seidel (no email)]

fast/canvas/webgl/draw-elements-out-of-bounds.html is crashing on Snow Leopard

http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r68740%20(18485)/fast/canvas/webgl/draw-elements-out-of-bounds-stderr.txt

Assertion failed: (0 && "InitThread(): Process hasn't been initalised."), function InitThread, file /Volumes/Data/WebKit-BuildSlave/snowleopard-intel-release/build/ANGLE/src/compiler/InitializeDll.cpp, line 71.

The crashing started after I turned off some canvas tests as part of this rollout:
http://trac.webkit.org/changeset/68737

Since skipping tests should not cause crashes, there was likely a crash here before, masked by other factors.

Going to skip this canvas test until zmo/kbr can take a look.

[WebKit Commit Bot]

Created attachment 69307 [details]
Skip the test

[WebKit Commit Bot]

Committed r68745: <http://trac.webkit.org/changeset/68745>

[Zhenyao Mo]

I'll take a look.  Looks like Angle issue, but it could be misleading.

[Adam Roben (:aroben)]

<rdar://problem/8496585>

[Zhenyao Mo]

Created attachment 69374 [details]
patch

After discussing with Alok and Dimitri, this is an educated attempt.

The cause (from the assertion failure) could be the previous test cleanup is delayed and the current test called ShInitialize() before the ShFinalize() from the previous test.

[Eric Seidel (no email)]

Do we finalize in some other way? or after this change would we never finalize?  And what are the consequences of such?

[Kenneth Russell]

Comment on attachment 69374 [details]
patch

We've discussed this earlier via email and this is the correct fix at the present time.

[Zhenyao Mo]

Committed r68833: <http://trac.webkit.org/changeset/68833>

[Eric Seidel (no email)]

My questions were never answered, but I guess I'm not on the right mailing list. :)

Does this need a FIXME for the future?

[Kenneth Russell]

(In reply to comment #9)
> My questions were never answered, but I guess I'm not on the right mailing list. :)
> 
> Does this need a FIXME for the future?

Not really. At best we would need an at-exit hook, but OS process separation will take care of the cleanup of the needed static and thread-local resources.

[Zhenyao Mo]

(In reply to comment #9)
> My questions were never answered, but I guess I'm not on the right mailing list. :)
> 
> Does this need a FIXME for the future?

alokp is working on getting rid of ShInitialize and ShFinalize.  We'll just live with this until that happens.