It was disabled by b3c814e2d015. It's not too bad, only the large-negative-value test is failing on 32bit machines. Regressed either in bug 732660 or with the porting to new bindings.
This was regressed by: changeset: 109965:4c134de65570 user: Benoit Jacob <firstname.lastname@example.org> date: Thu Oct 11 14:23:34 2012 -0400 summary: Bug 779611 - WebGL extensions should be [NoInterfaceObject] as their compliant IDL interface names would pollute the global object - r=bz
Erm, no, rather by changeset: 109966:78c16f313547 user: Benoit Jacob <email@example.com> date: Thu Oct 11 15:23:03 2012 -0400 summary: Bug 779611 - part 12: remove old-bindings remnants in the WebGL implementation - r=bz
Yeah OK I know what this almost certainly is --- the WebGLintptr typedef was changed to a typedef to GLintptr, and likely these were not actually the same.
OK, indeed: WebGLintptr was long long which is int64 GLintptr is ptrdiff_t. Whence the bug on 32bit platforms
Created attachment 671006 [details] [diff] [review] fix WebGL typedefs to reflect IDL types instead of being equal to OpenGL types Again, the bug was that while most WebGL types are equal to the corresponding OpenGL type, some are different: WebGLintptr is always long long (int64) in Web IDL while it is intptr_t in OpenGL WebGLsizeiptr is the same I also changed WebGLboolean to be bool, following Web IDL. Also removed now-useless #include "WebGLContext.h" in nsDOMCI. And the #undef None there was almost certainly added there for it (X11 headers have a None macro) and affected WebGLContext.h (which has or used to have a None enum value somewhere). Removed too. Reenabling the unit test.
Comment on attachment 671006 [details] [diff] [review] fix WebGL typedefs to reflect IDL types instead of being equal to OpenGL types r=me
Somehow, the part of that patch that actually reenables the test was dropped. Landed now: http://hg.mozilla.org/integration/mozilla-inbound/rev/088dfedf36d5