It passes on my local machine when I run-webkit-tests and also passes on webkit snowleopard bot, so I think it's a Mesa driver bug again.
Could its failure have something to do with how many tests are run in the run-webkit-tests invocation? We've seen cases before where certain tests set state in DumpRenderTree that affected later tests.
The failure is really weird. Basically I call uniform1i to set sampler2D and samplerCube uniforms, and read back the value. The failure is that the two values are swapped.
I could debug into Mesa code to see what's happening, but it has to wait.
LayoutTest failures for Chromium are being marked WontFix. The Bug is still accessible and referenced from TestExpectations.