THe error is with gcc-3.exe hanging on a could not load w, win 32 error 998 (ERROR_NO_ACCESS) It happens in different places throughout the test suite at different times. One observation is that there are always two gcc-3.exe processes running at the same time when the crash occurs. I'm not sure if this is to be expected. It's possible that the two gcc-3's are trying to use the same library, and one is blocking access from the other, causing the hang. One gcc-3 process has all the arguments passed in from the bindings generation tests and the other has no arguments, I have no clue where the latter is coming from but it may be the problem process. Skipping these tests on the Windows port for now. http://trac.webkit.org/changeset/146827 They aren't particularly high priority for the WinEWS bots regardless.
Is there a build log showing the problem? I'll try running the tests on my machine at work tomorrow and see if I can recreate it.