I think the issue in question is #373. This is really strange and I just managed to create a minimal test case for it: http://jsbin.com/ezidox/25

Note that the case only fails with Safari 6 and only if the inspector is closed.

Furthermore, it looks to me like Safari starts failing only after calling the function a few times. This suggests to me, that maybe it's due to a Nitro optimisation gone wrong.

Assuming that this issue is deterministic then it was resolved by 94e7004 which is available in rc3. I couldn't even begin to tell you WHY that might have resolved it but after bisecting this it seems pretty reliable that this fixed this issue.

Closing this PR as it feels like it actually masks the incorrect behavior where we actually want things to fail as there is something seriously wrong if that method receives undefined values.