(In reply to Boris Zbarsky [:bzbarsky, bz on IRC] from comment #2)
I'm not sure I'm OK with this until we actually stop using UniversalXPConnect...
Should we just disable this test when Fission is enabled for the moment? And if we don't plan to stop using UniversalXPConnect before Fission ships, we probably do want to fix this test....
We could just change it to use a cross-origin, same-process frame, but I'd honestly rather just delete it. The only things that still rely on UniversalXPConnect are a few scattered tests, and I don't think we need to ensure it preserves its current exact semantics strictly for their sake.
That said, I also don't think this test makes any kind of sense anymore. It calls
enablePrivilege in a function scope, and assumes that the privilege only lasts to the end of that scope. But that only made sense when
enablePrivilege worked by examining the stack for signed code which had added privileges. What remains of it now just works at the compartment level. It doesn't care about stacks, and it certainly doesn't go away at the end of the scope that it was called in.