A couple people asked me when it is ok to revert a patch.
The LLVM project uses a very aggressive form of code review. It is basically built on the assumption that people do the right thing most of the time, and optimizes for that case.
However, people do make mistakes. When something bad happens, don't be shy about reverting a broken patch, particularly when it is an obvious problem. The author may be away, and there is no reason to wait for them to fix it. When they get a chance, they can fix the patch and recommit.
There is nothing personal about it. I think too often we wait for patches to be fixed, rather than *immediately* reverting a known broken patch. If you have run up against a problem, you can be sure that many other people are hitting the same problem too, they just may not be saying anything about it.
When it doubt, revert