[cfe-dev] [3.3 Release] Release Candidate 2 Available

[Tim Northover]

> I don't really understand your logic here. We choose not to rely upon RTTI
> and exceptions because they violate the "only pay for what you use" policy
> of C++ - i.e. enabling them slows stuff down.
>
> Why would we disable use of something for speed reasons but then enable them
> when releasing? Isn't that somewhat pointless?

I suppose it depends on what the primary purpose of the release is. If
it's for the clang binary (or whatever) then yes, but if it's for the
libraries and API then leaving them disabled means a large number of
people can't actually use them.

Actually, I suppose at least the C++ API changes enough that the "test
against new LLVM binaries" isn't very plausible anyway.

Tim.