[PATCH] Downgrade the error about nested name specifiers when building with Microsoft extensions

[David Majnemer]

Usually we only downgrade this kind of error if it was blocking a system header, code that was automatically generated by some sort of IDL, or if it is impractical to change the code in some other way.  Is that the case here?

It's preferable, in my eyes, to keep it an error so that programmers make their code conforming instead of just ignoring another warning fly by on their terminal.

I don't think clang-cl should mirror every bug in cl, just the important ones.

http://reviews.llvm.org/D4333