[clang] [clang] Do not clear FP pragma stack when instantiating functions (PR #70646)

[John McCall via cfe-commits]

rjmccall wrote:

That makes sense.  Out of curiosity, since presumably the instantiation path is using the RAII class, and the RAII class seems to save and restore the entire stack, what actually ends up going wrong?  It doesn't look like your test case has local changes to the stack within the instantiated function.  I'm just wondering whether we maybe just have a bad API here.

https://github.com/llvm/llvm-project/pull/70646