[libcxx-commits] [libcxx] [libc++][math] Generalize workaround for UCRT to `signbit` (PR #212398)

[Lucas Mellone via libcxx-commits]

lknknm wrote:

> > I'm really not a big fan. This adds an incredible amount of additional code (as-in doubling the amount of code per math function we have to maintain) for supporting a single libc, which will inevitably have to implement this anyways.
> 
> Hmm, do we want to leave the workaround for MS UCRT half baked? I think I can live with removal of workarounds (which will also effectively revert #149234).

I don't know if this could be a strategy but I'm thinking about implementing P0533R9 in the order of:
- Finishing the libc++ impl for the `built-ins` already available, even if it needs the UCRT workaround;
- Finishing the clang-side implementations for built-ins (not myself completely but I could try to help);
- Finishing the remaining libc++ impl.
- Removing UCRT workarounds as they become available.

It's merely a suggestion but I'm not so fond of it either. Is there a way we can track the MS UCRT implementations? I saw from a Stephen Lavavej comment that MSVC's STL also has workarounds for the UCRT issues.

https://github.com/llvm/llvm-project/pull/212398