[libcxx-commits] [libcxx] [libc++] __need_infinity_nan doesn't work in non-gnu c++ standards (PR #172539)

[Louis Dionne via libcxx-commits]

https://github.com/ldionne commented:

Recording notes from live review:

- Looking at `float.h`, it's unclear to me why that header is needed at all. The underlying Clang `float.h` should define both `DECIMAL_DIG` and `FLT_EVAL_METHOD` itself, so there really isn't much point for libc++ to provide that header. We agreed to try and remove the header instead, like we've done for a few other C compatibility headers.
- **If** we end up needing this PR in some shape or form, I would prefer that we instead do the same treatment for all C compatibility headers, since they all need the same thing. That would make the pattern we're following clear, if not pretty.

For now, @ian-twilightcoder is going to open another PR where he tries to remove `<float.h>` entirely instead, and we'll see how far that takes us. That would certainly resolve the problems he was having with modules and would be OK from libc++'s perspective since that merely simplifies our codebase.

https://github.com/llvm/llvm-project/pull/172539