[libcxx-commits] [libcxx] [locale][num_get] Improve Stage 2 of string to float conversion (PR #65168)

[Mark de Wever via libcxx-commits]

mordante wrote:

> It seems possible to ship this without breaking the ABI. And assuming the new behavior doesn't start to reject meaningful previously accepted code then I think we should ship it retroactively.

I also don't see why this has to be an ABI break. What is the reason?

> [LWG2381](https://cplusplus.github.io/LWG/issue2381) (not implemented in libc++ yet) intendedly rejects Inf and NaN. Should we restrict the handling of Inf and NaN to some extension mode?

IMO we should just implement the LWG issue and not have an extension mode.


https://github.com/llvm/llvm-project/pull/65168