[libcxx-commits] [libcxx] [libc++] Format into stack buffer before allocating for floats (PR #212502)

[Adib Pathan via libcxx-commits]

adxb-pxthxn wrote:

Thanks for the review — you're right that mapping unspecified precision to `6` changes semantics and is not the correct fix.

I've updated the patch to keep unspecified precision equivalent to calling `to_chars` without a precision argument. `__float_buffer` now starts on the stack in that case, and we allocate only if `to_chars` returns `errc::value_too_large` (as suggested in #211909). The allocation regression test is unchanged.

https://github.com/llvm/llvm-project/pull/212502