[clang] [llvm] [LangRef] Clarify specification for float min/max operations (PR #172012)

[via cfe-commits]

ActuallyaDeviloper wrote:

Btw the signed zero behaviour of `minnum` in this lang ref change is identical to the one documented at the moment and released officially in LLVM 21. Although that change seemed to have been done without the necessary discussion, still if people are unable to reach consensus on any changes again, we seem to be stuck with this.

I hope the "solution" is not that Lang-ref remains half-baked (including SNaN) and one backend just "refuses" to follow it and makes up it's own definitions.

https://github.com/llvm/llvm-project/pull/172012