GeezerVison

@encukou wrote:

NaNs are part of IEEE 754, and the math functions are part of C11, so mentioning the earlier changes are redundant now.

Well, I chose to be extra explicit, since compilers have the bad habit of not implementing fully or exactly C standards. For example, see Modernizing Fedora's C code about GCC and clang which don't respect C99/C11 "for backward compatibility".

Anything blocking this to be merged?

Hey @vstinner @encukou do you think this should be changed or some of this reverted, based on the feedback above? Or should we go ahead with it?

Hey @vstinner @encukou , further thoughts on this?

Sorry, this PR fell through the cracks for me. I'll review this week.

Thanks @encukou . Also, it sounds like Victor is on hiatus for now and passed on reviewing this.

Well, I chose to be extra explicit, since compilers have the bad habit of not implementing fully or exactly C standards. For example, see Modernizing Fedora's C code about GCC and clang which don't respect C99/C11 "for backward compatibility".

Yeah, but that (the example at least) is about keeping old behavior in cases where C99 says something is illegal or undefined. (AFAIK, the standard mandates an error, so the implementation doesn't comply with the standard, but it's fine in the real world.)
Leaving out <math.h> (or a function from it) would be pretty hard to sell as a deliberate backwards-compatibility quirk. So in this particular case, only requiring a C11 libc should be fine.

Thanks @CAM-Gerlach for the PR, and @encukou for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11.
🐍🍒⛏🤖

GH-102497 is a backport of this pull request to the 3.11 branch.