[Bug tree-optimization/126473] [13/14/15/16/17 Regression] Wrong fold of signbit(x) -> x<0 for NaNs with -fno-signed-zeros

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126473

Drea Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |13.5
   Last reconfirmed|                            |2026-07-29
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
            Summary|Wrong fold of signbit(x) -> |[13/14/15/16/17 Regression]
                   |x<0 for NaNs                |Wrong fold of signbit(x) ->
                   |                            |x<0 for NaNs with
                   |                            |-fno-signed-zeros
      Known to work|                            |3.4.6
      Known to fail|                            |4.0.4

--- Comment #1 from Drea Pinski <pinskia at gcc dot gnu.org> ---
r0-56118-gef79730c93540e added it with __builtin_signbit.


It is a regression based on that.
You can reproduce it via `-O1 -g0 -ffast-math -fno-finite-math-only` and:
```
#include <math.h>

__attribute__((noipa)) int
sb (double x)
{
  return signbit (x) != 0;
}
```