https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110353
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Diego Garcia from comment #1)
> I would like to add to this that some <cmath> functions (such as nextafter
> and nexttoward) have been constexpr since C++23, they are marked as such,
> but the internals aren't constexpr, leading to errors:
>
> main.cpp:4:57: in ‘constexpr’ expansion of ‘std::nextafter(0.0f, 1.0e+0f)’
> /opt/gcc-15/include/c++/15.1.0/cmath:2629:32: error:
> ‘__builtin_nextafterf(0.0f, 1.0e+0f)’ is not a constant expression
> 2629 | { return __builtin_nextafterf(__x, __y); }
> | ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
>
>
> Not sure if this would be a separate bug (due to being C++23 related) or
> still part of this
It is a seperate bug, -fno-trapping-math works around it ... There might be a
bug already filed too.
