Re: [PATCH] libstdc++: Make chrono::hh_mm_ss more compact

[Jonathan Wakely <jwakely.gcc@gmail.com>]

On Mon, 12 Dec 2022 at 11:18, Jonathan Wakely wrote:
>
> On Mon, 12 Dec 2022 at 07:12, Stephan Bergmann via Libstdc++
> <libstdc++@gcc.gnu.org> wrote:
> >
> > On 11/21/22 21:43, Jonathan Wakely via Libstdc++ wrote:
> > > +     static constexpr bool __is_unsigned
> > > +       = __and_v<is_integral<typename _Duration::rep>,
> > > +                 is_unsigned<typename _Duration::rep>>;
> >
> > Using `__is_unsigned` as an identifier here causes compilation issues
> > with Clang, which predefines that as a builtin predicate.
>
> Great, those are going to break a lot of code:
>
> include/bits/charconv.h:      = ! __gnu_cxx::__int_traits<_Tp>::__is_signed;
> include/bits/locale_facets.tcc:   (__negative && __num_traits::__is_signed)
> include/bits/locale_facets.tcc:     if (__negative && __num_traits::__is_signed)
> include/bits/locale_facets.tcc:             &&
> __gnu_cxx::__numeric_traits<_ValueT>::__is_signed)
> include/bits/uniform_int_dist.h:
> static_assert(!_Up_traits::__is_signed, "U must be unsigned");
> include/bits/uniform_int_dist.h:
> static_assert(!_Wp_traits::__is_signed, "W must be unsigned");
> include/ext/numeric_traits.h:      static const bool __is_signed =
> (_Value)(-1) < 0;
> include/ext/numeric_traits.h:   =
> __is_integer_nonstrict<_Value>::__width - __is_signed;
> include/ext/numeric_traits.h:      static const _Value __max = __is_signed
> include/ext/numeric_traits.h:      static const _Value __min =
> __is_signed ? -__max - 1 : (_Value)0;
> include/ext/numeric_traits.h:    const bool
> __numeric_traits_integer<_Value>::__is_signed;
> include/ext/numeric_traits.h:      static const bool __is_signed = true;
> include/ext/numeric_traits.h:    const bool
> __numeric_traits_floating<_Value>::__is_signed;
> include/ext/numeric_traits.h:      static const bool __is_signed = true;
> include/ext/numeric_traits.h:      static const bool __is_signed = true;
>
> If Clang already treats __is_signed as a conditionally-active
> built-in, it should do the same for __is_unsigned.

The __is_unsigned use has been removed in r13-4613-gcb363fd9f19eb7