Add AVR __[u]int24 to type_traits

[kevinpt]

Support the 24-bit integer types __int24 and __uint24 in type_traits.

[jwakely]

Why are these not already enabled as __GLIBCXX_TYPE_INT_N_1 ?

That's how non-standard integer types are supposed to be supported, and how it works for other targets.

[kevinpt]

It didn't work for me when I tried to use type_traits on these types with avr-g++ 14.2.

…

-- Kevin Thibedeau

On Sun, Dec 29, 2024 at 7:18 PM Jonathan Wakely ***@***.***> wrote: Why are these not already enabled as __GLIBCXX_TYPE_INT_N_1 ? That's how non-standard integer types are supposed to be supported, and how it works for other targets. — Reply to this email directly, view it on GitHub <#37 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABTTUQ4YSITNXMGW2PTXCLD2ICGLXAVCNFSM6AAAAABULG4O5GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNRUHA4DANJXHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[jwakely]

If avr-g++ supports those types, then it should define the macros to tell libstdc++ that it supports them. Then everything in the library will just work.

[kevinpt]

This test program fails to compile with the existing code. It works fine with standard integer types:

  #include <type_traits>

  // Templated function to increment only integral types
  template <typename T>
  typename std::enable_if<std::is_integral<T>::value, T>::type increment(T value) {
      return value + 1;
  }

  bool demo(__int24 n) {
    return increment(n) > n;
  }

avr-g++ -std=c++20 -I lib/avr-libstdcpp/include -c type_traits_bug.cc -o type_traits_bug.o

  ...
  type_traits_bug.cc:5:62: note:   template argument deduction/substitution failed:
  type_traits_bug.cc: In substitution of 'template<class T> typename std::enable_if<std::is_integral<_Tp>::value, T>::type increment(T) [with T = __int24]':
  ...

[jwakely]

I think a better fix would be to add this to include/bits/c++config.h

#if defined(__INT24_MAX__) && !defined(__GLIBCXX_TYPE_INT_N_0)
# define __GLIBCXX_TYPE_INT_N_0 __int24
# define __GLIBCXX_BITSIZE_INT_N_0 24
#endif

This will make all the type traits and other metaprogramming utilities work for __int24 and unsigned __int24

I'll file an upstream GCC bug to get those macros implicitly defined by the compiler, which is how it's supposed to work.

Edit: filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118329

[kevinpt]

This is the best solution. Thanks for digging into the problem.

[jwakely]

The GCC avr maintainer has fixed this for GCC 14.3:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118329