Overriding std::is_arithmetic<> for program-defined type is undefined behaviour?

[saxbophone]

In the README, it's specified:

For modifier strong::arithmetic, the type trait std::is_arithmetic<> is true.

According to the standard, this is undefined behaviour:

• None of the templates defined in <type_traits> may be specialized for a program-defined type, except for std::common_type and std::basic_common_reference (since C++20). This includes the type traits and the class template std::integral_constant.

(source — cppreference.com)

Maybe std::numeric_limits<> should be specialised instead, to provide type-traits? That one is open to extension, provided one doesn't reopen the std namespace and just specialises the fully-qualified name of the template...

[rollbear]

Right. Bummer. I should've known this. Dropping it is also a potentially breaking change for some users, but it's the way to go. Thanks for reporting.

[timblechmann]

related to specializing std::is_arithmetic, this code gives me a compile error with catch2 v3

TEST_CASE( "testme", "" )
{
    using test_type
        = strong::type< std::int8_t, struct tag, strong::arithmetic, strong::default_constructible, strong::equality >;

    constexpr test_type a;
    test_type           b;
    CHECK( a == b );
}

my/test.cpp: error: use of overloaded operator '<=' is ambiguous (with operand types 'Catch::Decomposer' and 'const test_type' (aka 'const type<signed char, tag, strong::arithmetic, strong::default_constructible, strong::equality>'))
    catchAssertionHandler.handleExpr( Catch::Decomposer() <= a == b );
                                      ~~~~~~~~~~~~~~~~~~~ ^  ~
/path/to/catch2-v3/src/catch2/../catch2/internal/catch_decomposer.hpp:247:21: note: candidate function [with T = const strong::type<signed char, tag, strong::arithmetic, strong::default_constructible, strong::equality> &, $1 = 0]
        friend auto operator <= ( Decomposer &&, T && lhs ) -> ExprLhs<T const&> {
                    ^
/path/to/catch2-v3/src/catch2/../catch2/internal/catch_decomposer.hpp:252:21: note: candidate function [with T = strong::type<signed char, tag, strong::arithmetic, strong::default_constructible, strong::equality>, $1 = 0]
        friend auto operator <= ( Decomposer &&, T value ) -> ExprLhs<T> {
                    ^
1 error generated.

[rollbear]

Thanks @timblechmann , that was a very unexpected result of this. I'll get to fixing it as soon as I'm out of meetings.

[rollbear]

Fixed on main now. Leaving this open until a release is tagged.

[saxbophone]

Hurrah! How do you feel about going down the route of specializing std::numeric_limits<> instead? IIRC, there are members within that struct that can provide similar traits-like functionality as you were providing with is_arithmetic, though I forget the names of the exact members. I've got a previous specialisation of that struct I did for a personal project, maybe I'll dig that out (there's loads of definitions ☹️) and have a go hacking on strong_type...

[rollbear]

@saxbophone feel free to open a separate issue about that.

[rollbear]

Fix included in release v5

[saxbophone]

@saxbophone feel free to open a separate issue about that.

Thanks, I think I will and I will also try and send a PR with the structure of the type traits stuff I mentioned.