https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106085
Bug ID: 106085
Summary: It's unclear what the __has_trivial* builtins do after
P0848 (Conditionally Trivial Special Member Functions)
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: roi.jacobson1 at gmail dot com
Target Milestone: ---
I'm currently in the process of implementing P0848 for Clang, and I'm wondering
what to do about the __has_trivial_* builtins, especially since they consider
deleted functions as trivial, which goes against the spirit of 'deleted
functions are ineligible and should not affect the type traits'.
I tried to play with it a bit in GCC and the builtin seems pretty inconsistent:
#include <concepts>
template<class T>
struct A {
A() = delete;
A() requires std::same_as<T, int> = delete;
};
template<class T>
struct B {
B();
B() requires std::same_as<T, int> = delete;
};
template<class T>
struct C {
C();
C() requires std::same_as<T, int> = default;
};
static_assert(__has_trivial_constructor(A<int>));
static_assert(!__is_trivially_constructible(A<int>));
static_assert(!__has_trivial_constructor(B<int>));
static_assert(!__is_trivially_constructible(B<int>));
static_assert(__has_trivial_constructor(C<int>));
static_assert(__is_trivially_constructible(C<int>));
Is there logic behind this? Those builtins are technically deprecated in Clang
but I would like to support GCC compatibility if possible.
If not, it might be a good idea to add a warning about using those builtins or
potentially start to deprecate them in GCC as well.