https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113074
Bug ID: 113074
Summary: requires requires should be SFINAE
Product: gcc
Version: 13.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: jdapena at igalia dot com
Target Milestone: ---
Created attachment 56904
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56904&action=edit
Test case from godbolt
Calling std::to_address in a requires requires function should be SFINAE.
Instead of it, it fails to compile.
This is breaking GCC build of Chromium in
https://chromium.googlesource.com/chromium/src.git/+/refs/tags/122.0.6193.2/mojo/public/cpp/bindings/type_converter.h#98
when the type does not support std::to_address because it fails to find
operator->.
Code is:
template <typename T, typename U>
requires requires(const U& obj) {
not std::is_pointer_v<U>;
{ mojo::ConvertTo<T>(std::to_address(obj)) } -> std::same_as<T>;
}
inline T ConvertTo(const U& obj) {
return mojo::ConvertTo<T>(std::to_address(obj));
}
The code is designed to only declare ConvertTo in case it can convert from
std::to_address result.
A possible test case is attached. This case fails in Clang with the expected
outcome (it fails to resolve to a valid call).
