https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80858
--- Comment #9 from Jonathan Wakely <redi at gcc dot gnu.org> ---
The problem isn't limited to implicitly-declared special members. If we adjust
comment 4 so that B::operator=(const B&) is user-declared but defined as
defaulted:
struct Empty {
};
template<class T>
struct A
{
A &operator=(const A&)
{
T t(3);
return *this;
}
};
class B
{
A<Empty> a;
public:
B& operator=(const B&) = default;
};
int main(void)
{
B b1, b2;
b1 = b2;
}
Then we get almost the same error. The only difference is that the "required
from here" line points to the defaulted operator= instead of to the class head,
i.e. we get
head.cc:19:12: required from here
instead of:
head.cc:14:7: required from here
Which is a little better, but it's still easy to miss that line among the other
diagnostics.
This would still be greatly improved by "required from the implicitly-defined
copy assignment operator" instead of "required from here".
tl;dr we should name the special member that we're generating a defaulted
definition for, instead of just saying "here".
That's what EDG does, see comment 5:
detected during:
instantiation of "A<T> &A<T>::operator=(const A<T> &) [with
T=Empty]" at line 22
implicit generation of "B &B::operator=(const B &)" at line 22
