https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111608
--- Comment #4 from Julien Bernard <raplonu.jb at gmail dot com> ---
(In reply to Jonathan Wakely from comment #3)
I wasn't sure how to describe this issue, so my last sentence was probably
incorrect.
> I suspect this is covered by [temp.point] p7:
>
> "If two different points of instantiation give a template specialization
> different meanings according to the one-definition rule (6.3), the program
> is ill-formed, no diagnostic required."
If I take this example:
// 1. primary template
template<typename T>
struct X {
int f() { return 1; }
};
// 2. partial template specialization
template<typename T>
struct X<T*> {
int f() {return 2; }
};
// 3. full template specialization
template<>
struct X<int*> {
int f() { return 3; }
};
This program is well formed. In addition, the above 3. full template
specialization is equivalent to the following 4. implicit instantiation in the
sense they produce the same program (please, correct me if I'm wrong):
// 4. implicit instantiation
template<>
int X<int*>::f() { return 3; }
If I move 3. between 1. and 2., it still compiles but won't when 4. is between
1. and 2.
Here is a live demo with gcc, clang and msvc as a simple point of comparison.
https://godbolt.org/z/36aceTTb5
