https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85054
Jonathan Wakely <redi at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2018-03-23
Ever confirmed|0 |1
--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Confirmed. The problem was in <functional>, which was included by <string>, and
if GCC had caught it I wouldn't have committed the bug.
G++ does diagnose it if you try to use it:
struct Y { using type = int; };
X<Y, void> x;
I was initially surprised that the X<T, void_t<typename T::type>> partial
specialization isn't more specialized than the X<T, void> one, but it's not.
Given that, GCC should diagnose it.
FWIW the reduced example isn't very realistic, the real case used a default
argument:
template<typename ... > using void_t = void;
template<typename, typename = void> struct X;
template<typename T> struct X<T, void_t<typename T::type>> {};
template<typename T> struct X<T, void> {};
I expected the second partial specialization to be more specialized than the
first, and for it to be valid, just like the fixed code:
template<typename ... > using void_t = void;
template<typename, typename = void> struct X {};
template<typename T> struct X<T, void_t<typename T::type>> {};
Now when T::type is not valid we use the primary template to generate the
specialization X<T, void>, but when T::type is valid the partial specialization
is more specialized than the primary template.
