http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47125
--- Comment #7 from Jason Merrill <jason at gcc dot gnu.org> 2011-03-30 02:27:45 UTC --- (In reply to comment #6) > I think error message like followings is better. > ---- > testcase.cpp:11:1: error: need 'typename' before 'enable_if<true, T>::type' > because 'enable_if<true, T>' is a dependent scope You don't need 'typename' there because you have 'struct'. 14.6/5: "A qualified name used as the name in a mem-initializer-id, a base-specifier, or an elaborated-type-specifier is implicitly assumed to name a type, without the use of the typename keyword." But 'int' is not a struct, so substituting 'int' into struct enable_if<true,T>::type fails. So a better error message would be testcase.cpp:12:8: note: template<class T> class enable_if<true, T>::type f(T) testcase.cpp:12:8: note: 'class enable_if<true,int>::type' resolves to 'int', which is not a class type currently we don't attempt to explain SFINAE when printing candidates, but that would be a useful thing to do.
