OK, this time is for real. I have almost ready a library-only is_convertible
which purely uses SFINAE and noticed that apparently we have a remaining issue
with ambiguous bases, not fixed by your recent patch. The testcase goes like
this, I already checked that ICC likes it:
// #include <type_traits>
struct A { };
struct B
: A { };
struct C
: A { };
struct D
: B, C { };
template<bool, typename T = void> struct enable_if { typedef T type; };
template<typename T> struct enable_if<false, T> { };
template<typename From, typename To>
class mini_is_convertible
{
typedef char one;
typedef struct { char arr[2]; } two;
template<typename To1>
static void test_aux(To1);
template<typename To1, typename From1>
// static decltype(test_aux<To1>(std::declval<From1>()), one())
static typename
enable_if<(sizeof(test_aux<To1>(From1()), 1) > 0), one>::type
test(int);
template<typename, typename>
static two test(...);
public:
static const bool value = sizeof(test<To, From>(0)) == 1;
};
int main()
{
mini_is_convertible<D*, A*>::value; // compiles with 4.6 (false)
mini_is_convertible<D, A>::value; // does not compile (ok with ICC (false))
}
I would suggest double-checking the C++0x version too, using decltype and, more
correctly, declval.
--
Summary: SFINAE vs ambiguous base (the real one ;)
Product: gcc
Version: 4.6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: paolo dot carlini at oracle dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44907
