http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58972
Daniel Krügler <daniel.kruegler at googlemail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |daniel.kruegler@googlemail.
| |com
--- Comment #1 from Daniel Krügler <daniel.kruegler at googlemail dot com> ---
The behaviour looks like a gcc defect to me, not as a language problem. Note
that the similar code using a local class type causes the same kind of problem
for gcc:
//-------------------
class base
{
protected:
typedef int type;
};
template <typename T>
class derive: public base
{
using typename base::type;
public:
void foo(type a)
{
struct f_t {
f_t(type a) : a(a) {}
bool operator()(type x) const { return x == a; }
type a;
} f(a);
f(32);
}
};
int main()
{
derive<char> d;
d.foo(23);
}
//-------------------
It seems to be a regression, because gcc 4.4.6 and earlier accepted the code.
According to 9.8 [class.local] p1 the wording says that it should be
well-formed:
"The local class is in the scope of the enclosing scope, and has the same
access to names outside the function as does the enclosing function."