Nested class templates and incomplete type error

[Belebele]

The code below does not compile under g++.

There is an outer class template that contains and inner class
template. The outer class template is fully specialized for int. Then
a separate Foo class template inherits from the inner.

The compiler indicates that the inner class is incomplete when used as
the base class of the inner class template. Why?

-------------------------------------------------------------------
template <typename Xstruct Outer {

template <typename Ystruct Inner {};

};

template <struct Outer<int{ // If this full specialization is
removed, the compilation succeeds.

template <typename Y struct Inner {};

};

template <typename T>
struct Foo: Outer<T>::Inner<T{};

int main()
{
Foo<into;
return 0;
}
-------------------------------------------------------------------

Interestingly, if I remove the full specialization, the compilation
succeeds, to add to my confusion.

Thanks

[Victor Bazarov]

Belebele wrote:

The code below does not compile under g++.

There is an outer class template that contains and inner class
template. The outer class template is fully specialized for int. Then
a separate Foo class template inherits from the inner.

The compiler indicates that the inner class is incomplete when used as
the base class of the inner class template. Why?

-------------------------------------------------------------------
template <typename Xstruct Outer {

template <typename Ystruct Inner {};

};

template <struct Outer<int{ // If this full specialization is
removed, the compilation succeeds.

template <typename Y struct Inner {};

};

template <typename T>
struct Foo: Outer<T>::Inner<T{};

Try

template <typename T>
struct Foo: Outer<T>:: template Inner<T{};
// ^^^^^^^^^^

>
int main()
{
Foo<into;
return 0;
}
-------------------------------------------------------------------

Interestingly, if I remove the full specialization, the compilation
succeeds, to add to my confusion.

V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask

[Belebele]

Try

>
template <typename T>
struct Foo: Outer<T>:: template Inner<T{};
// ^^^^^^^^^^

Right on the money, as always. What's the issue here?

[Victor Bazarov]

Belebele wrote:

>Try

template <typename T>
struct Foo: Outer<T>:: template Inner<T{};
// ^^^^^^^^^^

Right on the money, as always. What's the issue here?

The compiler needs help determining that 'Inner' member of
'Outer<T>' is a template and that the meaning of the '<' after
it is the beginning of the template argument list and not the
comparison operator.

V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask

[Belebele]

The compiler needs help determining that 'Inner' member of
'Outer<T>' is a template and that the meaning of the '<' after
it is the beginning of the template argument list and not the
comparison operator.

Is there a legal construct that makes the use of the Inner class
template ambiguous if the "template" word is not used is this context?

[Victor Bazarov]

Belebele wrote:

>The compiler needs help determining that 'Inner' member of
'Outer<T>' is a template and that the meaning of the '<' after
it is the beginning of the template argument list and not the
comparison operator.

Is there a legal construct that makes the use of the Inner class
template ambiguous if the "template" word is not used is this context?

No, I don't believe there is. But that's the whole point. The C++
compiler cannot pick the meaning of the syntax based on its validity.

What if your 'Outer' specialisation for this particular 'T' would have
a member 'Inner' that is not a template, but instead, say, a static
constant of type 'size_t'? The expression 'Outer<T>::Inner < blah'
would be a valid expression (in some other context). The compiler
needs a nudge from you to help it interpret the _intention_. Don't
put the burden of guessing what your intention might have been onto
the poor compiler, it's already got its hands full.

But then again, I am not a compiler writer, nor am I formally educated
in computer language theory, so take whatever I say with a fistful of
salt.

V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask