Possible compiler bug in MSVC++

Question

I believe I've found a compiler bug in MSVC++ (present up to VS 2013). I want to check that it is indeed a bug before I report it.

The following code:

#include <map>

using std::map;

template <typename T>
class A
{
public:
    typedef T StoredType;
};

template <typename T>
map<typename T::StoredType, int> foo()
{
    map<typename T::StoredType, int> ret;
    return ret;
}  

template<>
map<char, int> foo<A<char>>()
{
    map<char, int> ret;
    return ret;
} // Error on this line

int main(int, char**)
{
    return 0;
}

Produces a compile error:

1>d:\documents\visual studio 2010\projects\proj\proj\source1.cpp(24): error C2785: 'std::map<T::StoredType,int> foo(void)' and 'std::map<_Kty,_Ty> foo(void)' have different return types
1>          with
1>          [
1>              _Kty=char,
1>              _Ty=int
1>          ]
1>          d:\documents\visual studio 2010\projects\proj\proj\source1.cpp(13) : see declaration of 'foo'
1>          d:\documents\visual studio 2010\projects\proj\proj\source1.cpp(20) : see declaration of 'foo'
1>d:\documents\visual studio 2010\projects\proj\proj\source1.cpp(24): error C2912: explicit specialization; 'std::map<_Kty,_Ty> foo<A<T>>(void)' is not a specialization of a function template
1>          with
1>          [
1>              _Kty=char,
1>              _Ty=int,
1>              T=char
1>          ]

However, it looks OK to me, and compiles fine on ideone.com. Is it a bug? Should it compile cleanly?

Out of curiosity, would it help if you made the primary template's body legal (i.e. changed the type of ret to map<typename T::StoredType, int>)? — Angew is no longer proud of SO
– Angew is no longer proud of SO, Commented Oct 30, 2014 at 10:04
@Angew same here, the code i wrong since map<T, int> is not map<typename T::StoredType, int>, even more curious if it compiles on GCC :D — Rudolfs Bundulis
– Rudolfs Bundulis, Commented Oct 30, 2014 at 10:05
Apologies - I was mucking about trying to find a minimal example and obviously didn't back up far enough. It still fails with a "correct" primary template body (as you'd expect - the template isn't instantiated, so it shouldn't ever be compiled). I'll edit the question. — Chowlett
– Chowlett, Commented Oct 30, 2014 at 10:09
It also appears I commented the wrong error line. Also fixed. — Chowlett
– Chowlett, Commented Oct 30, 2014 at 10:10

Rudolfs Bundulis · Accepted Answer · 2014-10-30 11:10:44Z

1

What is interesting that the same thing on a class specialization works fine:

#include <map>

using std::map;

template <typename T>
class A
{
public:
    typedef T StoredType;
};

template <typename T>
struct Z
{
    map<typename T::StoredType, int> foo()
    {
        map<T::StoredType, int> ret;
        return ret;
    }  // Error on this line
};

template<>
struct Z<A<char>>
{
    map<char, int> foo()
    {
        map<char, int> ret;
        return ret;
    }
};

int main(int, char**)
{
    return 0;
}

If the map is defined in the template then it also seems fine:

#include <map>

using std::map;

template <typename T>
class A
{
public:
    typedef map<T, int> MapType;
};

template <typename T>
typename T::MapType foo()
{
    T::MapType ret;
    return ret;
}

template<>
map<char, int> foo<A<char>>()
{
    map<char, int> ret;
    return ret;
}

int main(int, char**)
{
    return 0;
}

So the assumption about the bug seems to be possible.

edited Oct 30, 2014 at 11:10

answered Oct 30, 2014 at 10:42

Rudolfs Bundulis

12k7 gold badges41 silver badges81 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Lumen Over a year ago

Your answer would be easier to read if you just showed the parts that you’ve changed.

Collectives™ on Stack Overflow

Possible compiler bug in MSVC++

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related