[Bug c++/54755] New: Template instantiation confused by actual template argument definitions before and after template definition

[Bug c++/54755] New: Template instantiation confused by actual template argument definitions before and after template definition

[e678961 at yahoo dot co.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54755

             Bug #: 54755
           Summary: Template instantiation confused by actual template
                    argument definitions before and after template
                    definition
    Classification: Unclassified
           Product: gcc
           Version: 4.6.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: e678961@yahoo.co.uk


The following code:-

// ---------------------------------------------

#include <iostream>

// Before
void Function ( int  )  { std::cout << "int\n";   }
void Function ( long  ) { std::cout << "long\n";  }

template <class x>
struct TemplateClass
{
  x  Variable;
  void MemberFunction() { Function(Variable); }
};

// After
void Function ( float ) { std::cout << "float\n"; }

int main ( int, char** ) {
  TemplateClass<int>    I;
  I.MemberFunction();

  TemplateClass<long>   L;
  L.MemberFunction();

  TemplateClass<float>  F;
  F.MemberFunction();
}

// ---------------------------------------------

gives the compiler error:- error: call of overloaded 'Function(float&)' is
ambiguous

However, if all definitions of Function are in the "Before" block, and none in
"After" it works.

If all definitions are in the "After" block, and none in the "Before" code, it
also works.

If the int version of Function is in the "Before" code and the others in
"After" all instantiations of the template use the int version because of
implicit type conversion.

This looks like a bug in the symbol lookup. All overloaded versions of Function
can be found if they are either before or after the definition of the template
that happens to use them, but only if they are all defined together. Separate
the definitions before and after the template definition and it goes wrong.

Note: iostream output has only been used in the example to illustrate the bug
using std output. The code can be further simplified by removing iostream and
the same thing happens.

[Bug c++/54755] Template instantiation confused by actual template argument definitions before and after template definition

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54755

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-09-29 19:06:09 UTC ---
I haven't bothered analysing the code, but Comeau and Clang agree it's
ambiguous, so GCC is probably right.

[Bug c++/54755] Template instantiation confused by actual template argument definitions before and after template definition

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54755

Daniel Krügler <daniel.kruegler at googlemail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |daniel.kruegler at
                   |                            |googlemail dot com

--- Comment #2 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2012-09-29 23:07:59 UTC ---
(In reply to comment #1)
> I haven't bothered analysing the code, but Comeau and Clang agree it's
> ambiguous, so GCC is probably right.

I agree with Jonathan, the code is supposed to be ill-formed. This is so,
because the second stage of ADL is not allowed to consider overloads visible
*after* the point of the template definition where the call happens of function
argument types without associated namespaces.

[Bug c++/54755] Template instantiation confused by actual template argument definitions before and after template definition

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54755

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID

--- Comment #3 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-09-29 23:23:06 UTC ---
Closing.