Re: PR c++/11987: why is it invalid?

[Ian Lance Taylor <ian at airs dot com>]

Mark Mitchell <mark@codesourcery.com> writes:

> Ian Lance Taylor wrote:
> 
> > ==================================================
> > template <int dim> struct X { 
> >   struct I { I(); void foo(); }; 
> > }; 
> >  
> > template <int dim> struct Y : X<dim> { 
> >     typedef typename X<dim>::I I; 
> > }; 
> >  
> > template <int dim>
> > void Y<dim>::I::foo () {}
> > 
> > template struct Y<1>;
> > ==================================================
> 
> This should be invalid too, assuming I'm remembering correctly.  I think
> you have to say X<dim>::I::foo() when you make the definition.

(First of all, let me to say that I don't want to pick on Mark here,
he's just the one answering.)

OK, here is gcc/testsuite/g++.dg/eh/spec4.C:

==================================================
// PR c++/5104
// Test that a function with a throw spec is a valid template argument.

#include <exception>

typedef void (*HandlerFunction)();
typedef HandlerFunction (*SetHandlerFunction)(HandlerFunction);

template <SetHandlerFunction set_function>
class HandlerStack {
public:
  static void defaultHandler();
};

typedef HandlerStack<std::set_terminate> Terminate;

template<> void Terminate::defaultHandler() {}
==================================================

This test case uses a typedef when defining the function
defaultHandler.  Is this code legal?  If yes, then what is the
difference between this and the example above?

A similar yet different case appears in g++.dg/parse/ctor2.C:

==================================================
// PR c++/19244

typedef struct { void f(); } f;
void f::f() { }
==================================================

(These are, of course, cases I found with one version of my attempt to
patch this PR.)

Ian