Re: g++ 4: Unrecognised Friendship [REPOST/CROSS-POST]

[red floyd]

JH Trauntvein wrote:
> red floyd wrote:
>
> > JH Trauntvein wrote:
> >
> > > Consider the following example:
> > >
> > > namespace n1
> > > {
> > >   class cn1_base;
> > >
> > >
> > >   namespace n1_helpers
> > >   {
> > >      class helper1
> > >      {
> > >      private:
> > >         int private_member;
> > >         friend class cn1_base;
> > >      };
> > >   };
> > >
> > >
> > >   class cn1_base
> > >   {
> > >   public:
> > >      void foo()
> > >      {
> > >         n1_helpers::helper1 helper;
> > >         helper.private_member = 1;
> > >      }
> > >   };
> > > };
> > >
> > >
> > > While this compiled with earlier versions of G++, G++ version 4.x fails
> > > to compile this and gives the following message:
> > >
> > >  'int n1::n1_helpers::helper1::private_member' is prviate within this
> > >  context
> > >
> > > Fortunately, I can work around this by changing the friend declaration
> > > to the following:
> > >
> > >  friend class n1::n1_base;
> > >
> > > Is this a bug in the compiler or have I misunderstood something about
> > > friendship declarations?
> >
> > See Standard paragrah 7.3.1.2, subparagraph 3.  "If a friend declaration
> > in a non-local class first declares a class or function, the friend
> > class or function is a member of the innermost enclosing namespace."
> >
> > Sounds to me like it was a bug in earlier g++ versions.
>
>
> It seems to me that the key here is in the phrase, "first declares a
> class or function".  The sample code that I provided had a forward
> declaration of the class in the appropriate namespace that the
> friendship declaration should have used.

I am reposting this with a cross-post to comp.lang.c++, because this is 
not just a compiler issue, but a language issue.

I believe that in this case, the friend declaration is in fact the first 
declaration of n1_base, and that therefore 7.3.1.2/3 applies.