Qualified friends

[Nathan Sidwell]

Martin v. Loewis wrote:
> Correct. I've updated the previous patch; you'll have to reverse that
> one before applying this one. I guess I make a test case for
> everything once you don't find new problems, anymore :-)
Yup, that's all ok now - hurrah!

I attach the autoconf test I'm now using to check these features, feel free to
incorporate it into a testcase (you'll see your ambiguous case in there
anyway).

Unpatched snapshot gives,
nathan@manao:1066>ss-g++ -c alien.C
alien.C:5: parse error before `('
alien.C: In function `void foo(class S *)':
alien.C:9: member `m' is a private member of class `S'
alien.C: At top level:
alien.C:28: declaration of `bink(L::U *)' not in a namespace surrounding `M'
alien.C:48: parse error before `)'
alien.C:49: parse error before `)'
alien.C:60: parse error before `('
alien.C:65: parse error before `('
alien.C:79: parse error before `('
alien.C:80: declaration of `bonk(K::T *)' not in a namespace surrounding `::'
alien.C: In function `void bink(class K::T *)':
alien.C:86: member `m' is a private member of class `K::T'
alien.C: In function `void bonk(class K::T *)':
alien.C:89: member `m' is a private member of class `K::T'

patched compiler gives no diagnostics
nathan@manao:1067>devel-g++ -c alien.C 

> We had that in the beginning, until the global namespace appeared in
> the assembler output and the back-end people started complaining that
> they don't want see namespace nodes anywhere. This was the change
thanks for the background info on why it was changed.

> Given the confusion produced last time, I doubt that such a patch is
> accepted. I personally wouldn't object, provided that the fix is
> 'perfect' - there is no reason to change one hack into another.
Ok, I'll decline to `clean' this up, people like you (Martin) and Jason are in
a much better position to understand the issues. IMHO, this hack is getting
uglier and uglier -- we're lying to the compiler and it'll bite us!

Once again, thanks for taking the time to get these alien friends working.

nathan

-- 
Dr Nathan Sidwell :: Computer Science Department :: Bristol University
      You can up the bandwidth, but you can't up the speed of light      
nathan@acm.org  http://www.cs.bris.ac.uk/~nathan/  nathan@cs.bris.ac.uk
// try a global friend
class S;
void foo(S *);
class S {
  friend void ::foo(S *);
  int m;
};
void foo(S *s) {
  s->m = 1;
}

// try a namespace friend
class T;
namespace N {void bar(T *);}
class T {
  friend void N::bar(T *);
  int m;
};
void N::bar(T *t) {
  t->m = 1;
}

// try a different namespace friend
namespace L {class U;}
namespace M {void bink(L::U *);}
namespace L {
class U {
  friend void M::bink(U *);
  int m;
};
}
void M::bink(L::U *u) {
  u->m = 1;
}

// try an ambiguous case
struct A {
  struct B {};
};
    
A::B C();
    
namespace B {
  A C();
}
        
class Test {
  friend A (::B::C)();  // Ok
  friend A::B (::C)();  // Ok
};

// global friend from a namespace
namespace K {
class S;
}
void baz(K::S *);

namespace K {
class S {
  friend void ::baz(S *);
  int m;
};
}

void ::baz(K::S *ptr) {
  ptr->m = 1;
}

// std friend from a namespace
namespace K {
class T;
}
namespace std {
void bink(K::T *);
void bonk(K::T *);
}
namespace K {
class T {
  friend void ::std::bink(T *);
  friend void std::bonk(T *);
  int m;
};
}

void std::bink(K::T *ptr) {
  ptr->m = 1;
}
void std::bonk(K::T *ptr) {
  ptr->m = 1;
}