🚀 go-pugleaf

I found out that AT&T cfront 2.0 now complains if the return type is
specified for a type conversion operator:

    struct A
    {
      void * operator void * ();	// gives error message
    };

I am therefore making GNU C++ give a warning message about this, with
the intention of turning it into an error message later.  Such
redundancy should be avoided.

On the other hand, cfront also complains about multiple
initiaizations:

int f (int = 0);

int f (int a = 0)	// multiple initializations
{
  ...
}

I don't think this is such a hot idea, since the declaration of f and
its definition could be in different files, and it is nice to see
consistent facts at both sites.  I consider this form of redundancy
useful until all editors (including vi) implement hypertext.

Michael