[C++-sig] Bug: Pyste and pure virtual functions.

[Prabhu Ramachandran]

>>>>> "N" == nicodemus  <nicodemus at globalite.com.br> writes:

    N> Hi Prabhu!  Prabhu Ramachandran wrote:

    >> I just found an obvious way to get around this -- simply call
    >> the wrapped function (afterall this is only to satisfy the
    >> compiler).  Attached is a patch to fix the problem with Pyste
    >> from CVS.  This works fine for me.

    N> I tried this approach too, but then it doesn't link: it
    N> complains about the abstract function being referenced inside
    N> the default_() method and is unable to resolve find it (I'm
    N> using Intel 6 on Windows). 8/ Anyone got other ideas?

Huh?  How is that possible?  Did you use my patch?  I specifically
chose to use method.name and not method.FullName() since method.name
would resolve to the wrapped function.  Here is an example:

struct A_Wrapper: A
{
   // [ snip ]

   // f is a pure virtual function.
   unsigned int f() { return call_method< unsigned int >(self, "f"); }

    unsigned int default_f() {
        PyErr_SetString(PyExc_RuntimeError, "pure virtual function called");
        throw_error_already_set();
        return A::f(); // <---------- WRONG.
        return f(); // <------------- CORRECT.
    }
};

As far as I can tell this is perfectly legal and the compiler cannot
complain since f is a valid overridden virtual function.  I've also
tested this with my code and it builds everything fine -- no link
problems and everything seems to work ok.

cheers,
prabhu