Re: [Swig-user] change in swig behaviour with python ?

[Mark E. <ma...@mp...>]

On Fri, 2010-05-21 at 09:45 +0200, Haoyu Bai wrote:
> On Tue, May 11, 2010 at 2:00 AM, Mark Ellis <ma...@mp...> wrote:
> > Hi All
> >
> > I have encountered a problem which seems to have occurred between swig
> > 1.3.36 and 1.3.40, though by other accounts probably in 1.3.37.
> >
> > The opensync library generates a python wrapper using swig. The part of
> > the interface file causing the problem is the constructor of a wrapper
> > around a C struct.
> >
> > %extend OSyncChange {
> >        OSyncChange(PyObject *obj=NULL) {
> >                OSyncChange *change = NULL;
> >                if (obj)
> >                        change = (OSyncChange *)PyCObject_AsVoidPtr(obj);
> >                else
> >                        change = osync_change_new();
> >                return change;
> >        }
> >
> 
> Hi,
> 
> I would suggest to change the if(obj) to if(obj && obj!=Py_None). As
> Py_None is the Python equivalent of NULL, so the conversation of NULL
> -> None should be reasonable at here?
> 

Yep, that certainly works around the problem, many thanks.

I would suggest from a longer term swig perspective though that this
behaviour appears a bit broken.

Thanks again.
Mark