[Pyrex] __del__ not called on some objects

[John (J5) Palmieri]

You can find my simple test case at
http://people.redhat.com/johnp/files/pyrex_bug.tar.gz

tar xvfzp pyrex_bug.tar.gz
cd pyrex_bug
make
python test.py

It will show two objects being created with only one calling the __del__
handler.  The failing object is declared as a cdef which is what causes
it to not call __del__ however simply deriving from it with a regular
class does call Test's __del__ method on object destruction.

It occurred to me that this might be intentional (however confusing)
since the cdef object it technically a c structure so I am not sure if
it is subject to the same life cycle as a python object though I don't
know why not since it acts just like one for the most part.

Anyway, there it is.

On Thu, 2005-08-25 at 20:37 -0700, Bob Ippolito wrote:
> On Aug 25, 2005, at 7:57 PM, John (J5) Palmieri wrote:
> 
> > I just found a bug during my attempts to debug some memory leaks in  
> > the
> > dbus bindings.  We use __del__ to unref some C objects when the  
> > wrapper
> > Pyrex object goes out of scope.  The problem was our Message object  
> > was
> > not calling __del__ and causing a huge memory leak.  On further
> > inspection I noticed that other Message objects were.  The difference
> > was the one calling __del__ were all subclasses of Message.   
> > Subclassing
> > Message solved the problem though it seems dubious why this would  
> > be the
> > case.  It might have something to do with the fact that Message  
> > exports
> > cdef methods while the derived classes only export the __init__
> > constructor.
> >
> > If you need any more information let me know.
> 
> A simple reproducible example that demonstrates the problem would  
> sure help.
> 
> -bob

-- 
John (J5) Palmieri <johnp at redhat.com>