[Pyrex] Weakrefs patch

[Peter Johnson]

On Fri, 8 Dec 2006, Greg Ewing wrote:

> Peter Johnson wrote:
>> I recently ran into this problem myself (as the current code causes Python 
>> to crash), so I whipped up a quick patch that fixes it for me.  I think it 
>> follows all of the weakref guidelines now.
>
> Thanks, I'll see about incorporating this into the
> main version.
>
> One thing that bothers me a little is that if the
> __weakref__ slot can now be NULL, things will blow
> up if the user tries to reference it as a normal
> attribute, so I might have to put something in to
> prevent that. Probably a good idea anyway, since
> the user shouldn't be messing with it in the first
> place.
>
> Or should they? Are there any legitimate uses for
> Pyrex code to access the __weakref__ attribute?

I don't believe so, and hiding it is probably a good idea if doing so 
would cause blowup.

> And if not, perhaps it shouldn't have a user visible
> name at all? Would some other way of signalling
> weak-referenceability be better?

It looks like normal the normal Python way to do this is by doing:

   import weakref
   class foo(weakref.ref)

but I think this is ugly for Pyrex to try to handle.  Maybe instead derive 
the class from it as a keyword? e.g.

   cdef class foo(__weakref__)

?  But I've not yet looked to see how hard this would be to implement.

-Peter