An example to produce this behavior:

>>> f = lambda: None
>>> for i in range(1000000):
...     f = f.__call__
... 
>>> f = None
Segmentation fault

Logged In: YES 
user_id=55188

Oh. my patch doesn't fix another scenario that using
recursive by two or more types of slots. So I remove.

Logged In: YES 
user_id=764593

CVS for 2.4 has comments for (and a fix for) problems similar 
to this.  Does the bug still exist with that source code?

Logged In: YES 
user_id=108556

Python CVS as of April 7th consistently segfaults with the
above example for me:

smulloni@bracknell src $ ~/apps/python-cvs/bin/python
Python 2.4a0 (#1, Apr  7 2004, 23:10:34) 
[GCC 3.3.2 20031218 (Gentoo Linux 3.3.2-r5,
propolice-3.3-7)] on linux2
Type "help", "copyright", "credits" or "license" for more
information.
>>> f=lambda: None
>>> for x in xrange(1000000):
...     f=f.__call__
... 
>>> f=None
Segmentation fault

Of course, maybe that's a good thing :).

Logged In: YES 
user_id=89016

Python CVS from 2004-06-02 seems to work:
Python 2.4a0 (#5, Jun  2 2004, 20:23:30) 
[GCC 2.96 20000731 (Red Hat Linux 7.3 2.96-113)] on linux2
Type "help", "copyright", "credits" or "license" for more 
information.
>>> f = lambda: None
>>> for x in xrange(1000000):
...  f = f.__call__
... 
>>> f
<method-wrapper object at 0x4259478c>
>>> f = None
>>> f
>>>

Logged In: YES 
user_id=1188172

CVS HEAD of 2006/01/10 still segfaults (here GCC 3.4.5).

Segfault occurs at wrapper_dealloc in descrobject.c.

Logged In: YES 
user_id=366566

Note that in doerwalter transcript, _ is still bound to the
method wrapper, preventing Python from attempting to free it.

Logged In: YES 
user_id=21627

This is now fixed in r47217 and r47218, by putting the
method-wrapper objects into the trashcan.