[Python-bugs-list] [ python-Bugs-742911 ] Memory fault on complex weakref/weakkeydict delete

[SourceForge.net]

Bugs item #742911, was opened at 2003-05-24 18:29
Message generated for change (Comment added) made by tim_one
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=742911&group_id=5470

Category: Python Interpreter Core
Group: Python 2.2.2
Status: Open
Resolution: None
Priority: 5
Submitted By: Mike C. Fletcher (mcfletch)
Assigned to: Nobody/Anonymous (nobody)
Summary: Memory fault on complex weakref/weakkeydict delete

Initial Comment:
Attached find two modules which together form a
test-case.  The cache.py file is ripped out of a
production system (OpenGLContext), and I am seeing
memory faults under both Python 2.2.2 and 2.2.3 when I
run the code.  Under 2.2.2 while single-stepping
through the code I was able to provoke an error-message:

Fatal Python error: GC object already in linked list

The error message doesn't show up under 2.2.3, but the
memory-fault does.

Modules here don't use any extension modules, so there
shouldn't be any loose memory references or the like. 
Note, you'll likely need to make weakkeydictionary's
__delitem__ use keys instead of iterkeys to even get to
the crashing.

----------------------------------------------------------------------

>Comment By: Tim Peters (tim_one)
Date: 2003-05-24 22:53

Message:
Logged In: YES 
user_id=31435

Outstanding, Neal -- thanks!  I can confirm that this crashes 
in a current 2.3 debug build on Windows too.  I'm feeling 
sick and won't pursue it now, though.  When cleaning up 
from the call to makeSome() (the body of makeSome() has 
completed, and we're cleaning up its execution frame, 
decref'ing the locals), we're dying in _Py_ForgetReference 
with a NULL-pointer derefernce.  The refcount on an Items 
object has just fallen to 0, and the code is trying to verify 
that the object is in the debug-build "list of all objects".  But 
its prev and next pointers are both NULL -- it's not in the 
list, and simply trying to check that it is blows up.

I don't have a theory for what's causing this, but it's 
probably not a good thing <heh>.

----------------------------------------------------------------------

Comment By: Neal Norwitz (nnorwitz)
Date: 2003-05-24 19:31

Message:
Logged In: YES 
user_id=33168

I cut out a lot of stuff from the test.  The new file is
much more minimal, but still crashes for me in a 2.3 debug
build.  You only need the one file too (not both files).

There is an issue with new style classes.  If Items doesn't
derive from object, I don't get a crash.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=742911&group_id=5470