Apologies for the imprecise summary - I have no idea
where the problem is here. 

Thanks to JP Calderone for this little horror.

(distilled down from his example)

bonanza% cat foo.py 
print type('F', (object,), {})().__class__.__module__
bonanza% python2.3 -c "execfile('foo.py', {})"
Traceback (most recent call last):
  File "<string>", line 1, in ?
  File "foo.py", line 1, in ?
    print type('F', (object,), {})().__class__.__module__
SystemError: error return without exception set
bonanza% python2.4 -c "execfile('foo.py', {})"
Traceback (most recent call last):
  File "<string>", line 1, in ?
  File "foo.py", line 1, in ?
    print type('F', (object,), {})().__class__.__module__
SystemError: error return without exception set

Logged In: YES 
user_id=29957

The attached patch fixes this to raise an AttributeError if
the object has no __module__. The other approach to fixing
it would be to make sure that the object created always gets
a __module__, but I have no idea in that case what a
reasonable fix would be.

Logged In: YES 
user_id=6656

Ah, I was about to attach the same test :-)

Do add a test and use PEP 7 code if you check it in...

Logged In: YES 
user_id=29957

Is it better to fix this here, or in the type() call to make
sure there's always a __module__ ?

Logged In: YES 
user_id=6656

I don't think there's an answer to that.

OTOH, I think it's more important that this gets fixed than that it 
gets fixed 100% perfectly.  IOW, do what you like, but please do 
something :-)

Logged In: YES 
user_id=29957

Fix checked in, will be in 2.4a1 and 2.3.5.
Objects/typeobject 2.259, 2.241.6.10
Misc/NEWS 1.999,  1.831.4.120