$ ./python.exe 
Python 2.7a0 (trunk:67899, Dec 22 2008, 11:17:09) 
[GCC 4.0.1 (Apple Computer, Inc. build 5370)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
[36275 refs]
>>> sys._getframe(0).f_exc_traceback = 23
Python/ceval.c:2720: failed assertion `tstate->frame->f_exc_traceback ==
NULL'
Abort trap

Anybody willing and able to run with this?

As no response to msg109476 I'll close unless anyone objects.

Lack of response is not an appropriate reason to close.

Yup, still active:

$ ./python.exe 
Python 2.7.0+ (release27-maint:83247, Jul 30 2010, 00:41:50) 
[GCC 4.2.1 (Apple Inc. build 5659)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
[35067 refs]
>>> sys._getframe(0).f_exc_traceback = 23
Assertion failed: (tstate->frame->f_exc_traceback == NULL), function PyEval_EvalFrameEx, file Python/ceval.c, line 2998.
Abort trap

Not applicable to 3.x because the frame object does not have f_exc_traceback attribute anymore:

>>> import sys;sys._getframe(0).f_exc_traceback = 23
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'frame' object has no attribute 'f_exc_traceback'

The issue is present in 2.7 but appears to be rather benign.   It looks like the worst that can happen in non-debug build is a reference leak in set_exc_info().

I am not sure if this is worth fixing.  If this was affecting development version, I would argue that the simplest fix is to disallow setting f_exc_traceback or f_exc_value before f_exc_type is set.  In a bug-fix reliase, however this is probably not acceptable.

Given that sys._getframe() screams "beware, expert use only", I am not sure anything should be done other than possibly documenting that assigning to f_exc_* attributes of a transient frame object is not a good idea.

Note that the following works:

>>> f = sys._getframe(0)
>>> f.f_exc_traceback = 23

If there is a legitimate use for setting f_exc_traceback, the user should know to set f.f_exc_type and f.f_exc_value as well.

This issue looks like specific to 2.7, which is almost dead. Is there any interest about adding this to 2.7 branch?

2.7-only issue.