Platform Windows 7.
hg id: a3f2dba93743+ (3.2)

I do not know any reproduction steps for this bug outside as a side-effect within an incomplete and problematic Stackless merge.  This bug also seems to exist in cpython, and is not in Stackless specific code.

It seems to be bad exception handling, but maybe I do not understand all cases where this code is used.

1. ast_error_finish is called.
2. It calls PyErr_Fetch(&type, &value, &tback);
3. It them proceeds to treat 'value' as a tuple which results in PyErr_BadInternalCall being called and then being used in place of the original exception.

Reproduction steps:

1. Sync http://hg.python.org/stackless.
2. Update to 3.2-slp branch.
3. Merge in changes from v3.2.1 to v3.2.2.
4. Compile and run regression tests.
5. Several exceptions will happen along the lines of bad argument where ast exceptions have happened and hit the problem above.