The ctypes ArgumentError exception indicates the location of the problem by argument number. It counts arguments starting from 1, not 0 as is typical in Python.

Observed
--------
An example (anonymised) traceback:
    Traceback (most recent call last):
        ...
        foreign_function(a, b, c, d)
    ctypes.ArgumentError: argument 2: <class 'TypeError'>: wrong type

The error here was with the argument "b".

Expected
--------
Standard, zero-indexed Python counting would suggest that argument 2 should mean "c", and "b" would be argument 1.

Rationale
---------
This may be as intended, but for me it violates the principle of least surprise.

I *think* this is the vicinity of the bug:
http://hg.python.org/cpython/file/696d3631a4a1/Modules/_ctypes/callproc.c#l1103

    _ctypes_extend_error(PyExc_ArgError, "argument %d: ", i+1);

If I'm right and the "i+1" (here and/or in subsequent lines) is the cause, that definitely makes it look intentional.

IMO we always use 1-based numbering when counting things.

Try e.g.

>>> format('a', None)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: format expects arg 2 to be string or unicode, not NoneType