Thank you Alexander. Yes, there is still an issue for large operands, and the attached patch does fix it. Floats are explicitly checked for and rejected by PyArg_ParseTuple for the "l" format (as called by builtin_range) so to preserve this behavior we can explicitly check in the argument parsing of handle_range_longs as well.

This all goes away in Py3 due to the unification of int and long. (And I agree that using __index__ rather than __int__ fits better there).