Tests failing on Python 3.13

[JelleZijlstra]

Over on typing-extensions we run the typed-argument-parser tests daily on various Python versions. They started failing on Python 3.13 (python/typing_extensions#513):

https://github.com/python/typing_extensions/actions/runs/12172317648/job/33950775320

typed-argument-parser tests are failing on 3.13:

=========================== short test summary info ============================
FAILED tests/test_to_tap_class.py::test_subclasser_subparser[function-args_string_and_arg_to_expected_value7] - assert "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" in "usage: pytest [--foo] --arg_int ARG_INT [--arg_bool]\n              [--arg_list [ARG_LIST ...]] [-h]\n              {a,b} ...\npytest: error: argument {a,b}: invalid choice: 'X' (choose from a, b)\n"
 +  where "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" = str(SystemExit("error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')"))
FAILED tests/test_to_tap_class.py::test_subclasser_subparser[Class-args_string_and_arg_to_expected_value7] - assert "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" in "usage: pytest [--foo] --arg_int ARG_INT [--arg_bool]\n              [--arg_list [ARG_LIST ...]] [-h]\n              {a,b} ...\npytest: error: argument {a,b}: invalid choice: 'X' (choose from a, b)\n"
 +  where "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" = str(SystemExit("error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')"))
FAILED tests/test_to_tap_class.py::test_subclasser_subparser[_Args-args_string_and_arg_to_expected_value7] - assert "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" in "usage: pytest [--foo] --arg_int ARG_INT [--arg_bool]\n              [--arg_list [ARG_LIST ...]] [-h]\n              {a,b} ...\npytest: error: argument {a,b}: invalid choice: 'X' (choose from a, b)\n"
 +  where "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" = str(SystemExit("error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')"))
FAILED tests/test_to_tap_class.py::test_subclasser_subparser[class_or_function_3-args_string_and_arg_to_expected_value7] - assert "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" in "usage: pytest [--foo] --arg_int ARG_INT [--arg_bool]\n              [--arg_list [ARG_LIST ...]] [-h]\n              {a,b} ...\npytest: error: argument {a,b}: invalid choice: 'X' (choose from a, b)\n"
 +  where "error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')" = str(SystemExit("error: argument {a,b}: invalid choice: 'X' (choose from 'a', 'b')"))
======================== 4 failed, 301 passed in 1.58s =========================

https://github.com/python/typing_extensions/actions/runs/12172317648/job/33950775320

I suspect it's related to some of the argparse changes in Python 3.13.1. (Run git diff v3.13.0 v3.13.1 -- Lib/argparse.py in the CPython repo; a fair amount of code changed.) I haven't looked into this further.

[brianschubert]

This is likely a consequence of python/cpython#117766, which changed how the choices are displayed in the error message (switched from repr to str). That PR was backported to 3.13.1 and 3.12.8

[martinjm97]

Thank you @JelleZijlstra for raising this issue! We're happy to hear that Tap is being tested more broadly! We've accepted @brianschubert's fix, which seems to solve the string match problem for the error message.

Thanks again,
JK