[BUG] parallelization on python 3.13, windows, fails with `no module named _posixsubprocess`

[fkiraly]

New failures on windows 3.13 with no module named _posixsubprocess - this seems like it is a new failure?

Same failure occurs on sktime in tests involving parallelization backends (for benchmarking), see sktime/sktime#8722

FYI @jgyasu

Failing run link: https://github.com/SimonBlanke/Hyperactive/actions/runs/17117216642/job/48550557909

 =========================== short test summary info ===========================
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-3] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-4] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-5] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-6] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-7] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[GridSearchSk-8] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-3] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-4] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-5] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-6] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-7] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/tests/test_all_objects.py::TestAllOptimizers::test_opt_run[RandomSearchSk-8] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture1] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture2] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture3] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture4] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture5] - ModuleNotFoundError: No module named '_posixsubprocess'
FAILED src/hyperactive/utils/tests/test_parallel.py::test_parallelize_simple_loop[fixture6] - ModuleNotFoundError: No module named '_posixsubprocess'
=========== 18 failed, 2012 passed, 1 skipped in 251.89s (0:04:11) ============

[fkiraly]

The fixture name improvement in #170 indicates that these are two joblib backends: loky and multiprocessing
https://github.com/SimonBlanke/Hyperactive/actions/runs/17149614844/job/48652651607?pr=170

[fkiraly]

This is a bug in cypython, already reported here: python/cpython#138031

[fkiraly]

A fix should be available since yesterday, so we should revert the skips?
python/cpython#138031

[fkiraly]

FYI @SimonBlanke

[ogrisel]

FYI: a fix for this issue has been released yesterday in loky 3.5.6 and joblib 1.5.2 (that vendors loky 3.5.6), see joblib/loky#459 and joblib/loky#461 for more details.

[SimonBlanke]

A fix should be available since yesterday, so we should revert the skips? python/cpython#138031

Done. It works without an error now. Should we just close this PR?

[fkiraly]

yes, though we also need to revert the skip of python 3.13 windows in the CI?

[SimonBlanke]

yes, though we also need to revert the skip of python 3.13 windows in the CI?

Done