Closed ogrisel closed 1 year ago
test_loky linux-python-py39-high-memory
has a series of failures at teardown (unexpected loky workers) that in the end causes a deadlock during the main process shutdown.
I cancelled the CI run will trigger other runs from master
in a dedicated PR to see if we can reproduce.
This is very unlikely unrelated to this PR though as the first failure happens in tests/test_process_executor_loky.py
:
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_shutdown_with_sys_exit_at_pickle FAILED [ 36%]
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_shutdown_with_sys_exit_at_pickle ERROR [ 36%]
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_interpreter_shutdown ERROR [ 36%]
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_interpreter_shutdown ERROR [ 36%]
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_hang_issue12364 ERROR [ 37%]
tests/test_process_executor_loky.py::TestsProcessPoolLokyShutdown::test_hang_issue12364 ERROR [ 37%]
while the tests/test_reusable_executor.py
test file edited in this PR has not been reached by pytest yet.
Apparently there was a race condition I would always trigger on my local machine in this test utility.
Weirdly enough it would not be triggered by the CI...