Propagate pytest filterwarnings to subprocesses via PYTHONWARNINGS

[parthsanan]

Closes #1634

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[Copilot]

Pull request overview

This PR updates the test configuration to ensure pytest’s filterwarnings settings are propagated to Python subprocesses by exporting them via PYTHONWARNINGS, addressing warnings not being elevated to errors in subprocess-based tests (closes #1634).

Changes:

• Read filterwarnings from pytest config during pytest_configure.
• Set PYTHONWARNINGS (if not already set) to a comma-separated list of the configured warning filters so subprocesses inherit the same behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[parthsanan]

Hey @TomAugspurger, this is ready for review when you get a chance :)

[wence-]

/ok to test 58bba88

[wence-]

I think this makes sense, thanks.

[TomAugspurger]

/ok to test ca29091

[TomAugspurger]

/ok to test 7d15605

[TomAugspurger]

Thanks @parthsanan, and sorry for the delay. It looks like there's some CI failures: https://github.com/rapidsai/dask-cuda/actions/runs/25125997838/job/73639887965?pr=1638#step:12:5705

Details

FAILED tests/test_dask_cuda_worker.py::test_unknown_argument - assert b'Scheduler address: --my-argument' in b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n"
 +  where b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n" = CompletedProcess(args=['dask', 'cuda', 'worker', '--my-argument'], returncode=2, stdout=b'', stderr=b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n").stderr
FAILED tests/test_dask_cuda_worker.py::test_pre_import_not_found - assert b"ModuleNotFoundError: No module named 'my_module'" in b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n"
 +  where b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n" = CompletedProcess(args=['dask', 'cuda', 'worker', '127.0.0.1:9369', '--pre-import', 'my_module'], returncode=2, stdout=b'', stderr=b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n").stderr
FAILED tests/test_dask_cuda_worker.py::test_cuda_visible_devices_uuid - AssertionError: assert False
 +  where False = wait_workers(<Client: 'tcp://172.18.0.2:9359' processes=0 threads=0, memory=0 B>, n_gpus=1)
FAILED tests/test_dask_cuda_worker.py::test_rmm_track_allocations - AssertionError: assert False
 +  where False = wait_workers(<Client: 'tcp://172.18.0.2:9369' processes=0 threads=0, memory=0 B>, n_gpus=1)
 +    where 1 = get_n_gpus()
FAILED tests/test_dask_cuda_worker.py::test_get_cluster_configuration - AssertionError: assert False
 +  where False = wait_workers(<Client: 'tcp://172.18.0.2:9369' processes=0 threads=0, memory=0 B>, n_gpus=1)
 +    where 1 = get_n_gpus()
FAILED tests/test_dask_cuda_worker.py::test_worker_fraction_limits - AssertionError: assert False
 +  where False = wait_workers(<Client: 'tcp://172.18.0.2:9369' processes=0 threads=0, memory=0 B>, n_gpus=1)
 +    where 1 = get_n_gpus()
FAILED tests/test_dask_cuda_worker.py::test_worker_timeout - assert 'closing nanny at' in "/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: userwarning: while registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nusage: dask [options] command [args]...\ntry 'dask -h' for help.\n\nerror: no such command 'cuda'.\n"
 +  where "/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: userwarning: while registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nusage: dask [options] command [args]...\ntry 'dask -h' for help.\n\nerror: no such command 'cuda'.\n" = <built-in method lower of str object at 0x78763af62fa0>()
 +    where <built-in method lower of str object at 0x78763af62fa0> = "/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n".lower
 +      where "/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n" = CompletedProcess(args=['dask', 'cuda', 'worker', '192.168.1.100:7777', '--death-timeout', '1'], returncode=2, stdout='', stderr="/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n").stderr
FAILED tests/test_dask_cuda_worker.py::test_worker_cudf_spill_warning[True] - assert b'UserWarning: cuDF spilling is enabled' in b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n"
 +  where b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n" = CompletedProcess(args=['dask', 'cuda', 'worker', '127.0.0.1:9369', '--enable-cudf-spill', '--death-timeout', '1'], returncode=2, stdout=b'', stderr=b"/opt/conda/envs/test/lib/python3.11/site-packages/dask/__main__.py:7: UserWarning: While registering the command with name 'cuda', an exception occurred; the distributed.cli.utils module is deprecated.\n  run_cli()\nUsage: dask [OPTIONS] COMMAND [ARGS]...\nTry 'dask -h' for help.\n\nError: No such command 'cuda'.\n").stderr
FAILED tests/test_dask_setup.py::test_dask_cuda_worker_cli_integration[tcp] - AssertionError: assert False
 +  where False = wait_workers(<Client: 'tcp://172.18.0.2:57847' processes=0 threads=0, memory=0 B>, n_gpus=1)
FAILED tests/test_dask_setup.py::test_dask_cuda_worker_cli_integration[ucx] - AssertionError: assert False
 +  where False = wait_workers(<Client: 'ucx://172.18.0.2:34951' processes=0 threads=0, memory=0 B>, n_gpus=1)
FAILED tests/test_dask_setup.py::test_dask_cuda_worker_cli_integration[ucxx] - AssertionError: assert False
 +  where False = wait_workers(<Client: 'ucxx://172.18.0.2:33345' processes=0 threads=0, memory=0 B>, n_gpus=1)
= 21 failed, 1029 passed, 12 skipped, 3 xfailed, 16 warnings in 541.18s (0:09:01) =

At a glance, they look related to the changes in this PR. Are you interested in investigating those issues?

[parthsanan]

Hey @TomAugspurger, thanks for taking a look and I'll absolutely push a fix for the pipeline. Will ping once done :)