Closed musicinmybrain closed 1 year ago
Thank you, I'm aware of this issue.The culprit is eventlet monkey patching, which does not play well with multiprocessing. I did not manage to figure out how to isolate the test cases, nor how to isolate eventlet. So I physically separated the tests into two directories and I run them now as
pytest tests/
pytest tests_eventlet/
Is this something you could do as well? The only alternative that comes to my mind is to try run the eventlet tests in a child process, so that the monkey patch would not affect the main one.
Hang on, I think running in a child works fine. Give me a sec :running_man:
Thank you, I'm aware of this issue.The culprit is eventlet monkey patching, which does not play well with multiprocessing. I did not manage to figure out how to isolate the test cases, nor how to isolate eventlet. So I physically separated the tests into two directories and I run them now as
pytest tests/ pytest tests_eventlet/
Thank you. This works for me, too.
I fixed it in https://github.com/harlowja/fasteners/commit/80a3eaed75276faf21034e7e6c626fd19485ea39 on my side as well - eventlet test is back in main
, and I run it in a subprocess, so there is no more conflict.
I fixed it in 80a3eae on my side as well - eventlet test is back in
main
, and I run it in a subprocess, so there is no more conflict.
Thanks again; I can confirm that commit works for me, too.
On Fedora Linux 36, x86_64,
test_no_concurrent_writers
always hangs beginning with 383d9b6. It never hangs with 59dec43 and earlier.Beginning with 1875874,
test_double_reader_writer
also hangs: