STINNER Victor added the comment: > test.test_multiprocessing_spawn.WithProcessesTestPool.test_traceback
Oh, this test only "leaks" dangling processes and threads because PR 2841 was too strict. The problem is that multiprocessing.Pool has thread attributes which hold references to the pool: reference cycle. I don't see any easy way to break these cycles; since I don't understand if a poll is still supposed to be usable after terminate() or not. Anyway, my intent here was to do a first cleanup. So I modified my PR 2841 to call support.gc_collect() to break reference cycles. If someone wants to break reference cycles, please open a new issue. IMHO this issue already became enough complex :-) ---------- _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue26762> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com