René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Various data races in subinterpreter tests under TSAN","articleBody":"# Bug report\n\nThread sanitizer reports a number of data races when running subinterpreter tests\n\n### `InterpreterPoolMixin` (`test_concurrent_futures`)\n\n- [ ] `run_eval_code_obj`: see https://github.com/python/cpython/issues/128130\n- [ ] [`posixmodule_exec`](https://github.com/python/cpython/blob/49bd47d5f14993d37b97aa2bbf257f5df16b96a9/Modules/posixmodule.c#L18014) modifies globals `waitid_result_desc` and `stat_result_desc`\n- [x] `_PyBuiltins_AddExceptions`: modifies globals `PyExc_EnvironmentError`, etc. https://github.com/python/cpython/blob/49bd47d5f14993d37b97aa2bbf257f5df16b96a9/Objects/exceptions.c#L4354-L4355\n- [ ] `_PyInterpreterState_FinalizeAllocatedBlocks`: modifies _PyRuntime variable _PyInterpreterState_FinalizeAllocatedBlocks\n- [ ] `_globals_fini` + [`clear_interpreter`](https://github.com/python/cpython/blob/49bd47d5f14993d37b97aa2bbf257f5df16b96a9/Modules/_interpqueuesmodule.c#L1454-L1456): data race on `_globals.module_count` (the read in `clear_interpreter` happens outside of a lock)\n- [ ] `init_static_exctypes`: modifies globals `_PyExc_InterpreterError`, etc.\n- [ ] `_structmodule_exec`: data race reported on [`ptr-\u003eunpack = native-\u003eunpack`](https://github.com/python/cpython/blob/49bd47d5f14993d37b97aa2bbf257f5df16b96a9/Modules/_struct.c#L2815)\n\nI think for now it would make sense to skip the `InterpreterPoolMixin` tests when running with TSAN.\n\n### `test__interpchannels`\n\n- [ ] [`find_name_in_mro`](https://github.com/python/cpython/blob/7b2e01bb555359913939d7ff168363f1760d3f8e/Objects/typeobject.c#L5415) and [`_waiting_release`](https://github.com/python/cpython/blob/7b2e01bb555359913939d7ff168363f1760d3f8e/Modules/_interpchannelsmodule.c#L497): I don't understand this one, but it happens during `test_send_recv_different_interpreters_and_threads`\n\n### `test__interpreters` and `test_interpreters`\n\n- [ ] `type_ready_set_new`: https://github.com/python/cpython/blob/7b2e01bb555359913939d7ff168363f1760d3f8e/Objects/typeobject.c#L8509\n- [ ] `type_ready`: (see https://github.com/python/cpython/issues/129817)\n- [ ] File descriptor race in `test_api.LowLevelTests.test_is_running` and `test_running`: `close()` concurrent with `read()` on the same file descriptor.\n- [ ] `managed_static_type_state_init` in `test.test_interpreters.test_stress.StressTests.test_create_many_threaded`: data race on the non-atomic read of `interp_count`.\n- [x] pegen [`memo_statistics`](https://github.com/python/cpython/blob/5fa7e1b7fd57e8c6297e9eb79d79cede42e5ce0f/Parser/pegen.c#L367) mutex is currently only used in free threading build (seen in test_create_many_threaded) \n- [x] version tags on static types: `NEXT_GLOBAL_VERSION_TAG`, `type-\u003etp_versions_used`, and `tp-\u003etp_version_tag` modifications are not thread-safe (seen in test_create_many_threaded)\n\n### `test_capi.test_misc.assign_version_tag`\n\n- [x]  race on `NEXT_GLOBAL_VERSION_TAG` in `assign_version_tag` (see above)\n\n\u003c!-- gh-linked-prs --\u003e\n### Linked PRs\n* gh-129826\n* gh-129829\n* gh-135794\n* gh-136266\n* gh-143446\n* gh-143530\n\u003c!-- /gh-linked-prs --\u003e\n","author":{"url":"https://github.com/colesbury","@type":"Person","name":"colesbury"},"datePublished":"2025-02-07T18:44:52.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":2},"url":"https://github.com/129824/cpython/issues/129824"}