Various data races in subinterpreter tests under TSAN
Open
Nobody has claimed this yet.
interpreter-core
topic-subinterpreters
type-bug
- Dominant language
- Python
- Stars
- 77.2k
- Forks
- 35.9k
- PR merge metrics
- PR metrics pending
Description
Bug report
Thread sanitizer reports a number of data races when running subinterpreter tests
InterpreterPoolMixin (test_concurrent_futures)
-
run_eval_code_obj: see https://github.com/python/cpython/issues/128130 -
posixmodule_execmodifies globalswaitid_result_descandstat_result_desc -
_PyBuiltins_AddExceptions: modifies globalsPyExc_EnvironmentError, etc. https://github.com/python/cpython/blob/49bd47d5f14993d37b97aa2bbf257f5df16b96a9/Objects/exceptions.c#L4354-L4355 -
_PyInterpreterState_FinalizeAllocatedBlocks: modifies _PyRuntime variable _PyInterpreterState_FinalizeAllocatedBlocks -
_globals_fini+clear_interpreter: data race on_globals.module_count(the read inclear_interpreterhappens outside of a lock) -
init_static_exctypes: modifies globals_PyExc_InterpreterError, etc. -
_structmodule_exec: data race reported onptr->unpack = native->unpack
I think for now it would make sense to skip the InterpreterPoolMixin tests when running with TSAN.
test__interpchannels
-
find_name_in_mroand_waiting_release: I don't understand this one, but it happens duringtest_send_recv_different_interpreters_and_threads
test__interpreters and test_interpreters
-
type_ready_set_new: https://github.com/python/cpython/blob/7b2e01bb555359913939d7ff168363f1760d3f8e/Objects/typeobject.c#L8509 -
type_ready: (see https://github.com/python/cpython/issues/129817) - File descriptor race in
test_api.LowLevelTests.test_is_runningandtest_running:close()concurrent withread()on the same file descriptor. -
managed_static_type_state_initintest.test_interpreters.test_stress.StressTests.test_create_many_threaded: data race on the non-atomic read ofinterp_count. - pegen
memo_statisticsmutex is currently only used in free threading build (seen in test_create_many_threaded) - version tags on static types:
NEXT_GLOBAL_VERSION_TAG,type->tp_versions_used, andtp->tp_version_tagmodifications are not thread-safe (seen in test_create_many_threaded)
test_capi.test_misc.assign_version_tag
- race on
NEXT_GLOBAL_VERSION_TAGinassign_version_tag(see above)
Linked PRs
- gh-129826
- gh-129829
- gh-135794
- gh-136266
- gh-143446
- gh-143530
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the unchecked races in test_concurrent_futures, test__interpchannels, and test__interpreters, then read the linked locations in Modules/posixmodule.c, Objects/typeobject.c, Modules/_interpchannelsmodule.c, and Modules/_struct.c. Run the affected subinterpreter tests under TSAN; done means each reported race has a reviewed resolution or documented disposition, with the linked PRs considered.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- c, python
- Domain
- backend, testing-qa
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100