search

python / cpython

The Python programming language
https://www.python.org
Other
62.87k stars 30.12k forks source link

test_interpreters Failing Intermittently Under Py_DISABLE_GIL #118731

Open ericsnowcurrently opened 5 months ago

ericsnowcurrently commented 5 months ago

Bug report

Bug description:

On one of my recent PRs, gh-118157, I was getting consistent failures on CI in test_interpreters. Different tests would fail, but the error was always the same:

(expand) ``` 0:08:24 load avg: 4.18 Re-running 1 failed tests in verbose mode in subprocesses 0:08:24 load avg: 4.18 Run 1 test in parallel using 1 worker process (timeout: 10 min, worker timeout: 15 min) 0:08:26 load avg: 3.93 [1/1/1] test_interpreters failed (6 failures) Re-running test_interpreters in verbose mode (matching: test_display_preserved_exception, test_gh_109793, test_sys_path_0, test_sys_path_0, test_sys_path_0, test_sys_path_0) test_display_preserved_exception (test.test_interpreters.test_api.TestInterpreterExec.test_display_preserved_exception) ... FAIL test_gh_109793 (test.test_interpreters.test_lifecycle.FinalizationTests.test_gh_109793) ... FAIL test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) ... ### start subtest debug ### # python3 script.py failed: --- stdout --- { "main": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4", "sub": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4" } ------ --- stderr --- Exception ignored in: Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__ File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref TypeError: catching classes that do not inherit from BaseException is not allowed Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters Python runtime state: finalizing (tstate=0x00005639aed8ea10) ------ ### end subtest debug ### test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 script.py] ... FAIL ### start subtest debug ### # python3 -m script failed: --- stdout --- { "main": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4", "sub": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4" } ------ --- stderr --- Exception ignored in: Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__ File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref TypeError: catching classes that do not inherit from BaseException is not allowed Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters Python runtime state: finalizing (tstate=0x0000564591b58a10) ------ ### end subtest debug ### test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m script] ... FAIL ### start subtest debug ### # python3 -m pkg failed: --- stdout --- { "main": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4", "sub": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4" } ------ --- stderr --- Exception ignored in: Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__ File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref TypeError: catching classes that do not inherit from BaseException is not allowed Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters Python runtime state: finalizing (tstate=0x000055aaa11eaa10) ------ ### end subtest debug ### test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m pkg] ... FAIL ### start subtest debug ### # python3 -m pkg.script failed: --- stdout --- { "main": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4", "sub": "/tmp/test_python_tk7ha1xh/test_interpreters_76l2y2f4" } ------ --- stderr --- Exception ignored in: Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__ File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref TypeError: catching classes that do not inherit from BaseException is not allowed Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters Python runtime state: finalizing (tstate=0x00005637ca7cfa10) ------ ### end subtest debug ### test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m pkg.script] ... FAIL ====================================================================== FAIL: test_display_preserved_exception (test.test_interpreters.test_api.TestInterpreterExec.test_display_preserved_exception) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/__init__.py", line 2603, in wrapper return func(*args, **kwargs) File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_api.py", line 764, in test_display_preserved_exception self.assertEqual(stderr, dedent(f"""\ ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^ Traceback (most recent call last): ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ...<18 lines>... RuntimeError: uh-oh! ^^^^^^^^^^^^^^^^^^^^ """)) ^^^^^ AssertionError: 'Trac[778 chars]oh!\nException ignored in: interp.exec(script) ~~~~~~~~~~~^^^^^^^^ File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 227, in exec raise ExecutionFailed(excinfo) test.support.interpreters.ExecutionFailed: RuntimeError: uh-oh! Uncaught in the interpreter: Traceback (most recent call last): File "/tmp/test_python_tk7ha1xh/tmpkqxiakqz/script.py", line 6, in script spam.eggs() ~~~~~~~~~^^ File "/tmp/test_python_tk7ha1xh/tmpkqxiakqz/spam.py", line 6, in eggs ham() ~~~^^ File "/tmp/test_python_tk7ha1xh/tmpkqxiakqz/spam.py", line 3, in ham raise RuntimeError('uh-oh!') RuntimeError: uh-oh! - Exception ignored in: - Traceback (most recent call last): - File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__ - File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref - TypeError: catching classes that do not inherit from BaseException is not allowed - Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters - Python runtime state: finalizing (tstate=0x000055b06f0a1a10) - ====================================================================== FAIL: test_gh_10[979](https://github.com/python/cpython/actions/runs/8978599353/job/24660038142#step:19:980)3 (test.test_interpreters.test_lifecycle.FinalizationTests.test_gh_109793) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 185, in test_gh_109793 self.assertEqual(proc.returncode, 1) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ AssertionError: -6 != 1 ====================================================================== FAIL: test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 script.py] ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 29, in subTest yield ctx File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 157, in test_sys_path_0 out = self.run_python(argv, cwd=cwd) File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 114, in run_python self.assertEqual(proc.returncode, 0) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ AssertionError: -6 != 0 ====================================================================== FAIL: test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m script] ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 29, in subTest yield ctx File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 157, in test_sys_path_0 out = self.run_python(argv, cwd=cwd) File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 114, in run_python self.assertEqual(proc.returncode, 0) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ AssertionError: -6 != 0 ====================================================================== FAIL: test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m pkg] ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 29, in subTest yield ctx File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 157, in test_sys_path_0 out = self.run_python(argv, cwd=cwd) File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 114, in run_python self.assertEqual(proc.returncode, 0) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ AssertionError: -6 != 0 ====================================================================== FAIL: test_sys_path_0 (test.test_interpreters.test_lifecycle.StartupTests.test_sys_path_0) [python3 -m pkg.script] ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 29, in subTest yield ctx File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 157, in test_sys_path_0 out = self.run_python(argv, cwd=cwd) File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_interpreters/test_lifecycle.py", line 114, in run_python self.assertEqual(proc.returncode, 0) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ AssertionError: -6 != 0 ---------------------------------------------------------------------- Ran 3 tests in 1.720s FAILED (failures=6) test test_interpreters failed 1 test failed again: test_interpreters ```

The error looks like this:

Exception ignored in: <function Interpreter.__del__ at 0x2000117c0d0>
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 157, in __del__
  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/support/interpreters/__init__.py", line 173, in _decref
TypeError: catching classes that do not inherit from BaseException is not allowed
Fatal Python error: PyInterpreterState_Delete: remaining subinterpreters
Python runtime state: finalizing (tstate=0x0000564591b58a10)

I'm fairly sure that means one of the 3 calls in _interpreters._decref() is raising a bogus exception.

It isn't clear to me how gh-118157 would have caused this to happen. As a very temporary solution, I've skipped test_interpreters on free-threading builds.

CC @colesbury

CPython versions tested on:

CPython main branch

Operating systems tested on:

No response

colesbury commented 5 months ago

I think what's happening is:

By the time Interpreter.__del__ is called, a lot of global variables including _interpreters have already been deleted from the module's dictionary so you get a bunch of weird errors, and the interpreter is not deleted.

ericsnowcurrently commented 5 months ago

That makes sense. You have any thoughts already on how to resolve this?

colesbury commented 5 months ago

I'm not sure. We could call the finalizers on all GC-enabled objects during shutdown in the free-threaded build (before we start clearing module's dictionaries). I think that makes logical sense: everything is going away, why not call their finalizers? That seems to fix test_interpreters, but a few other test cases would need tweaking.