René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Race condition in subinterpreters during subinterpreter creation on Windows debug build","articleBody":"Windows reproducer on main branch: run ``.\\PCbuild\\amd64\\python_d.exe -m test test__xxsubinterpreters -F`` and leave it running. If by run 100 nothing crashes, cancel and re-run again until you get a segfault.\r\n\r\nTraceback:\r\n```\r\nThread 0x000022ec (most recent call first):\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\test__xxsubinterpreters.py\", line 788 in setUp\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\case.py\", line 576 in _callSetUp\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\case.py\", line 619 in run\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\case.py\", line 678 in __call__\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 122 in run\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 84 in __call__\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 122 in run\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 84 in __call__\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 122 in run\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\unittest\\suite.py\", line 84 in __call__\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\support\\testresult.py\", line 140 in run\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\support\\__init__.py\", line 1099 in _run_suite\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\support\\__init__.py\", line 1225 in run_unittest\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\runtest.py\", line 281 in _test_module\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\runtest.py\", line 317 in _runtest_inner2\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\runtest.py\", line 360 in _runtest_inner\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\runtest.py\", line 235 in _runtest\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\runtest.py\", line 265 in runtest\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\main.py\", line 456 in run_tests_sequential\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\main.py\", line 574 in run_tests\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\main.py\", line 752 in _main\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\main.py\", line 711 in main\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\libregrtest\\main.py\", line 775 in main\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\test\\__main__.py\", line 2 in \u003cmodule\u003e\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\runpy.py\", line 88 in _run_code\r\n  File \"D:\\Ken\\Documents\\GitHub\\cpython\\Lib\\runpy.py\", line 198 in _run_module_as_main\r\n```\r\n\r\nSomehow after calling `__xxsubinterpreters.create`: in `drop_gil` the `\u0026ceval2-\u003egil_drop_request` is non-NULL but invalid, so when we try to `_Py_atomic_load` load from that address it segfaults.\r\n\r\nI have my suspicions it's due to swapping of GIL states but this is incredibly hard to debug because the race condition sometimes triggers when I run the test 4 times in a row, and sometimes doesn't trigger after 2 hours of continuous running.","author":{"url":"https://github.com/Fidget-Spinner","@type":"Person","name":"Fidget-Spinner"},"datePublished":"2023-01-03T14:09:28.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":11},"url":"https://github.com/100711/cpython/issues/100711"}