René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"TSan data races from test_queue with free threading","articleBody":"# Bug report\n\n### Bug description:\n\nIn a `--with-thread-sanitizer --disable-gil` build, running the `test_queue` test with high levels of `--parallel-threads` causes TSan to report several data races within the garbage collector.  All appear low-severity to me -- the worst they can do is cause the garbage collector to make bad decisions about whether to run a collection -- but I thought I'd report them anyway.\n\nStack traces have been truncated as how we got into the garbage collector isn't germane to the problem, and each individual conflict is only noted once.\n\n```\n$ ./python -E -Wd -m test --parallel-threads 16 test_queue\nUsing random seed: 3119004105\n0:00:00 load avg: 0.87 Run 1 test sequentially in a single process\n0:00:00 load avg: 0.87 [1/1] test_queue\n==================\nWARNING: ThreadSanitizer: data race (pid=373472)\n  Write of size 4 at 0x55ba3261501c by thread T469:\n    #0 gc_should_collect_rss Python/gc_free_threading.c:2072 (python+0x4e9921)\n    ...\n  Previous atomic read of size 4 at 0x55ba3261501c by thread T481:\n    #0 _Py_atomic_load_int_relaxed Include/cpython/pyatomic_gcc.h:307 (python+0x4e978b)\n    #1 gc_should_collect Python/gc_free_threading.c:2081 (python+0x4e978b)\n    ...\n\n  Atomic read of size 8 at 0x55ba32615100 by thread T481:\n    #0 _Py_atomic_load_ssize_relaxed Include/cpython/pyatomic_gcc.h:383 (python+0x4e9829)\n    #1 gc_should_collect_rss Python/gc_free_threading.c:2050 (python+0x4e9829)\n    ...\n  Previous write of size 8 at 0x55ba32615100 by thread T469:\n    #0 gc_should_collect_rss Python/gc_free_threading.c:2071 (python+0x4e9912)\n    ...\n\n  Write of size 4 at 0x55ba3261501c by thread T493:\n    #0 gc_should_collect_rss Python/gc_free_threading.c:2072 (python+0x4e9921)\n    ...\n  Previous atomic write of size 4 at 0x55ba3261501c by thread T509:\n    #0 _Py_atomic_add_int Include/cpython/pyatomic_gcc.h:15 (python+0x4eb18b)\n    #1 record_allocation Python/gc_free_threading.c:2111 (python+0x4eb18b)\n    ...\n\n  Atomic write of size 4 at 0x55ee8e6c301c by thread T115:\n    #0 _Py_atomic_add_int Include/cpython/pyatomic_gcc.h:15 (python+0x4eb18b)\n    #1 record_allocation Python/gc_free_threading.c:2111 (python+0x4eb18b)\n    ...\n  Previous read of size 4 at 0x55ee8e6c301c by thread T114:\n    #0 gc_should_collect_rss Python/gc_free_threading.c:2071 (python+0x4e9903)\n    ...\n\n  Write of size 8 at 0x558a95f2c0f8 by thread T982:\n    #0 gc_collect_internal Python/gc_free_threading.c:2242 (python+0x4f1227)\n    ...\n  Previous read of size 8 at 0x558a95f2c0f8 by thread T975:\n    #0 gc_should_collect_rss Python/gc_free_threading.c:2060 (python+0x4e9846)\n    ...\n```\n\nI can send a patch for these if it is felt to be worthwhile.\n\n### CPython versions tested on:\n\nCPython main branch\n\n### Operating systems tested on:\n\nLinux","author":{"url":"https://github.com/zackw","@type":"Person","name":"zackw"},"datePublished":"2025-05-05T21:24:55.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":2},"url":"https://github.com/133473/cpython/issues/133473"}