René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Unpickler BUILD can traverse a concurrently mutated shared state dictionary","articleBody":"**Description**\n\nThe ordinary-state and slot-state loops in `_pickle.load_build()` traverse their state dictionary with `PyDict_Next()` without a critical section or owned snapshot. A reduction can supply an existing dictionary shared with other threads, so key-set changes can invalidate traversal and replacements can reclaim borrowed keys or values before assignment.\n\n**Observed Behavior**\n\nOn a free-threaded build, ordinary-state mode terminates with SIGSEGV in about 0.23 seconds and slot-state mode in about 0.15 seconds. Both compatibility-GIL controls complete normally.\n\n**Affected Version**\n\nCPython 3.14.7 at commit `823f0323ee6ec1402088b73bce1a38473cac36dc`, using the free-threaded build with `PYTHON_GIL=0`.\n\n**Reproduction**\n\nRun both modes:\n\n```sh\nPYTHON_GIL=0 python3.14t poc/reproduce.py 10 state\nPYTHON_GIL=0 python3.14t poc/reproduce.py 10 slot\n```\n\n**PoC Source Code**\n\n`poc/reproduce.py`:\n\n```python\n#!/usr/bin/env python3\n\"\"\"Race Unpickler BUILD iteration with mutation of its shared state dict.\"\"\"\n\nimport pickle\nimport sys\nimport threading\nimport time\n\n\nMODE = sys.argv[2] if len(sys.argv) > 2 else \"state\"\nif MODE not in {\"state\", \"slot\"}:\n    raise ValueError(\"mode must be 'state' or 'slot'\")\n\nSHARED = {f\"field_{i}\": object() for i in range(4096)}\nKEYS = tuple(SHARED)\nSTOP = threading.Event()\nFAILURES = []\n\n\ndef get_shared():\n    return SHARED\n\n\nclass SharedState(dict):\n    def __reduce__(self):\n        return get_shared, ()\n\n\nclass Victim:\n    def __reduce__(self):\n        state = SharedState()\n        if MODE == \"slot\":\n            state = (None, state)\n        return Victim, (), state\n\n\nPAYLOAD = pickle.dumps(Victim(), protocol=2)\n\n\ndef mutate():\n    while not STOP.is_set():\n        SHARED.clear()\n        SHARED.update({key: object() for key in KEYS})\n\n\ndef load_worker(deadline):\n    count = 0\n    try:\n        while time.monotonic() \u003c deadline and not STOP.is_set():\n            obj = pickle.loads(PAYLOAD)\n            count += 1\n    except BaseException as exc:\n        FAILURES.append((type(exc).__name__, repr(exc), count))\n        STOP.set()\n\n\ndef main():\n    seconds = float(sys.argv[1]) if len(sys.argv) > 1 else 10.0\n    deadline = time.monotonic() + seconds\n    threads = [threading.Thread(target=mutate) for _ in range(2)]\n    threads += [threading.Thread(target=load_worker, args=(deadline,)) for _ in range(8)]\n    for thread in threads:\n        thread.start()\n    for thread in threads[2:]:\n        thread.join()\n    STOP.set()\n    for thread in threads[:2]:\n        thread.join()\n    print(\"mode=\", MODE, \"gil=\", sys._is_gil_enabled(), \"failures=\", FAILURES)\n    return bool(FAILURES)\n\n\nraise SystemExit(main())\n```\n","author":{"url":"https://github.com/Nievesjyl","@type":"Person","name":"Nievesjyl"},"datePublished":"2026-09-07T18:48:53Z","url":"https://github.com/python/cpython/issues/157124"}