gh-129967: Fix race condition in repr(set) (gh-129978)

The call to `PySequence_List()` could temporarily unlock and relock the
set, allowing the items to be cleared and return the incorrect
notation `{}` for a empty set (it should be `set()`).

Co-authored-by: T. Wouters <thomas@python.org>
This commit is contained in:
Sam Gross 2025-02-11 17:29:27 -05:00 committed by GitHub
parent 1f233f56d6
commit a7427f2db9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 54 additions and 2 deletions

View file

@ -0,0 +1,41 @@
import unittest
from threading import Thread, Barrier
from unittest import TestCase
from test.support import threading_helper
@threading_helper.requires_working_threading()
class TestSet(TestCase):
def test_repr_clear(self):
"""Test repr() of a set while another thread is calling clear()"""
NUM_ITERS = 10
NUM_REPR_THREADS = 10
barrier = Barrier(NUM_REPR_THREADS + 1)
s = {1, 2, 3, 4, 5, 6, 7, 8}
def clear_set():
barrier.wait()
s.clear()
def repr_set():
barrier.wait()
set_reprs.append(repr(s))
for _ in range(NUM_ITERS):
set_reprs = []
threads = [Thread(target=clear_set)]
for _ in range(NUM_REPR_THREADS):
threads.append(Thread(target=repr_set))
for t in threads:
t.start()
for t in threads:
t.join()
for set_repr in set_reprs:
self.assertIn(set_repr, ("set()", "{1, 2, 3, 4, 5, 6, 7, 8}"))
if __name__ == "__main__":
unittest.main()

View file

@ -0,0 +1,2 @@
Fix a race condition in the :term:`free threading` build when ``repr(set)``
is called concurrently with ``set.clear()``.

View file

@ -535,9 +535,18 @@ set_repr_lock_held(PySetObject *so)
return PyUnicode_FromFormat("%s()", Py_TYPE(so)->tp_name);
}
keys = PySequence_List((PyObject *)so);
if (keys == NULL)
// gh-129967: avoid PySequence_List because it might re-lock the object
// lock or the GIL and allow something to clear the set from underneath us.
keys = PyList_New(so->used);
if (keys == NULL) {
goto done;
}
Py_ssize_t pos = 0, idx = 0;
setentry *entry;
while (set_next(so, &pos, &entry)) {
PyList_SET_ITEM(keys, idx++, Py_NewRef(entry->key));
}
/* repr(keys)[1:-1] */
listrepr = PyObject_Repr(keys);