gh-144356: Make set iterator `__length_hint__` and `iternext` race-safe under no-gil #144357

hyongtao-code · 2026-01-31T03:34:26Z

Long log:

In free-threaded builds, setiter_len() and setiter_iternext() could race with concurrent set mutation and iterator exhaustion.

We now serialize operations on the iterator itself (and keep the set stable) using a critical section, and ensure the iterator’s exhaustion state is handled consistently.

The non-free-threaded build keeps the existing semantics. Add free-threading-style concurrency tests (similar to the itertools FT tests) to exercise __length_hint__ and iternext exhaustion under concurrent execution.

setiter_len() was reading so->used without atomic access while concurrent
mutations update it atomically under Py_GIL_DISABLED.

In free-threaded builds, setiter_len() could race with concurrent set
mutation and iterator exhaustion.

Use an atomic load for so->used to avoid a data race. This preserves the
existing semantics of __length_hint__ while making the access thread-safe.

Signed-off-by: Yongtao Huang yongtaoh2022@gmail.com

Issue: Data race in set iterator length_hint under no-gil #144356

setiter_len() was reading so->used without atomic access while concurrent mutations update it atomically under Py_GIL_DISABLED. Use an atomic load for so->used to avoid a data race. This preserves the existing semantics of __length_hint__ while making the access thread-safe. Signed-off-by: Yongtao Huang <yongtaoh2022@gmail.com>

Lib/test/test_free_threading/test_set.py

Objects/setobject.c

hyongtao-code · 2026-02-01T16:21:39Z

Thanks for the review. I’ve decided to address both issues in this PR. I also added a corresponding test case for the issue you pointed out.

Objects/setobject.c

eendebakpt

I left some more review comments. I think we can get this right, but a simpler approach here would be to put a critical section on the set iterator itself.

Lib/test/test_free_threading/test_set.py

Objects/setobject.c

hyongtao-code · 2026-02-08T11:56:22Z

put a critical section on the set iterator

Thanks for the suggestion. I simplified the implementation by protecting both the iterator and the set with Py_BEGIN_CRITICAL_SECTION2(self, so), so concurrent use of the same iterator is properly serialized. This keeps the logic aligned with other iterators and avoids the races.

It looks like there is an unrelated flaky test case.

eendebakpt

@hyongtao-code Thanks for your work on this! The core idea of the PR has shifted a bit, which is why I have some more review comments, but we are getting closer I believe.

Lib/test/test_free_threading/test_set.py

Objects/setobject.c

Co-authored-by: Pieter Eendebak <pieter.eendebak@gmail.com>

This reverts commit 78241a8.

Lib/test/test_free_threading/test_set.py

Objects/setobject.c

Lib/test/test_free_threading/test_set.py

eendebakpt · 2026-02-10T19:40:17Z

Objects/setobject.c

+#ifdef Py_GIL_DISABLED
+    PySetObject *so = si->si_set;
+    assert(so != NULL);
+
+    Py_BEGIN_CRITICAL_SECTION2(op, so);
+    if (si->si_pos >= 0 && si->si_used == so->used) {
        len = si->len;
+    }
+    Py_END_CRITICAL_SECTION2();
+#else
+    if (si->si_set != NULL && si->si_used == si->si_set->used) {
+        len = si->len;
+    }
+#endif
+


Suggested change

#ifdef Py_GIL_DISABLED

PySetObject *so = si->si_set;

assert(so != NULL);

Py_BEGIN_CRITICAL_SECTION2(op, so);

if (si->si_pos >= 0 && si->si_used == so->used) {

len = si->len;

}

Py_END_CRITICAL_SECTION2();

#else

if (si->si_set != NULL && si->si_used == si->si_set->used) {

len = si->len;

}

#endif

PySetObject *so = si->si_set;

Py_BEGIN_CRITICAL_SECTION(op);

if (si->si_pos >= 0 && si->si_used == FT_ATOMIC_LOAD_SSIZE_RELAXED(so->used)) {

len = si->len;

}

Py_END_CRITICAL_SECTION();

This is less code and a bit more efficient (only a single lock):

From the set object so = si->si_setwe only need to read the fieldusedwhich we can do with an atomic load. In the FT buildsois never NULL. In the non-FT buildsocan only be NULL ifsi->si_set < 0`)

@hyongtao-code I did not test the above suggestion

eendebakpt · 2026-02-10T19:47:48Z

Objects/setobject.c

+    else {
+        /* exhausted */
+        si->si_pos = -1;
+        si->len = 0;


Is this needed?

eendebakpt · 2026-02-10T19:48:56Z

Objects/setobject.c

    if (key == NULL) {
-        si->si_set = NULL;
+        /* exhausted */
        Py_DECREF(so);


Can you move the decref to the section

/* exhausted */ si->si_pos = -1; si->len = 0;

?

hyongtao-code added 2 commits January 31, 2026 10:51

Add test case

3e3785c

hyongtao-code requested a review from rhettinger as a code owner January 31, 2026 03:34

bedevere-app bot mentioned this pull request Jan 31, 2026

Data race in set iterator length_hint under no-gil #144356

Open

bedevere-app bot added the awaiting review label Jan 31, 2026

📜🤖 Added by blurb_it.

229ced3

eendebakpt reviewed Jan 31, 2026

View reviewed changes

Lib/test/test_free_threading/test_set.py Outdated Show resolved Hide resolved

Objects/setobject.c Show resolved Hide resolved

hyongtao-code added 2 commits February 1, 2026 23:25

Resovle comments

cdcf88a

Update test case: used_race and exhaust_race

21f1478

hyongtao-code changed the title ~~gh-144356: fix data race in setiter_len() under no-gil~~ gh-144356: Avoid races when computing set_iterator.__length_hint__ under no-gil Feb 1, 2026

eendebakpt reviewed Feb 1, 2026

View reviewed changes

Objects/setobject.c Show resolved Hide resolved

colesbury reviewed Feb 2, 2026

View reviewed changes

Objects/setobject.c Outdated Show resolved Hide resolved

colesbury requested a review from kumaraditya303 February 2, 2026 18:58

hyongtao-code added 3 commits February 7, 2026 00:07

Add test case

a18c698

Try to address the comments

79b5fbc

Post fix

6ac15e0

eendebakpt reviewed Feb 6, 2026

View reviewed changes

hyongtao-code added 2 commits February 8, 2026 11:33

Address comments

3222eef

Post fix

7fb39bc

eendebakpt requested changes Feb 9, 2026

View reviewed changes

Lib/test/test_free_threading/test_set.py Outdated Show resolved Hide resolved

Objects/setobject.c Outdated Show resolved Hide resolved

Objects/setobject.c Show resolved Hide resolved

bedevere-app bot added awaiting core review and removed awaiting review labels Feb 9, 2026

hyongtao-code and others added 2 commits February 9, 2026 19:09

Update Lib/test/test_free_threading/test_set.py

78241a8

Co-authored-by: Pieter Eendebak <pieter.eendebak@gmail.com>

Revert "Update Lib/test/test_free_threading/test_set.py"

52285a2

This reverts commit 78241a8.

eendebakpt reviewed Feb 9, 2026

View reviewed changes

Lib/test/test_free_threading/test_set.py Outdated Show resolved Hide resolved

hyongtao-code added 2 commits February 9, 2026 22:15

Address comments

c9ece31

Enhance the test case as test_itertools.py

e41e852

hyongtao-code changed the title ~~gh-144356: Avoid races when computing set_iterator.__length_hint__ under no-gil~~ gh-144356: Make set iterator __length_hint__ and iternext race-safe under no-gil Feb 10, 2026

Update NEWS

b32b5ec

eendebakpt reviewed Feb 10, 2026

View reviewed changes

Lib/test/test_free_threading/test_set.py Outdated Show resolved Hide resolved

Lib/test/test_free_threading/test_set.py Outdated Show resolved Hide resolved

Objects/setobject.c Outdated Show resolved Hide resolved

Objects/setobject.c Outdated Show resolved Hide resolved

hyongtao-code added 2 commits February 10, 2026 21:44

Address comment

d14856c

Address comments

2a512a8

eendebakpt reviewed Feb 10, 2026

View reviewed changes

Update test cases

1ce898a

Uh oh!

gh-144356: Make set iterator __length_hint__ and iternext race-safe under no-gil #144357

Are you sure you want to change the base?

gh-144356: Make set iterator __length_hint__ and iternext race-safe under no-gil #144357

Conversation

hyongtao-code commented Jan 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hyongtao-code commented Feb 1, 2026

Uh oh!

Uh oh!

Uh oh!

eendebakpt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hyongtao-code commented Feb 8, 2026

Uh oh!

eendebakpt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

eendebakpt Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

eendebakpt Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

eendebakpt Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

gh-144356: Make set iterator `__length_hint__` and `iternext` race-safe under no-gil #144357

gh-144356: Make set iterator `__length_hint__` and `iternext` race-safe under no-gil #144357

hyongtao-code commented Jan 31, 2026 •

edited

Loading