Skip to content

GH-125603: Don't count executing generators and coroutines as referrers in gc.gc_referrers. #125640

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
+34 −32
Open

GH-125603: Don't count executing generators and coroutines as referrers in gc.gc_referrers. #125640

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
01bf540
Don't count executing generators and coroutines as referrers in gc.ge…
markshannon Oct 16, 2024
9673b8a
Mark generators left on stack when clearing frame as zombies.
markshannon Oct 22, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions Include/internal/pycore_frame.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,11 +46,12 @@ typedef enum _framestate {
FRAME_SUSPENDED_YIELD_FROM = -1,
FRAME_EXECUTING = 0,
FRAME_COMPLETED = 1,
FRAME_CLEARED = 4
FRAME_CLEARED = 4,
FRAME_ZOMBIE = 5, /* For generators left on the stack of cleared threads */
} PyFrameState;

#define FRAME_STATE_SUSPENDED(S) ((S) == FRAME_SUSPENDED || (S) == FRAME_SUSPENDED_YIELD_FROM)
#define FRAME_STATE_FINISHED(S) ((S) >= FRAME_COMPLETED)
#define FRAME_STATE_FINISHED(S) ((S) > FRAME_EXECUTING)

enum _frameowner {
FRAME_OWNED_BY_THREAD = 0,
Expand Down
5 changes: 2 additions & 3 deletions Lib/test/test_asyncio/test_streams.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1214,9 +1214,8 @@ async def main():
# can't use assertRaises because that clears frames
exc = excs.exceptions[0]
self.assertIsNotNone(exc)
self.assertListEqual(gc.get_referrers(exc), [main_coro])
main_coro = main()
asyncio.run(main_coro)
self.assertListEqual(gc.get_referrers(exc), [])
asyncio.run(main())


if __name__ == '__main__':
Expand Down
19 changes: 5 additions & 14 deletions Lib/test/test_asyncio/test_taskgroups.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,15 +29,6 @@ def get_error_types(eg):
return {type(exc) for exc in eg.exceptions}


def no_other_refs():
# due to gh-124392 coroutines now refer to their locals
coro = asyncio.current_task().get_coro()
frame = sys._getframe(1)
while coro.cr_frame != frame:
coro = coro.cr_await
return [coro]


class TestTaskGroup(unittest.IsolatedAsyncioTestCase):

async def test_taskgroup_01(self):
Expand Down Expand Up @@ -923,7 +914,7 @@ class _Done(Exception):
exc = e

self.assertIsNotNone(exc)
self.assertListEqual(gc.get_referrers(exc), no_other_refs())
self.assertListEqual(gc.get_referrers(exc), [])


async def test_exception_refcycles_errors(self):
Expand All @@ -941,7 +932,7 @@ class _Done(Exception):
exc = excs.exceptions[0]

self.assertIsInstance(exc, _Done)
self.assertListEqual(gc.get_referrers(exc), no_other_refs())
self.assertListEqual(gc.get_referrers(exc), [])


async def test_exception_refcycles_parent_task(self):
Expand All @@ -963,7 +954,7 @@ async def coro_fn():
exc = excs.exceptions[0].exceptions[0]

self.assertIsInstance(exc, _Done)
self.assertListEqual(gc.get_referrers(exc), no_other_refs())
self.assertListEqual(gc.get_referrers(exc), [])

async def test_exception_refcycles_propagate_cancellation_error(self):
"""Test that TaskGroup deletes propagate_cancellation_error"""
Expand All @@ -978,7 +969,7 @@ async def test_exception_refcycles_propagate_cancellation_error(self):
exc = e.__cause__

self.assertIsInstance(exc, asyncio.CancelledError)
self.assertListEqual(gc.get_referrers(exc), no_other_refs())
self.assertListEqual(gc.get_referrers(exc), [])

async def test_exception_refcycles_base_error(self):
"""Test that TaskGroup deletes self._base_error"""
Expand All @@ -995,7 +986,7 @@ class MyKeyboardInterrupt(KeyboardInterrupt):
exc = e

self.assertIsNotNone(exc)
self.assertListEqual(gc.get_referrers(exc), no_other_refs())
self.assertListEqual(gc.get_referrers(exc), [])


if __name__ == "__main__":
Expand Down
3 changes: 2 additions & 1 deletion Objects/genobject.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ gen_traverse(PyObject *self, visitproc visit, void *arg)
PyGenObject *gen = _PyGen_CAST(self);
Py_VISIT(gen->gi_name);
Py_VISIT(gen->gi_qualname);
if (gen->gi_frame_state != FRAME_CLEARED) {
if (gen->gi_frame_state < FRAME_EXECUTING || gen->gi_frame_state == FRAME_ZOMBIE) {
_PyInterpreterFrame *frame = &gen->gi_iframe;
assert(frame->frame_obj == NULL ||
frame->frame_obj->f_frame->owner == FRAME_OWNED_BY_GENERATOR);
Expand Down Expand Up @@ -375,6 +375,7 @@ gen_close(PyObject *self, PyObject *args)

if (gen->gi_frame_state == FRAME_CREATED) {
gen->gi_frame_state = FRAME_COMPLETED;
_PyFrame_ClearLocals(&gen->gi_iframe);
Py_RETURN_NONE;
}
if (FRAME_STATE_FINISHED(gen->gi_frame_state)) {
Expand Down
34 changes: 22 additions & 12 deletions Python/pystate.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1636,18 +1636,28 @@ PyThreadState_Clear(PyThreadState *tstate)

int verbose = _PyInterpreterState_GetConfig(tstate->interp)->verbose;

if (verbose && tstate->current_frame != NULL) {
/* bpo-20526: After the main thread calls
_PyInterpreterState_SetFinalizing() in Py_FinalizeEx()
(or in Py_EndInterpreter() for subinterpreters),
threads must exit when trying to take the GIL.
If a thread exit in the middle of _PyEval_EvalFrameDefault(),
tstate->frame is not reset to its previous value.
It is more likely with daemon threads, but it can happen
with regular threads if threading._shutdown() fails
(ex: interrupted by CTRL+C). */
fprintf(stderr,
"PyThreadState_Clear: warning: thread still has a frame\n");
if (tstate->current_frame != NULL) {
_PyInterpreterFrame *frame = tstate->current_frame;
if (verbose) {
/* bpo-20526: After the main thread calls
_PyInterpreterState_SetFinalizing() in Py_FinalizeEx()
(or in Py_EndInterpreter() for subinterpreters),
threads must exit when trying to take the GIL.
If a thread exit in the middle of _PyEval_EvalFrameDefault(),
tstate->frame is not reset to its previous value.
It is more likely with daemon threads, but it can happen
with regular threads if threading._shutdown() fails
(ex: interrupted by CTRL+C). */
fprintf(stderr,
"PyThreadState_Clear: warning: thread still has a frame\n");
}
do {
if (frame->owner == FRAME_OWNED_BY_GENERATOR) {
PyGenObject *gen = _PyGen_GetGeneratorFromFrame(frame);
gen->gi_frame_state = FRAME_ZOMBIE;
}
frame = frame->previous;
} while (frame != NULL);
Comment on lines +1654 to +1660
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's not sufficient to mark generators as zombies in PyThreadState_Clear(). We call PyThreadState_Clear() one at a time on each non-main thread:

  • At that point, they are already unlinked from the interpreter's list of PyThreadStates
  • The PyThreadState_Clear() calls makes escaping calls (Py_CLEAR()) that can trigger the GC

So the GC may still see generators that are marked as "running", but aren't in any accessible PyThreadState's frame stack.

}

/* At this point tstate shouldn't be used any more,
Expand Down