Skip to content

gh-126083: Fix a reference leak in asyncio.Task when reinitializing with new non-None context #126103

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

Merged
merged 4 commits into from
Oct 31, 2024
Merged

gh-126083: Fix a reference leak in asyncio.Task when reinitializing with new non-None context #126103

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
614d20f
Fix ref leak and add tests
Nico-Posada Oct 29, 2024
8e5effe
Address review comments
Nico-Posada Oct 29, 2024
a588f7b
fix linter errors
Nico-Posada Oct 29, 2024
2e70b51
Address sobolevn's review
Nico-Posada Oct 30, 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
26 changes: 26 additions & 0 deletions Lib/test/test_asyncio/test_tasks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2688,6 +2688,32 @@ def test_get_context(self):
finally:
loop.close()

def test_proper_refcounts(self):
# see: https://github.com/python/cpython/issues/126083
class Break:
def __str__(self):
raise RuntimeError("break")

obj = object()
initial_refcount = sys.getrefcount(obj)

coro = coroutine_function()
loop = asyncio.new_event_loop()
task = asyncio.Task.__new__(asyncio.Task)

for _ in range(5):
try:
task.__init__(coro, loop=loop, context=obj, name=Break())
except Exception as e:
Copy link
Member

Choose a reason for hiding this comment

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

question: do we expect this exception in all cases? if so, use with self.assertRaisesRegex(RuntimeError, 'break'):

Copy link
Contributor Author

Choose a reason for hiding this comment

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

So update it to be this?

for _ in range(5):
    with self.assertRaisesRegex(RuntimeError, 'break'):
        task.__init__(coro, loop=loop, context=obj, name=Break())

Copy link
Member

Choose a reason for hiding this comment

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

FTR: In order to catch the correct exception in #126120, I used a custom exception + a special argument so that I'm sure that I catch and expect the correct one (and not just a random one because of something else). So what I did was something like:

# what I want to catch
def foo():
	raise ReachableCode(1)

# what I don't want to catch yet
def bar():
	raise ReachableCode(2)

# where I call the thing I want to catch
with self.assertRaises(ReachableCode) as cm:
	something_that_calls_foo()
self.assertEqual(len(cm.exception.args), 1)
self.assertIs(cm.exception.args[0], 1)

I find this pattern quite good (maybe a bit too exhaustive) because I can really see what I'm catching.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I like the idea of the ReachableCode exception class, but I still think it's easier and more readable to just do

with self.assertRaisesRegex(ReachableCode, 'str here'):
    ...

and use a string in the arg instead of an int

Copy link
Member

Choose a reason for hiding this comment

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

I think I'll go with that as well to simplify the tests on my side (but not today).

# exception should only come from Break.__str__ as it's used to
# avoid having to do too much setup for this test
self.assertIs(type(e), RuntimeError)
self.assertEqual(e.args[0], "break")
coro.close()
del task

self.assertEqual(sys.getrefcount(obj), initial_refcount)
Copy link
Contributor

@kumaraditya303 kumaraditya303 Nov 2, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

Please avoid tests for exact refcounts, the test should test that no references are leaked rather than reference count should be equal, both are not same thing. In this particular case all of this seems unnecessary, our existing refleak checker is sufficient for this.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Oh I was unaware of the existing refleak checker. Do you have any examples of where it's used so i can check it out?

Copy link
Member

Choose a reason for hiding this comment

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

You can run python -m test XXX -R : (it's with -R :)



def add_subclass_tests(cls):
BaseTask = cls.Task
Expand Down
1 change: 1 addition & 0 deletions Misc/NEWS.d/next/Library/2024-10-28-22-35-22.gh-issue-126083.TuI--n.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Fixed a reference leak in :class:`asyncio.Task` objects when reinitializing the same object with a non-``None`` context. Patch by Nico Posada.
2 changes: 1 addition & 1 deletion Modules/_asynciomodule.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2120,7 +2120,7 @@ _asyncio_Task___init___impl(TaskObj *self, PyObject *coro, PyObject *loop,
return -1;
}
} else {
self->task_context = Py_NewRef(context);
Py_XSETREF(self->task_context, Py_NewRef(context));
}

Py_CLEAR(self->task_fut_waiter);
Expand Down
Loading