@kumaraditya303 I don't know why but this started showing on our tests after #98572. I can't tell if that PR introduced a bug or it just exposed something wrong with our implementation.

IMO, it seems it's something wrong with our implementation.

asyncio always has surprises hence it is my favorite :)

Seriously though this has uncovered another race in our implementation. The code does not guard against concurrent closing of the stream which leads to this failure. The connection_lost callback should be called only once and here it is being called more than once.

Here's a better fix:

diff --git a/Lib/asyncio/proactor_events.py b/Lib/asyncio/proactor_events.py
index 2685a3376c..c6aab408fc 100644
--- a/Lib/asyncio/proactor_events.py
+++ b/Lib/asyncio/proactor_events.py
@@ -152,6 +152,8 @@ def _force_close(self, exc):
         self._loop.call_soon(self._call_connection_lost, exc)
 
     def _call_connection_lost(self, exc):
+        if self._called_connection_lost:
+            return
         try:
             self._protocol.connection_lost(exc)
         finally:

Feel free to update this PR with the patch above, I have verified this.

Feel free to update this PR with the patch above, I have verified this.

Awesome! Thank you for the fix. I was trying to find something more elegant but I'm a noob at asyncio internals :). Your fix is way nicer.

I created #98730 to add tests for this.

Thanks @Fidget-Spinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.10, 3.11.
🐍🍒⛏🤖

Sorry @Fidget-Spinner, I had trouble checking out the 3.11 backport branch.
Please retry by removing and re-adding the "needs backport to 3.11" label.
Alternatively, you can backport using cherry_picker on the command line.
cherry_picker 8a755423eba8e19704d96905730cf7f50083eb23 3.11

GH-98753 is a backport of this pull request to the 3.10 branch.

Thanks @Fidget-Spinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.11.
🐍🍒⛏🤖

GH-98754 is a backport of this pull request to the 3.11 branch.