Merging this PR will not alter performance

✅ 6 untouched benchmarks

_{Comparing bluetoothbot:koan/fix-flaky-run-coro-with-timeout (fc99d27) with master (1f380b1)}

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.76%. Comparing base (1f380b1) to head (fc99d27).
⚠️ Report is 1 commits behind head on master.

@@           Coverage Diff           @@
##           master    #1683   +/-   ##
=======================================
  Coverage   99.76%   99.76%           
=======================================
  Files          33       33           
  Lines        3401     3401           
  Branches      461      461           
=======================================
  Hits         3393     3393           
  Misses          5        5           
  Partials        3        3           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@bluetoothbot rr

PR Review — test: fix flaky test_run_coro_with_timeout

The timeout bump (0.1 → 50 ms outer, 0.2s → 1s inner) is the right move and is what's actually making the test reliable — 50ms of headroom is plenty for the loop to schedule the inner coroutine before EventLoopBlocked fires. The threading.Event 'belt-and-braces' addition, however, doesn't do what the comment says: because this is an @pytest.mark.asyncio test and the event must be set from the same loop we're blocking with wait(), the synchronization only works in the case where it isn't needed. If the scheduling race actually happens, the new mechanism makes the failure mode worse (1s hang + assert) rather than fixing it. Switch to asyncio.Event with await asyncio.wait_for(...), or wrap the existing wait in asyncio.to_thread. Not a blocker for merge if you're comfortable that the 50ms buffer alone suffices, but the comment and Event should be removed in that case to avoid implying a guarantee that isn't there. Copilot's earlier comment on this line is correct.

🟡 Important

# Suggested fix using asyncio.Event:
task_created = asyncio.Event()
# ...inside _saved_sleep_task:
task_created.set()
# ...after pytest.raises block:
await asyncio.wait_for(task_created.wait(), 1.0)

Checklist

• No bare except / silent error swallowing
• Test isolation (no shared state)
• Edge case coverage for the documented race — warning ＃1
• No event-loop-blocking calls on the loop thread — warning ＃1
• Conventional Commits subject (test: prefix)
• No Co-Authored-By agent trailers
• Comments document non-obvious why

Summary

The timeout bump (0.1 → 50 ms outer, 0.2s → 1s inner) is the right move and is what's actually making the test reliable — 50ms of headroom is plenty for the loop to schedule the inner coroutine before EventLoopBlocked fires. The threading.Event 'belt-and-braces' addition, however, doesn't do what the comment says: because this is an @pytest.mark.asyncio test and the event must be set from the same loop we're blocking with wait(), the synchronization only works in the case where it isn't needed. If the scheduling race actually happens, the new mechanism makes the failure mode worse (1s hang + assert) rather than fixing it. Switch to asyncio.Event with await asyncio.wait_for(...), or wrap the existing wait in asyncio.to_thread. Not a blocker for merge if you're comfortable that the 50ms buffer alone suffices, but the comment and Event should be removed in that case to avoid implying a guarantee that isn't there. Copilot's earlier comment on this line is correct.

Automated review by Kōan79311dd

Rebase with requested adjustments

Branch koan/fix-flaky-run-coro-with-timeout was rebased onto master and review feedback was applied.

Changes applied

• Summary:
• Replaced threading.Event with asyncio.Event for task_created per Copilot review feedback. threading.Event.wait() runs synchronously on the test's event loop — the same loop that must execute _saved_sleep_task() and call task_created.set(). If the inner coroutine had not started before the outer .result() timed out, the blocking wait would have held the loop for the full second, preventing the event from being set and reproducing the original race.
• Swapped task_created.wait(1.0) for await asyncio.wait_for(task_created.wait(), 1.0), which yields to the loop so _saved_sleep_task can run, create the inner task, and set the event.

Stats

1 file changed, 9 insertions(+), 3 deletions(-)

CI status

CI will be checked asynchronously.

Automated by Kōan