You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Flaky test: "a mandatory call deadline overrides requestTimeoutMs: 0" races a fixed sleep against a child's 300 ms reply #478
test/mcpl-request-timeout.test.ts:164-189 (main e670363) is flaky. The test:
spawns a stub server whose tools/call answers 300 ms after it receives the call;
calls with a 75 ms deadline;
sleeps a fixed 350 ms after the rejection;
asserts that exactly one orphaned-response arrived.
That leaves about 125 ms of slack. On a loaded runner the reply lands after the sleep, so the test sees [] and fails.
Why it matters
It turns main red with no code change. CI for cabf3d0 (the merge of #412, Oct 18) failed only on macOS / Node 24, only in this test (actual: []). The next merge passed. #407 fixed the same class of problem in the guard's #281 tests.
How it was found
While checking whether main's red CI matched the red check on #372 (Actions run 39087371623).
Suggested fix
Wait for the event rather than sleeping: once(connection, 'orphaned-response'), raced against a timeout of several seconds.
Activity
test(mcpl): the request-timeout tests decide when a late reply comes,…