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
Commit cfdc89b
Browse filesBrowse the repository at this point in the historyBrowse files
test(server): idle-resume replays wait for the session release before the next prompt
After an `advance_clock` past the idle timeout, the scenario dispatched the
next message right away. The idle release runs on the fiber the clock jump
woke, and it yields (background-work probe, child archive) before it removes
the live entry. When the next turn's `markBusy` got there first, it
interrupted the release, so the turn went to the old process. The Kiro
replay agent expected `session/close` and a restart, saw `session/prompt`, and
the test hung until Vitest timed out. CPU contention from running four server
test files per shard made the losing order show up in CI.
Fixtures whose jump releases the session now say so
(`releasesProviderSession`), and the scenario awaits the committed
`provider-session.updated` stopped event before running later steps. Codex's
rollback fixture keeps its shared process across the jump, so it does not opt in.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
0 commit comments