Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
Description
Found an interesting defect where the final partition of the `RepartitionExec::execute` thread spawner was consistently not being spawned via `tokio::spawn`. This meant that `RepartitionStream::poll_next` was sitting waiting forever for data that never arrived.
It looks like a race condition where the `JoinHandle` was not being `await`ed and something strange going on with the internals of tokio like lazy evaluation?