@ilooner please review -- This PR was changed, reduced from its original scope as other work (e.g. DRILL-6453) handled the main issue (a STOP outcome with uninitialized container). The main addition is the test file, checking that incoming uninitialized input containers (for STOP or NONE outcomes) are handled correctly. (Basically the same test as before, only now calling `next()` twice (as the first call always returns OK_NEW_SCHEMA)). A small fix was needed for the hash join, as in the case the first real incoming batch was NONE, the second `next()` was returning an OK (with 0 rows). With the change it returns NONE. And a minor change in the MockRecordBatch to avoid getting record count on uninitialized containers.
[ Full content available at: https://github.com/apache/drill/pull/1373 ] This message was relayed via gitbox.apache.org for [email protected]
