sunchao commented on PR #6439: URL: https://github.com/apache/datafusion-comet/pull/6439#issuecomment-5916409412
@andygrove @viirya Addressed in 1627bd7d5. Added an end-to-end Parquet-backed `REPARTITION(4)` / grouped `AVG` regression with AQE enabled, native hash partitioning disabled, and native round-robin partitioning enabled. It checks nullable AVG results, the executed native round-robin shuffle, restored Spark aggregates, and absence of placeholders. The planner-level test now builds its round-robin input stage under the same configuration, with no hash-partitioning toggle. Restoring only the old cleanup reproduces `CometNativeExec should not be executed directly without a serialized plan` on a leftover sink around the materialized native repartition stage. All three placeholder regressions fail in that control; the other 83 tests pass. With the fix restored, all 86 `CometExecRuleSuite` tests pass on Spark 4.1.3 / JDK 21. Native build, formatting, and Scala style checks also pass. The existing Spark SQL and all-profile CI labels remain applied. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
