JingsongLi commented on PR #11178: URL: https://github.com/apache/paimon/pull/11178#issuecomment-6921957458
Revalidated `7955643468` after the retraction/flush-boundary revision. Requirement fit: SUPPORTED. Implementation: FINDINGS (the existing Native P1 remains). The revised Java contract preserves other writers' columns across the tested buffer boundaries, and sequence-group removal is again outside this change. All 76 selected Java merge-function, real-table and chain-table tests passed on JDK 11 with normal Maven checks; no additional Java regression was established. The cross-reader disagreement is still reproducible on this exact head. A real fixed-bucket Parquet partial-update table with `partial-update.remove-record-on-delete=true` contains a committed `+I(1,10,20)` followed by a lone `-U(1,10,NULL)`. Java classes compiled from this worktree return zero rows, while supported `pypaimon-rust 0.5.0` returns the stale `(1,10,20)` row with Python fallback forbidden. Both readers used the identical persisted files, whose SHA-256 hashes stayed unchanged. The current dependency still permits `pypaimon-rust>=0.5.0`, and the Native routing has no semantic capability/version guard. Aligning Rust is necessary; please also require the compatible runtime or explicitly reject the incompatible path before shipping this Java behavior. The Python fallback currently rejects these removal options, so an assumed automatic fallback is insufficient. This is the same [existing P1](https://github.com/apache/paimon/pull/11178#discussion_r4583489915), rather than a duplicate inline finding. The exact-head CI is still queued. -- 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]
