li3zhi4 commented on PR #11746: URL: https://github.com/apache/seatunnel/pull/11746#issuecomment-6099921329
Thanks for the complete re-review and for catching the inherited-dialect gap. All requested source/documentation follow-ups are now pushed in `34af774c36` on top of the verified head.\n\n**Issue 1 (must-fix): validated-dialect gate is now explicit.** `StarRocksDialect`, `DsqlDialect`, `OpenGaussDialect`, and `PostgresLowDialect` each override `supportCompositeKeySplit(DatabaseMetaData)` to return `false`. The validated MySQL, PostgreSQL, SQLite, SQL Server and Oracle paths are unchanged. `CompositeKeyChunkSplitterTest#testUnvalidatedDerivedDialectsDisableCompositeKeySplit` asserts all four derived dialects remain opted out.\n\n**Issue 3:** Added matching EN/ZH entries to `introduction/concepts/incompatible-changes.md`: numeric/date composite-PK tables on the validated dialects may have different boundaries, chunk counts and startup time after upgrade while reading the same rows; `partition_column` or `enable_concurrent_read = false` are the migration options.\n\n**Issue 2 (deferre d, documented):** I kept the behavior unchanged as agreed: composite boundary probes still use the base table while filtered reads apply `where_condition`, which can skew or empty chunks but does not lose or duplicate filtered rows. The EN/ZH JDBC composite section now states this limitation explicitly and leaves the where-scoped boundary query as a follow-up requiring its own E2E.\n\n**Issue 6:** imported `SqlType` normally, changed the duplicate-boundary guard to use numeric-aware `compareArrays(...) == 0`, and clarified the composite docs as numeric/date key columns with STRING fallback.\n\nIssue 5 remains the non-blocking fixed-port/retry-safety observation; no unrelated E2E setup redesign was added.\n\nVerification on the pushed head: `CompositeKeyChunkSplitterTest` **21 passed**, JDBC Spotless check passed, JDBC module verify passed, and `git diff --check` is clean. The official MySQL SplitIT was already run on the same feature behavior before this small dialect/documentation follow-up: **7/7, zero failures/errors/skips**.\n\nThe only remaining gate is external: the fork has not produced a Build run for the latest SHA, so the Apache-side check remains ACTION_REQUIRED rather than green. Please take the final review on `34af774c36` when convenient.\n -- 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]
