SEZ9 commented on PR #11780: URL: https://github.com/apache/seatunnel/pull/11780#issuecomment-5409353121
Thanks @DanielLeens for the fresh re-review against `380d77ae19` and for confirming via CI run `32637942572` that the `BasicType.LONG_TYPE` fix resolves the compile break from `b411c63bd8` — the build is unblocked. That commit only addresses the test compile issue, though. The remaining points from the previous review are still open: 1. **PR11780-F1 (HIGH, blocking):** The collector schema restore is still not gated on whether the Debezium deserializer actually applied the restored schema. On CDC dialects without a `SchemaChangeResolver` (Postgres/OpenGauss/MongoDB) this deterministically produces an `ArrayIndexOutOfBoundsException` after recovery. 2. **PR11780-F2:** The legacy `MultipleRowType` fallback can hard-fail via `TablePath.of()` on non 2/3-segment table ids, turning a previously-restorable checkpoint into a fatal restore error. Please guard this path. 3. **PR11780-F3:** `restoredCheckpointTables` is (re)set on every `initializedState` call, so mid-run split re-adds can rewind the collector schema after DDL has already advanced it. Please make the restore one-shot. 4. **PR11780-F4 / F6:** `Collector#restoreSchema` is Zeta-only (only `SeaTunnelSourceCollector` overrides it) and the default is a silent no-op on other engines. Please state the engine scope and recovery behavior in the Javadoc and schema-evolution docs. 5. **PR11780-F5:** `restoreCheckpointHistoryTableChanges` clears the target map before `putAll`, which self-destructs if the caller ever passes the live `tableChangesStructMap`. Please make this safe or defensively copy. 6. **PR11780-F7 / F8:** Minor cleanups — mark the package-private `restoreCheckpointState` as test-only, trim the INFO log that prints the full `CatalogTable` list, drop the fully-qualified `SeaTunnelRowType` in the `toLegacyCheckpointTable` signature, and extract the duplicated magic catalog name. One note: your review comment appears truncated mid-sentence at "Issue 2 (carried over from 202" — could you repost the rest so the full context is on the thread? Still blocking on F1 at minimum. Thanks! <!-- streview-comment:557 --> -- 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]
