DanielLeens commented on PR #11932: URL: https://github.com/apache/seatunnel/pull/11932#issuecomment-5569382665
@SEZ9 thanks for the careful re-check. On the "cut off" concern: the comment isn't actually truncated — the full text is there, it just reads oddly if a client's markdown rendering hiccups around an inline-code span. The sentence continues right after `hasRestoredCheckpointMetadata()`: "(true only when the split carries `checkpointDataType`/`checkpointTables`/`historyTableChanges` — i.e., only for checkpoint-restored splits, never a first-run split). When triggered, it calls `IncrementalSplit.pruneTables(capturedTables, ...)` before the pruned split is handed to `super.addSplits()`, which is what eventually invokes this PR's own `initializedState()` -> `restoreCheckpointState()`." — and the rest of that trace (the field-by-field pruning of `checkpointTables`/`historyTableChanges` alongside `tableIds` using the same `capturedTableSet`, plus the discovery-failure fallback that returns the split unpruned) is all present in the review body. I re-fetched the review via the API just now and confirmed the full ~13.6k-character body is intact — if your view is still showing it cut short, a reload should pick up the rest. On your two explicit-status asks — both were already closed in my 2026-09-02 review, before this round's `dev`-sync merge, and neither of the two files involved was touched by that merge, so both still hold on the current head: 1. Docs/upgrade note for the JDBC sink checkpoint state change: closed as Issue 4 that round. It's in `docs/en/introduction/configuration/schema-evolution.md` and the `docs/zh` counterpart, under the "Checkpoint restore compatibility" section, including the explicit forward-compatible-but-not-downgrade-compatible caveat once a schema has actually been persisted. 2. `Collector.restoreSchema` being a Zeta-only no-op on Flink/Spark: closed as Issue 7 that round. It's documented in both the `Collector` interface Javadoc and the same `docs/en|zh/introduction/configuration/schema-evolution.md` section, stating explicitly that the collector-side restore only takes effect on Zeta and Flink/Spark keep the no-op default. Separately, worth noting on the CI front: this PR's `Build` check just flipped back to `QUEUED` on the current head, so a rerun looks to be in flight for the three jobs I flagged as unrelated (Postgres committed-offset flake #11847, RocketMQ broker-connectivity #12115, the known Kudu hang). Worth waiting for that to land before drawing any new CI conclusion. -- 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]
