hesam-oxe commented on PR #11077: URL: https://github.com/apache/seatunnel/pull/11077#issuecomment-6843964451
New head is up, with the fork `Build` now running and linked. Thanks for the corrections to the earlier plan — they saved a lot of rework. **Fork Build run for this head:** https://github.com/hesam-oxe/seatunnel/actions/runs/43171582616 - **Style gates:** all green (Code style, License header, Dead links). - **unit-test:** all green, including the two new tests below. - **Integration tests:** 83 jobs succeeded. One unrelated failure: `jdbc-connectors-it-part-7 (11)` — the jdbc OpenGauss e2e container test fails at container startup (`start-seatunnel-flink-13/15-connector-v2.sh` STDERR), which is not in this PR's diff (the diff is `seatunnel-api` multi-table sink + file connector + docs only). Per your offer: could you check whether it also fails on `dev`? Happy to retrigger if it is flaky. **What this head addresses (per the consolidated list):** 1. **Issue 2 (docs accuracy)** — `incompatible-changes.md` EN and ZH entries no longer promise a startup `IllegalStateException` for the file sink. They now describe the actual behavior: the file destination identifier folds in the physical row layout, so tables with divergent layouts keep separate writers (no merge, no error); the coordinator-level `IllegalStateException` is described as applying to connectors whose destination identifiers do not already distinguish row layouts. The PR release note was updated the same way. 2. **Issue 3 (tests)** — `HdfsFileSinkTest#testPhysicalDestinationIdentifierPinsEveryDestinationComponent` pins every identifier component (path, filesystem endpoint, format, tmp path, file-name expression, row type, and the order-independent fs-options hash) plus a metadata-only variant keeping the identifier; `MultiTableSinkWriterTest#testSharedDestinationWithMetadataOnlySchemaDifferencesSharesOneWriter` covers sharing with metadata-only schema differences (column length/scale/comment). 3. **Conflicts against `dev`** — resolved in `cea28e327`: `MultiTableSinkAggregatedCommitter` keeps the identity-grouped shared-committer iteration and picks up the dev `restore`/`commit` routing; the test file keeps both the shared-committer lifecycle test and the restore-routing test. A merge-resolution mistake that briefly broke the test file header was caught by the CI gates and repaired (`6a70ae62`, `8f208be4`); final formatting applied via spotless (`72ea816e`, `352e5a27`). Nothing was rewritten from the already-covered areas (writer creation outside `computeIfAbsent`, alias context registration, `validateSharedDestinationSchemas`, shared-writer identity grouping) — kept as-is per your note. -- 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]
