li3zhi4 commented on PR #11746: URL: https://github.com/apache/seatunnel/pull/11746#issuecomment-6165460314
@DanielLeens @SEZ9 — pushed `e7586c06a` to address the shared-image teardown issue. Both Oracle and SQL Server SplitITs now use `PullPolicy.alwaysPull()`, so Testcontainers does not rely on its cached image-presence result after a sibling removes the image. I checked actual class ordering rather than treating a standalone SplitIT run as proof. With `-Dfailsafe.runOrder=alphabetical` and `-Dit.test=JdbcSqlServerIT,JdbcSqlServerSplitIT`, the log shows JdbcSqlServerIT running first, deleting the shared image, then JdbcSqlServerSplitIT re-pulling and starting it. Results: sibling 10/10 and SplitIT 2/2, zero failures/errors/skips; BUILD SUCCESS. Spotless passes for both affected E2E modules and diff-check is clean. Oracle ordered validation selected both classes correctly. After JdbcOracleIT teardown, JdbcOracleSplitIT successfully re-pulled the deleted image: the original image-not-found failure is gone. However, both Oracle classes timed out waiting for DATABASE IS READY TO USE during first-time database/PDB initialization with the current floating image. Oracle test methods did not execute, so this is NOT a green Oracle acceptance result. I have not changed the image version or weakened assertions to hide it. The next CI run is needed for final acceptance. For SEZ9's F1–F8 status on the current implementation: - F1 done: NULL key components excluded from boundary probes, included in first split and excluded from middle/last splits; SQLite NULL regression is retained. - F2 done: Oracle runtime major-version gate (<12 and SQLException fall back with WARN); version tests cover 11/12/23. - F3/F4 done: any STRING PK component prevents composite splitting and retains the single-column behavior; official MySQL mixed-case/non-ASCII/collation regression covers fallback and exact row membership. - F5 intentional metadata-only limitation: no separator decoding exists; tuples/key names are carried structurally. Comma-containing values are not encoded into this display field. No runtime rejection was added. - F6 done: mixed Number tuples compare numerically via BigDecimal.compareTo; duplicate guard uses compareArrays == 0, with Integer/Long and differing BigDecimal-scale tests. - F7 done: EN/ZH Upgrading/opt-out and incompatible-change entries; partition_column selects a single key and enable_concurrent_read=false selects a full-table split. - F8 done: findSplitKey has a dynamic-only gate, and FixedChunkSplitter defensively rejects multi-column keys; both have tests. All statements above distinguish implementation from reviewer sign-off and remote Build acceptance; the final approval/green Build gates are still outstanding. -- 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]
