Dev-next-gen commented on PR #12269: URL: https://github.com/apache/seatunnel/pull/12269#issuecomment-5640575217
@DanielLeens workflows are now enabled on the fork, and the empty commit (5f09e8b) triggered a full Build run: https://github.com/Dev-next-gen/seatunnel/actions/runs/34614982562 The plugin-discovery tests passed on all four unit-test jobs (JDK 8 and 11, Ubuntu and Windows), including `getSinkPluginJarPathsUnderTurkishLocale`. Because the change sits in core, the e2e jobs ran against every engine container rather than Zeta only, and several of them failed. I went through each failing log, and none of them points at plugin discovery: - all-connectors-it-6 (JDK 11, HudiSeatunnelS3MultiTableIT and HudiSparkS3MultiTableIT) and all-connectors-it-2 (JDK 11, IcebergSourceIT): the MinIO container can't be pulled. Docker answers `pull access denied for minio/minio, repository does not exist`, and when I checked today hub.docker.com returned 404 for the `minio/minio` repository. Every test that starts `minio/minio:RELEASE.2024-06-13T22-53-53Z` should fail the same way until that image reference changes. - transform-v2-it-part-2 (JDK 8 and 11, TestPythonTransformIT): the test installs python3 with apt-get inside the Flink 1.13.6 container, and `apt-get update` stops on `Release file for http://deb.debian.org/debian-security/dists/bullseye-security/InRelease is expired`. That InRelease is dated 2026-08-31 with Valid-Until 2026-09-07. The same test passed today in the Build run for #12271, but that run only starts the Zeta container (`Run all container: false` in its log, `true` in this one). - all-connectors-it-1 (JDK 8): Maven Central returned 502 Bad Gateway for `retrofit-converters-2.6.0.pom` while resolving the connector-datahub dependencies, before any test ran. - engine-v2-it (JDK 8 and 11): `SplitClusterFaultToleranceIT.testStreamJobCancelResolvesWhenWorkerCrashesBeforeCancelAck` expected CANCELED and got FAILED. The same assertion fails at the same line in the Build run for #12265. - transform-v2-it-part-1 (JDK 11): `TestFilterRowKindIT.testFilterRowKindMultiTable` on Flink 1.18, where the assert sink counted 93 and 97 rows against a MIN_ROW of 100 and 125 and 150 against a MAX_ROW of 100. The same test fails the same way in the Build run for #12251. - all-connectors-it-2 (JDK 8): `CouchbaseIT.testFakeSourceToCouchbaseSink`, where the Couchbase client timed out in WAIT_FOR_CONFIG after an authentication failure against the container. The job was already running the Couchbase sink at that point, so the connector jar had been resolved. The same test also fails at the same assertion in the Build run for #12265. - unit-test (JDK 11, Windows) and kudu-connector-it (JDK 11) were cancelled by the 90-minute job timeout. The Windows job had already run SeaTunnelSourcePluginDiscoveryTest (6 tests, 0 failures) before it stalled. The diff only changes behaviour when the JVM default locale is Turkish or Azerbaijani. Under any other default locale, `toUpperCase()` and `toUpperCase(Locale.ROOT)` return the same string for `source`, `sink` and `transform`. The MinIO and bullseye failures are in the test environment and don't depend on this branch, so I expect them to show up on any PR that runs the full e2e matrix until they are fixed separately. -- 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]
