DanielLeens commented on PR #11545: URL: https://github.com/apache/seatunnel/pull/11545#issuecomment-5968655601
Thanks @SEZ9 - answering from the commit history of the current head `ac6c41cdb` (a dev sync on top of `7c91ebec2`), so you can verify against the real diff. **Where each item lives (all already pushed on this PR, none are description-only):** - F5 / F1 (launcher side): `504eb8b29` "Enforce the Java 11 baseline and mandatory module flags in launch scripts" - touches `seatunnel.sh` and `seatunnel-cluster.sh` under `seatunnel-core/seatunnel-starter/src/main/bin/`. - F2 / F4: `1b1da9e86` "Report denied Kerberos reflective access instead of failing silently" - touches `KuduUtil.java` and `PaimonSecurityContext.java`. - F3: the shipped `config/jvm_options`, `jvm_master_options` (and siblings) now carry a short comment per flag naming its consumer (Hazelcast internals, the `URLClassLoader#addURL` plugin loaders, or the Kerberos reload helpers), so operators can see what each flag is for before pruning. The list itself is unchanged: I kept all six because Hazelcast's own reflective access is not visible from first-party call sites. If you still want the list narrowed, please tell me which flag you believe is unnecessary and I will either drop it or add the evidence. - F6 / F8: the workflow-level `JAVA_TOOL_OPTIONS` is gone from `backend.yml`. The six flags are carried by `SUREFIRE_JVM_ARGS` in the unit-test steps (Linux and Windows), and `7c91ebec2` moves the pom side into a plain `surefire.module.args` property, because the old JDK-activated profile deactivated the `release` profile and dropped `seatunnel-dist` from the reactor (the cause of the "Dependency licenses" failure). **One thing I found while re-reading, which I should fix:** the explanatory comment at the top of `backend.yml` still says the flags come from the `jdk9-plus-test-opens` profile, but `7c91ebec2` deleted that profile in favor of `surefire.module.args`. I will correct that wording in my next push. **Not done yet:** the `incompatible-changes.md` note about the launcher re-injecting flags (F1 docs) - agreed to land it together with the next fix commit. I have not run anything locally for this reply; CI on the fork is the verification path, and the current head's fork run was cancelled/failed after the dev sync, so I will re-trigger it after the next push rather than quote it as evidence. -- 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]
