shangeyao commented on PR #11174:
URL: https://github.com/apache/seatunnel/pull/11174#issuecomment-4862038943

   Hi @DanielLeens, thanks again for the detailed review.
   
   ## CI update (head `7fb1071ad`)
   
   The dedicated Spark 4.1 merge gates are now green on fork run `28495262009`:
   
   | Job | Status |
   |-----|--------|
   | `spark-4-1-unit-test` | ✅ |
   | `spark-4-1-it` | ✅ |
   | `dist-spark41` | ✅ |
   
   The top-level `Build` is still red because of unrelated lanes (e.g. `Dead 
links`, some connector IT timeouts), but the Spark 4.1 path you previously 
asked to be validated is passing end to end.
   
   ## What changed since `80fec7880`
   
   The remaining `WrappedArray$ofRef` failure in `spark-4-1-it` was caused by 
`seatunnel-translation-spark-common` still being compiled against Spark 3.3 / 
Scala 2.12 and shaded into the 4.1 starter jar. The latest commit:
   
   - moves the common translation runtime (`execution/`, `serialization/`, 
`utils/`) into `seatunnel-translation-spark-4.1` and rebuilds it against Spark 
4.1 / Scala 2.13
   - adapts row converters away from Scala 2.12-only types (`WrappedArray`, 
`HashTrieMap`)
   - excludes the entire `seatunnel-translation-spark-common` artifact from the 
shaded 4.1 starter jar
   - fixes `dist-spark41` by using `-Dmaven.test.skip=true` (spark41 tarball 
intentionally does not ship transforms-v2)
   
   ## On the transform support boundary (Issue 1)
   
   You are right that the current contract is inconsistent. This is intentional 
for the first merge slice, but the docs have not caught up yet:
   
   - **Runtime / packaging**: `seatunnel-transforms-v2` is **not** bundled in 
`-spark41-bin` because transform plugins are still built against Scala 2.12 and 
conflict with Spark 4.1's Scala 2.13 runtime.
   - **E2E**: `FakeSpark41IT` exercises a minimal **source → sink** path 
(`fake_to_assert_spark41.conf`) without a transform stage, matching what the 
spark41 tarball actually ships today.
   - **Docs**: `docs/en/engines/spark.md` (and zh) still incorrectly list 
`seatunnel-transforms-v2` as bundled — I agree this needs to be corrected.
   
   **Proposed next step (Option B from your review):** update en/zh docs and 
the quick-start example to clearly state that Spark 4.1 currently supports the 
minimal source/sink path, and that transform support is tracked separately 
(design #11184, Phase 2). I can push that doc-alignment commit in this PR if 
you prefer, or as an immediate follow-up.
   
   Full transform E2E on Spark 4.1 (Option A) depends on making 
`seatunnel-transforms-v2` Scala 2.13 compatible, which is a larger follow-up.
   
   ## Upstream sync
   
   This branch is currently **16 commits ahead / 41 commits behind** 
`apache/seatunnel:dev`. I plan to rebase onto the latest `dev` next so CI 
signal is not mixed with upstream drift.
   
   Please let me know whether you would like the doc/contract alignment 
included in this PR before merge, or if the current CI-green Spark 4.1 
source/sink path is sufficient for a first merge with docs fixed immediately 
after.
   
   Thanks!


-- 
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]

Reply via email to