Rangsh commented on PR #12081: URL: https://github.com/apache/seatunnel/pull/12081#issuecomment-5674745117
@SEZ9 thanks — answering the two remaining points directly so you can start the synced-head pass. @DanielLeens already checked both against `7f69d7f69` ([comment](https://github.com/apache/seatunnel/pull/12081#issuecomment-5674304136)); I'm confirming the same reading and correcting my earlier wording where it was imprecise. ### On the apparent cutoff Agreed with Daniel: that looks like the same GitHub rendering artifact, not a server-side truncation. Comment [`5659635055`](https://github.com/apache/seatunnel/pull/12081#issuecomment-5659635055) is complete via the API — item 2 ends at `...covered by WALWorkHandlerSurvivabilityTest.`, and item 3 (Mockito / MiniDFSCluster out-of-scope note) follows immediately after. No missing remainder on my side. ### 1. Mockito dependency for `imap-storage-file` **There is a new explicit `test`-scope line in this module's `pom.xml` on the current head** (`mockito-junit-jupiter`, for `HdfsWriterFlushCallCountTest` / `WALWorkHandlerSurvivabilityTest`). My earlier “no new module dependency” was imprecise: the artifact was already available transitively from the root POM, so resolution does not change, but this PR does add an explicit module-local opt-in (with a comment saying as much). Harmless / redundant for classpath purposes; kept so the module’s test classpath is obvious without reading the parent. What remains **intentionally out of scope** for this PR is still a real `MiniDFSCluster` end-to-end harness for the HDFS branches — follow-up, as discussed earlier. ### 2. `queryExecuteStatus` timeout logging **Changed intentionally, not left as ERROR+full stack for routine timeouts.** In both `queryExecuteStatus(long, long)` and `batchQueryExecuteFailsStatus`: - `TimeoutException` → `WARN` with `requestId` + elapsed/limit context; full stack only at `DEBUG` - any other unexpected `Exception` → still `ERROR` with the full stack (now also carrying `requestId`) So a routine timed-out wait no longer floods ERROR with a stack trace; a real failure still does. ### Scope reminder (unchanged) This PR stays **Related to #12058 only** (correctness track). On #12058 the isolated flush-collapse A/B on `file:///` is already concluded as a null result and does **not** authorize a CV claim from this PR; variance follow-up continues on that issue separately. Please treat the RequestFuture / WALWorkHandler / logging / Mockito items above as closed for this pass and review the synced head (`7f69d7f69`) directly when convenient. cc @DanielLeens — thanks for the independent diff check; nothing here changes the “Ready to merge” read from my side either. -- 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]
