SEZ9 commented on PR #11885:
URL: https://github.com/apache/seatunnel/pull/11885#issuecomment-5595086864
@DanielLeens thanks for the detailed write-up and for tracing the rebase —
that explains why `dc699a34b` is no longer reachable and what `befad345c`
("Repair test seam damaged by rebase conflict resolution") is for. Your comment
looks truncated, though: it stops at "At `2026-09-08T05:07:10Z` the PR head
was". Could you post the rest of that trace?
Mapping your description against my earlier review points, here is what I
still need to see in the diff before I can close any of them:
- **PR11885-F1 / F6**: you describe the resolved offset as carried in
`IncrementalPhaseState.stopOffset`, with `addSplits()` adopting the offset from
an in-flight split for legacy checkpoints. My concern was that the field was
not written/read by the checkpoint serializer and that the restore test in
`IncrementalSplitAssignerTest` bypassed it. If `befad345c` touched the
serializer or that test, a pointer to the relevant hunk would help.
- **PR11885-F2**: your explanation relies on `HybridSplitAssigner` only
creating incremental splits after snapshot completion. I still want to confirm
from the code that the `resolvedStopOffset == null` guard is the only path
producing a stop offset for `latest` mode and that all splits from one
assignment share the same cached value.
- **PR11885-F5**: you mention the resolution is retried with backoff. I have
not yet seen where that lives or whether it is tested, so I'll leave this open
until I can check the diff.
- **PR11885-F3**: not covered in your summary — do you know whether the
readiness gate in `MysqlCDCStopModeSpecificIT` changed in the rebase?
- **PR11885-F4 / F7 / F8**: these concern the connector docs and the PR
description; I'll verify them against the current PR body and diff rather than
the summary here.
Once the rest of the trace is up I'll re-review the current head.
<!-- streview-comment:917 -->
--
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]
