SEPURI-SAI-KRISHNA commented on PR #11721: URL: https://github.com/apache/seatunnel/pull/11721#issuecomment-5392902132
Thanks @SEZ9 and @DanielLeens, F1 and F2 closed out, and the CI question is now answered by events rather than by a rerun. Two things happened since the last round. ## The branch went red-flagged as conflicting, and had to be updated #11925 merged on 2026-08-23, and it rewrote **both** listings this PR patches. The blocks I was editing, `extractPrimaryKeyIfPresent(row)` / `% replicaNum`, no longer exist on `dev`; they are now `element.getField(primaryKey.get())` / `% blockingQueues.size()`. GitHub therefore moved this PR to `CONFLICTING`, and it could not be merged in that state by anyone. @SEZ9 asked me *not* to push here for something cosmetic and I agreed, so to be explicit about why I pushed anyway: this isn't that. The PR was unmergeable, so the update was mandatory rather than optional, I'd have left it alone otherwise. @DanielLeens's approval does still show as standing on the new head, but the merge pulled in 64 commits of `dev`, so a re-review is very welcome if you'd rather re-confirm against the new base. ## The merged callout asked for exactly this update #11925 landed a `:::caution Known issue` block in §5.3 saying, verbatim: > This page documents the behaviour currently on `dev`; the defect is tracked in [#11720](https://github.com/apache/seatunnel/issues/11720), and this section should be updated when a fix lands. This PR is that fix, so the callout is retired. It becomes a `:::note` that explains *why* the sign bit is masked rather than describing a defect that no longer exists, and both listings move to `(object.hashCode() & Integer.MAX_VALUE) % blockingQueues.size()`. Three follow-on edits that are easy to miss: - **`docs/zh` carried the same callout.** #11925 added a matching `已知问题` block and the formula `$replica = \mathrm{Math.abs}(hash(pk)) \bmod replicaNum$`. This PR previously touched only `en`; leaving `zh` alone would have left the Chinese page asserting a bug that `dev` no longer has. Both are updated together. - **The random-strategy note referred back to `Math.abs`**, "for exactly the same reason `Math.abs` can above", in both languages. With `Math.abs` gone from the listing above it, that sentence pointed at nothing, so it is reworded to refer to the masking. - **§5 walkthrough listing**, not just §5.3. #11925 put the same known-issue comment inline there too. Net effect: after this merges, no page on the site describes `Math.abs` as the routing implementation, and no page carries a known-issue callout for #11720. ## The `all-connectors-it-1` rerun @SEZ9, your remaining ask was the rerun result for the cancelled `all-connectors-it-1 (8, ubuntu-latest)` job on run `32461786477`. The push above supersedes that run entirely, the new head triggers a full fresh run, which is a stronger signal than rerunning one job on a stale head. I'll report the result here once it settles, and if `all-connectors-it-1` cancels again in the same spot I'll take @DanielLeens up on digging into whether it matches a tracked flake. @DanielLeens, thanks for going back and re-diagnosing rather than restating the earlier read. Distinguishing "same flake recurring" from "second, distinct flake" is the difference between a known issue and an unknown one, and I'd have written it up wrong without that. ## Where this leaves the PR The code change is untouched: `MultiTableSinkWriter.java` and its test are byte-identical to the reviewed `11599d94d`. Only the docs hunks moved, and only because their surroundings were replaced underneath them. F1 resolved via the tracked deferral, F2 withdrawn, no open findings. -- 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]
