SEPURI-SAI-KRISHNA commented on issue #12332: URL: https://github.com/apache/seatunnel/issues/12332#issuecomment-5741198143
All four done or answered. **1 and 2.** The body is updated. The verified/not-verified caveat is now the first thing under "What happened", ahead of the mechanism, with a dated correction notice linking to the correction comment so the history stays readable. The `<topic>` wording is replaced by `%RETRY%<group>` throughout, including in the Error Exception block. The "unlucky once" and discovery-interval argument is gone, replaced by the narrow window it actually is: between `run():179` and `run():180` of one synchronized block, on a group that already has committed offsets. The "Possible fix" section now describes the two-branch behaviour and points at #12349. **4. Confirmed.** Both tests are in #12349's current head, `f98275e5`, in `RocketMqAdminUtilTest`: * `testCurrentOffsets_retryTopicMissingWhileRouteHealthyReturnsEmpty` * `testCurrentOffsets_routeOutageSurfacesAsFailure` alongside `testCurrentOffsets_otherResponseCodeSurfacesEvenWhenRouteHealthy`, which pins that the route probe applies only to code 17, and the two pre-existing successful-lookup cases. I also checked they fail for the right reasons rather than passing by luck: removing the route probe fails only the outage test, and removing the code 17 branch entirely fails only the retry-topic test. **3. I do not think I can do this cheaply, and I would rather say so than let it sit as an open promise.** The window is between two statements in the same synchronized block, so reproducing it means killing the name server inside that gap. `RocketMqIT` shares one container across every engine leg, so stopping the name server mid-test would take the rest of the class down with it, and a delete-and-restore of just `%RETRY%<group>` races the broker's re-registration heartbeat. I could force it with a test seam that injects the failure, but that proves the handler's behaviour rather than the window's existence, and the unit tests already do that. So the "by construction" caveat stays, and it is now the first thing a triager reads. If you would still like the injected version, say so and I will add it, with the caveat written on it. -- 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]
