SEZ9 commented on issue #12562: URL: https://github.com/apache/seatunnel/issues/12562#issuecomment-5923261870
Thanks for writing this up so clearly. For F2, the state-driven option seems preferable to the fixed sleeps: a package-private `isTracking(Path)` on `FileCollectReader`, awaiting `!isTracking(logFile)` and then `isTracking(logFile)` before appending the second line, removes the dependency on poll timing entirely, so a stall on a loaded runner can only delay the test rather than reorder it. If the sleep-based sequence is chosen instead, it would be worth documenting the `CLOSE_INACTIVE_MS` / `GLOB_SCAN_INTERVAL_MS` ordering it relies on. For F5, accumulating polled events outside the Awaitility lambda (draining `poll(10)` into a shared list each iteration and asserting on the accumulated list) keeps the condition monotonic and makes the failure message show what was actually observed instead of `expected: <1> but was: <0>`. Two questions: has this race been observed in an actual CI run (if so, could you link it here so the before/after can be tracked), and should both fixes land in one PR, or F5 separately since it only affects diagnostics? <!-- streview-comment:1431 --> -- 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]
