SEPURI-SAI-KRISHNA commented on PR #12381: URL: https://github.com/apache/seatunnel/pull/12381#issuecomment-5781923265
A second read of this, plus failure data that post-dates @DanielLeens's approval. I looked because this flake is currently blocking #12438 of mine, which is also the third row in the table below. **The change checks out, read against the source rather than run.** `PayPalClientTest` has exactly two `arrived.await` sites, at lines 395 and 416, and both are widened. The two bounds that actually test client behaviour are left at 15s: `result.get` at 397 for "close aborts the active body", and `complete.await` at 423 for "close wakes the retry wait". So a genuine hang in the close wiring still fails in 15 seconds, and only the environmental waits get the 60s budget. That split is the part worth keeping in review notes, because it is what stops this from being a blanket timeout bump. `FileCollectReaderBehaviorTest` follows the same principle: the three `atMost(3, TimeUnit.SECONDS)` ceilings at lines 74, 84 and 99 are all widened to the shared 10s budget, while `CLOSE_INACTIVE_MS` and `GLOB_SCAN_INTERVAL_MS`, which are the reader behaviour being exercised, are left untouched. Same split, so both halves of the PR widen only what the runner controls. One observation, not an objection: line 417's `deadline` for the worker reaching `TIMED_WAITING` is widened too. That is test synchronisation rather than client behaviour so the change is right, but it does mean a genuine failure to enter the retry wait now takes 60s to surface instead of 15s. Slower failure, same detection. **Fresh data on how often this fires.** Across three of my PRs between 2026-09-21 and 2026-09-22, `unit-test (11, windows-latest)` failed on `closeWakesRetryWait:416` in 4 of 6 attempts: | PR | attempt 1 | attempt 2 | |---|---|---| | #12416 | fail | pass | | #12182 | fail | pass | | #12438 | fail | fail | So it is closer to a two-thirds failure rate than an occasional flake, and rerunning is not a reliable workaround: #12438 has now failed it twice in a row with a comment-only diff that shares no code with `connector-http-paypal`. This has had @DanielLeens's approval since 09-18 and the diff is unchanged since. It needs a committer to take 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]
