goutamadwant commented on PR #12444: URL: https://github.com/apache/seatunnel/pull/12444#issuecomment-5841261416
I ran `PayPalClientTest#closeWakesRetryWait` in a loop locally (macOS, a loaded 10-core host, one fresh JVM per run). Base was `dev` at `deb16a3c3`. Each head was the PR's diff applied to that commit. **Unmodified test:** | | JDK 8 | JDK 11 (`-XX:ActiveProcessorCount=2`) | |---|---|---| | base | 0 / 50 | 5 / 50 failed, all at `PayPalClientTest.java:416` (`arrived.await`, same line as the Windows CI failures) | | this PR | 0 / 50 | 0 / 50 | | #12381 (arrival wait 60 s) | 0 / 50 | 1 / 50 failed, after the full 60 s wait | | #12454 (bodyless 503) | 0 / 50 | 0 / 50 | The base failures take 15 s, the whole wait. The #12381 failure takes the whole 60 s. So the latch is never counted down; the request is not just slow to arrive. **Forcing the failure:** I added a local-only switch that makes the server's body write throw `IOException`. That is the case this PR describes for Windows. - base: 20 / 20 failed at :416 - #12381: 20 / 20 failed at the 60 s wait - this PR: 0 / 20 failed - #12454: 0 / 20 failed (its bodyless reply never writes a body) Counting down before the write, plus the extra `countDown()` in `finally`, covers the case where headers or body cannot be written. The assertions are unchanged. This looks like the right fix to me. #12454 changes the same test differently, by making the 503 bodyless. Both work in my runs. It might be simpler to land this one and drop the PayPal part from #12454, so the two don't conflict. -- 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]
