ym0506 commented on issue #39765: URL: https://github.com/apache/shardingsphere/issues/39765#issuecomment-5577162217
I added a smaller [standalone reproducer using the official Agent 5.5.3 advice classes](https://gist.github.com/ym0506/451d135eac9f2b793e1753721c3f7701). After staging the official archive and nine checksum-pinned application dependency JARs, it runs with Python 3.10+ and Java 17, without a Maven reactor, Mockito, Docker or a database. The four text files include the source, runner, dependency coordinates/download URLs/checksums and instructions. The runner verifies the archive and dependencies before compiling; it makes no network requests. Both controlled two-thread scenarios start two SQL spans but complete only one: - A succeeds: only B's span completes, with OK; A remains unfinished. - A fails: A's synthetic exception is recorded on B's ERROR span; A remains unfinished. This scenario calls A's `afterMethod` in `finally` after `onThrowing`, matching the executor lifecycle. The command deliberately exits 1 for the failed `expected=2 actual=1` assertion. It checks SQL identity, status and the exception event as well as completion count. This is direct advice invocation with synthetic metadata, not a bytecode-instrumentation or database test. The current-master regression and separate MySQL integration results in the issue remain distinct evidence. Codex assisted with this additional reproducer and its validation. -- 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]
