SEZ9 commented on issue #12344:
URL: https://github.com/apache/seatunnel/issues/12344#issuecomment-5704346488

   Additional data after four more rerun rounds on both branches, and a 
methodology note that *strengthens* rather than weakens this one.
   
   The note first: GitHub carries unchanged job records forward into a new 
attempt created by `gh run rerun --failed`, so a job's presence in an attempt 
does not mean it ran there. A job only really executed if its `started_at` is 
distinct from the previous attempt's. I have since found that this inflated the 
counts in a different issue I filed (#12345), so I re-checked this one. It does 
not affect these numbers: `all-connectors-it-2` **failed on every attempt**, 
which means `--failed` re-scheduled it every time, so every attempt is a 
genuine execution. Spot-checking `started_at` on run 
[34993534806](https://github.com/SEZ9/seatunnel/actions/runs/34993534806) 
confirms it — a1 `2026-09-15T16:17:09Z`, a4 `2026-09-16T05:25:06Z`, a8 
`2026-09-16T18:40:58Z`, all distinct.
   
   Updated tally for `OpengaussCDCIT.testAddFieldWithRestore`:
   
   | Run | Base | Real executions | Failures |
   |---|---|---|---|
   | [34926657845](https://github.com/SEZ9/seatunnel/actions/runs/34926657845) 
(#12298) | `0d9f9e230` | 4 attempts × 2 JDKs = 8 | 8 |
   | [34993525332](https://github.com/SEZ9/seatunnel/actions/runs/34993525332) 
(#12298) | `75b60fa14` | 8 attempts × 2 JDKs = 16 | 16 |
   | [34993534806](https://github.com/SEZ9/seatunnel/actions/runs/34993534806) 
(#12299) | `75b60fa14` | 8 attempts × 2 JDKs = 16 | 16 |
   | | | **40** | **40** |
   
   **40 of 40**, across two unrelated pull requests, two base commits, and both 
JDKs, with the same assertion at `OpengaussCDCIT.java:476` every time and the 
other 14 tests in the class passing every time. No pass has ever been observed. 
Rerunning is not a workaround for this one.
   


-- 
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]

Reply via email to