I believe we have found a possible cause for this - and a fix - however it’s going to take a while to figure out a test scenario for this - we need to be able to reproduce it so we can confirm the fix.
We have not been able to reproduce this here, however we have been testing a potential fix for this - so far we have not seen any side effects. The fix will be in the next update (next week).
I think we figured out what the problem is. Last night for 20 builds, 15 failed with this error so we started to investigate. It seems the culprit is the import of junit to Continua.
Before we use Continua action to load junit (which a blocking action), but now we use custom message in python script [@@continuacontinuacontinuacontinua[importUnitTestResults]. Before it took a long time (even couple of minutes) as we have around 20k of testcases. Right now from script perspective it takes 0s, as it’s only “logged”, so it moves on, but probably a thread is created for the import. Then when the stage ends (last action) I think you are terminating any still existing threads and that’s where it crashes.
P.S. On closer look, it crashes on ending the same action not stage. So python script executes tests, logs this custom messages, then exits, this crashes the thread.
After further investigation, it looks like this is fixed by the changes we made a few weeks ago. We did some more work on this today, adding some more unit testing around the offending code.
The issue was that the I/O threads that handle process output were timing out before the process was done - that was causing the ThreadAborted exception.