Orchestration fails if single build job cancelled
|Assignee:||Jan Moringen||% Done:|
Cancelling a single job stops the whole orchestration because the CancellationException that is thrown when done.get() is called, is not caught.
The attached patch provides a dirty workaround, ignoring cancelled jobs completely. However this way:
- other potential exceptions thrown by done.get() are not caught yet
- the failure is not reported back to the main spinning loop
Probably, Jan, you have a better idea how to fix the problem.