The test verifies time order of historic tasks. Since the resolution is ms, the order my not predictable if the engine is faster to complete the tasks.
On my computer I inserted a calculation to show the distance between to tasks: 3 ms.
Stephan Itzenplitz could not compile and test camunda bpmn, his computer is probably a little bit faster. He did all the analyze work to locate the problem. Compilation within Ecipse slows the process so that there have been no problems.
Please have a look at the following test case:
If you run it with the Thread.sleep call, it is ensured, that there will be the required delay on milli second level to get it all time working.
Please think about concepquences.
- Is the resolution at the API sufficient?
- are there probably further tests affected?