erma-core team mailing list archive

[Doug Barth <dougbarth@xxxxxxxxx>]

On Sep 16, 2008, at 9:10 PM, Ray Krueger wrote:
> Anyone have any hints where the usual suspects are in that sort of
> failure? For instance, when running the tests in IDEA the
> MonitoringEngineTest fails consistently, unless I run it alone, then
> it passes fine.

Any test that depends on the behavior of a running MonitoringEngine  
should have "MonitoringEngine.getInstance().restart();" in its setup  
method. That should (assuming there aren't any bugs) reset the  
MonitoringEngine to a valid state and solve any inter test dependencies.

The one potential issue I did run into previously was the dependence  
on system properties. There is no way to clear a system property, so  
some tests were expecting things like the VMID to not be setup, but it  
was set if run in a different order. It looks like that logic was  
moved out of the MonitoringEngine, which is good. However, looks like  
that logic is completely gone from the codebase.

Greg/Stephen, I thought we had decided to keep the initialization of  
Monitor.HOSTNAME and Monitor.VMID in the codebase.
--
Doug Barth