The problem is in the test case. It is written to start a CPM pipeline, run a bunch of docs thru it, and while processing, it calls (async) "stop". It does this by waiting 500 milliseconds before calling stop. If you run this on a fast CPU, then in some cases, the CPM pipeline actually finishes, before the 500 milliseconds. In this case, the callback routine is called with the "finished" callback, not the "abort" callback. The testcase is looking only for the abort callback, and since it doesn't get this, it just waits forever.
The fix is to adjust the test case to issue the "stop" sooner, perhaps also checking to see if it's getting the finished call back (which would indicate the stop wasn't being done soon enough). I confirmed this by putting in various little tracings in the code (temporarily), and also trying to adjust the delay till the stop and seeing that had the expected behavior. -Marshall On 11/30/2011 3:21 PM, Marshall Schor wrote: > In doing the build, for the first time I've seen it, the cpm test hung in the > CpmInitTest. This is an intermittent failure. I am able to reproduce by > running "mvn test" in a loop... > > I'm going to see if I can debug this - so for now the release is stopped. > > -Marshall > >
