Joe wrote:

> For example, here are some real life results:
> - mvn clean from root
> - remove aries artifacts from maven local repo to ensure a clean 
> starting point (rm -rf ./m2/repository/org/apache/aries )
> - mvn install from root
> - failures in application tests - IsolatedRuntimeTest, UpdateAppTest, 
> OBRResolverTest, and OBRResolverAdvancedTest
[...]

This is very similar to what I'm seeing. Failures every single time, but 
it feels like the same test never fails twice. 

I believe that the quiesce tests in particular do have some concrete 
timing criteria - "has everything managed to quiesce before the timeout?" 
and so on. 

The fact that they work reliably when run individually makes me wonder if 
the performance of the build gets slower as it progresses through the 
child components, and that means some tests are running slowly enough to 
hit timing glitches. This could be heap fragmentation - in which case 
running with a bigger heap could help. It could be that the Hudson builds 
are running with a larger heap, or it could be that the absence of other 
processes on the machine means the memory and disk caches are staying 
clean enough that the performance is ok?

I wonder if running with a bigger heap would help?

Holly








Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 
741598. 
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU





Reply via email to