RFR: 7154113: jcmd, jps and jstat tests failing when there are unknown Java processes on the system

[Alan Bateman]

On 19/03/2012 07:34, Staffan Larsen wrote:
> On 17 mar 2012, at 23:47, David Holmes wrote:
>> On 16/03/2012 6:10 PM, Staffan Larsen wrote:
>>> Fixes have been verified in JPRT.
>> How have the fixes been "verified" in JPRT? We would expect these tests to pass in JPRT, where there should be no unknown Java processes. So how have we verified that they work in the presence of unknown Java processes?
> I meant "verified" in the sense that they do not break any existing tests in JPRT. The added functionality was verified manually on my machine.
>
> The tests initially failed in JPRT because of unknown Java processes. This was on mac osx and I think the process was a Firefox running a Java plugin.
>
> /Staffan
I haven't reviewed the fix (just don't have cycles just now) but just to 
say that I've seen these tests failing on several platforms because of 
the same issue. Even on machines that are just running do test runs 
sequentially it is possible for something to go wrong that causes a java 
process to remain on the system and this is visible to tests that use 
jps or other jvmstat tools.

-Alan