RFR 8059034: ProcessTools.startProcess() might leak processes

[Jaroslav Bachorik]

On 09/25/2014 12:13 PM, Staffan Larsen wrote:
> I wonder if the p.waitFor() is needed? What if the process launching expired with a timeout and now we are still waiting for the process to end - wouldn’t that kind of defeat the timeout? In any case, the destroyForcibly() should end the process whether we wait for it or not.

It would be wonderful but the javadoc states that the result of 
destroyForcibly() call depends on the implementation and may actually 
not force close the process and one should use waitFor() to make sure 
that the process has in fact died.

I wonder whether JTReg kills the process tree on timeout - in case it 
does using waitFor() would guarantee that there would be no zombies 
left. Without using waitFor() and semantics of destroyForcibly() there 
might be situations when non-functional stuck processes are left behind 
(not sure how probable, however).

-JB-

>
> /Staffan
>
>
> On 25 sep 2014, at 11:54, Jaroslav Bachorik <jaroslav.bachorik at oracle.com> wrote:
>
>> Please, review the following change to the JDK test library class
>>
>> Issue : https://bugs.openjdk.java.net/browse/JDK-8059034
>> Webrev: http://cr.openjdk.java.net/~jbachorik/8059034/webrev.00
>>
>> Currently, the ProcessTools.startProcess() might leave a dangling process behind when a timeout or interrupt happens. The solution is to try and forcibly terminate the forked process when this happens.
>>
>> Thanks,
>>
>> -JB-
>