RFR: 8081682: AbstractWorkGang::_terminate is never used
Kim Barrett
kim.barrett at oracle.com
Wed Jun 3 16:46:11 UTC 2015
On Jun 3, 2015, at 5:31 AM, Stefan Karlsson <stefan.karlsson at oracle.com> wrote:
>
>
>
> On 2015-06-03 10:07, Stefan Karlsson wrote:
>>
>>
>> On 2015-06-03 01:06, Kim Barrett wrote:
>>> On Jun 2, 2015, at 9:51 AM, Stefan Karlsson <stefan.karlsson at oracle.com> wrote:
>>>> Hi all,
>>>>
>>>> Please review this patch to remove the AbstractWorkGang::_terminate variable. The work gangs are never deleted and therefore the code which sets _terminate is never executed.
>>>>
>>>> http://cr.openjdk.java.net/~stefank/8081682/webrev.00/
>>>> https://bugs.openjdk.java.net/browse/JDK-8081682
>>>>
>>>> Thanks,
>>>> StefanK
>>> I assume the idea here is that, rather than having a (presently
>>> unused) generic early stop request mechanism provided by the workgroup
>>> framework, any workgroup that needs an early stop mechanism should
>>> roll its own.
>>
>> Yes, for the time being. If we later find a compelling reason to add a used, tested stop mechanism to the workgroup then we should do it.
>>
>>
>>> Are there any that already do so,
>>
>> One example is CMConcurrentMarkingTask.
>>
>>> and could have been
>>> using the generic mechanism?
>>
>> I don't know.
>
> I realize that we might be talking about different stop mechanisms.
> 1) A stop mechanism to tear down and exit the worker threads.
> 2) A stop mechanism to abort the current executing task.
>
> I'm removing the code for (1), while the CMConcurrentMarkingTask has its own mechanism for (2).
>
> I discussed this briefly with Per, and we probably want to reimplement (1) and tear down all the worker threads when DestroyJavaVM is called. However, that will be handled as a separate RFE.
That all sounds fine to me.
>
> StefanK
>
>>
>>>
>>> The changes look ok.
>>>
>>
>> Thanks,
>> StefanK
More information about the hotspot-gc-dev
mailing list