Request for Review (xs) - 8155966: Assertion failures when -XX:+UseParallelGC -XX:ParallelGCThreads=1
Jon Masamitsu
jon.masamitsu at oracle.com
Tue May 3 23:28:27 UTC 2016
On 5/3/2016 4:19 PM, Kim Barrett wrote:
>> On May 3, 2016, at 6:51 PM, Jon Masamitsu <jon.masamitsu at oracle.com> wrote:
>>
>> 8155966: Assertion failures when -XX:+UseParallelGC -XX:ParallelGCThreads=1
>> https://bugs.openjdk.java.net/browse/JDK-8155966
>>
>> The change for 8150994 change the stealing task to the main task for compaction.
>> Mistakenly left in the stealing task definition was code to skip the stealing task
>> if there was only 1 GC thread (no other GC thread from which to steal).
>>
>> The code changes are an improved assertion message and removal of the code
>> from the stealing tasks for the special case of 1 GC worker.
>>
>>
>> http://cr.openjdk.java.net/~jmasa/8155966/webrev.00/
>>
>> Failure was reproducible with GCOld and ran GCOld with
>> ParallelGCThreads=1 and default number of GC threads for
>> testing.
>>
>> Thanks.
>>
>> Jon
> Looks good.
Thanks for the review.
>
> Something to consider for the future, the class StealRegionCompactionTask seems somewhat poorly named after 8150994.
Yes, I created https://bugs.openjdk.java.net/browse/JDK-8155992 to take
care of that.
Jon
>
More information about the hotspot-gc-dev
mailing list