RFR(xs): 8152180: SIGFPE in CMSCollector::preclean with big CMSScheduleRemarkSamplingRatio

sangheon sangheon.kim at oracle.com
Tue Apr 5 21:47:41 UTC 2016


Thank you Jesper for reviewing this.

Sangheon


On 04/05/2016 02:46 PM, Jesper Wilhelmsson wrote:
> Looks good!
> /Jesper
>
> Den 5/4/16 kl. 23:44, skrev sangheon:
>> Hi all,
>>
>> After a discussion with Jesper, I added a parentheses for better 
>> readability.
>>
>> Updated webrev:
>> http://cr.openjdk.java.net/~sangheki/8152180/webrev.02
>>
>> Thanks,
>> Sangheon
>>
>>
>> On 04/05/2016 06:59 AM, sangheon wrote:
>>> Hi all,
>>>
>>> I reverted the change for this flag's type to 'uintx' as there's no 
>>> benefit of
>>> changing it on 32bit.
>>> This new webrev only contains the calculation rephrase.
>>>
>>> http://cr.openjdk.java.net/~sangheki/8152180/webrev.01
>>>
>>> Thanks,
>>> Sangheon
>>>
>>>
>>> On 04/02/2016 11:38 PM, sangheon wrote:
>>>> Hi all,
>>>>
>>>> Could I have a couple of reviews for this tiny change?
>>>>
>>>> There would be an arithmetic overflow with 
>>>> CMSScheduleRemarkSamplingRatio at
>>>> CMSCollector::preclean() as the flag(uintx type with max limit of 
>>>> max_uintx)
>>>> is multiplied by 100.
>>>> This change is simply rephrased the calculation routine to avoid 
>>>> the overflow
>>>> and changed the flag's type to 'uint' because 'uintx' seems too 
>>>> large for the
>>>> flag.
>>>>
>>>> CR: https://bugs.openjdk.java.net/browse/JDK-8152180
>>>> Webrev: http://cr.openjdk.java.net/~sangheki/8152180/webrev.00
>>>> Testing: JPRT, runtime/commandline JTREG for all platforms
>>>>
>>>> Thanks,
>>>> Sangheon
>>>
>>



More information about the hotspot-gc-dev mailing list