RFR 8014024: NPG: keep compiled ic methods from being deallocated in redefine classes

Dean Long dean.long at oracle.com
Tue May 7 09:42:21 PDT 2013


Sounds good (reviewed).

dl

On 5/7/2013 6:58 AM, Coleen Phillimore wrote:
>
> On 05/06/2013 10:04 PM, Dean Long wrote:
>> Why do you only handle methods for the !is_icholder_call() case? 
>> Isn't that assuming that the only caller is RedefineClasses,
>> that f is mark_on_stack, and that it's a no-op for klass metadata?
>
> I do make that assumption which isn't cool.  There's code that asserts 
> it's live that also calls this code.  I'll change it to call the 
> function on both types of metadata and rerun the testing with the 
> verification code.
>
> thanks,
> Coleen
>
>>
>> dl
>>
>> On 5/6/2013 5:09 PM, Coleen Phillimore wrote:
>>> Summary: Walk the compiledIC relocation records to keep Method* from 
>>> being deallocated.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8014024/
>>> bug link at http://bugs.sun.com/view_bug.do?bug_id=8014024
>>>
>>> Found when fixing 8013063.   Tested same as 8013063.
>>>
>>> Coleen
>>
>



More information about the hotspot-runtime-dev mailing list