RFR (S) 8057570: RedefineClasses() tests fail assert(((Metadata*)obj)->is_valid()) failed: obj ,is valid

Coleen Phillimore coleen.phillimore at oracle.com
Fri Sep 5 12:11:45 UTC 2014


Thank you Dan.

On 9/4/14, 10:57 PM, Daniel D. Daugherty wrote:
> On 9/4/14 6:44 PM, Coleen Phillimore wrote:
>> Summary: Revert two small changes from the previous-versions cleanup 
>> with marking code cache.
>>
>> These weren't really related to the cleanup, but I thought they'd 
>> make it more efficient.  One caused the crash and the other I'm just 
>> being paranoid about.  Remind me to stay away from deoptimization.
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/8057570/
>
> src/share/vm/code/nmethod.cpp
>     This looks like a proper reversion for lines 2091-2 to what
>     they were before the fix for:
>
>     8055008: Clean up code that saves the previous versions of 
> redefined classes
>
>     I presume that leaving the change on this line is ok:
>
>     -          f(md);
>     +          if (md != _method) f(md);

It's done unconditionally now in the end so I don't see how it could go 
wrong.

>
> src/share/vm/memory/universe.cpp
>     This looks like a proper reversion for lines 1259-60 to what
>     they were before the fix for:
>
>     8055008: Clean up code that saves the previous versions of 
> redefined classes
>
>
> Thumbs up.
>
> The bug needs an evaluation when you get the chance.

I just added an evaluation.

Thanks for the quick code review (and heads up about this bug!)

Coleen
>
> Dan
>
>
>
>> bug link https://bugs.openjdk.java.net/browse/JDK-8057570
>>
>> Tested with failing test case in loop on machine where it failed 
>> every time.
>>
>> Coleen
>



More information about the hotspot-runtime-dev mailing list