RFR (S) 8164692: InstanceKlass::_previous_version_count goes negative

[Coleen Phillimore]

Thanks!
Coleen

On 8/30/16 5:39 PM, dean.long at oracle.com wrote:
> Looks good to me.
>
> dl
>
> On 8/30/16 2:32 PM, Daniel D. Daugherty wrote:
>> On 8/30/16 2:49 PM, Coleen Phillimore wrote:
>>> Summary: decrement previous_version_count when it's removed from the 
>>> list.
>>>
>>> The previous_version_count was decremented in the wrong place. This 
>>> contributed to the bug that Dean's just sent out the RFR and the fix 
>>> is essentially contributed by Dean.
>>>
>>> Ran jdk/test/java/lang/instrument, and internal closed (colocated 
>>> and non-colocated) redefinition tests.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8164692.01/webrev
>>
>> src/share/vm/classfile/classLoaderData.cpp
>>     L965: // and only if class redefinition and if there are previous 
>> versions of
>>         Perhaps:
>>
>>         // and only if class redefinition occurred and if there are 
>> previous versions of
>>
>> src/share/vm/oops/instanceKlass.cpp
>>     No comments.
>>
>> src/share/vm/oops/instanceKlass.hpp
>>     No comments.
>>
>> test/runtime/RedefineTests/RedefineCount.java
>>     L48: // previous_version_list and the count should stay zero, but 
>> not go negative
>>
>>         Perhaps:
>>
>>             "should stay zero and not go negative"
>>
>>         for that last part.
>>
>> Does the new test (RedefineCount.java) reproduce the crash that Dean L
>> is chasing in https://bugs.openjdk.java.net/browse/JDK-8156137?
>>
>> Thumbs up!
>>
>> Dan
>>
>>
>>> bug link https://bugs.openjdk.java.net/browse/JDK-8164692
>>>
>>>
>>> Thanks,
>>> Coleen
>>
>