RFR(XS): 4965252: JvmtiExport::post_raw_field_modification jni ref handling is odd

[Rickard Bäckman]

Serguei,

thank you for the review!

/R

On May 15, 2013, at 4:58 PM, serguei.spitsyn at oracle.com wrote:

> Hi Rickard,
> 
> I think, the fix is good.
> But the code that was fixed was harmless as a memory leak
> described in the report never really happened.
> It is still a nice clean up. :)
> 
> Thanks,
> Serguei
> 
> On 5/15/13 2:37 AM, Rickard Bäckman wrote:
>> Hi all,
>> 
>> can I please have this change reviewed?
>> 
>> My interpretation is that this isn't really a bug, since the parameter sig_type is never set to [.
>> The suggested change is to remove the check for [ in the if and add an assert. I also created a boolean
>> to track handle creation to simplify cleanup.
>> 
>> The caller of this method is InterpreterRuntime::post_field_modification
>> which sets the sig_type to:
>> 
>>   switch(cp_entry->flag_state()) {
>>     case btos: sig_type = 'Z'; break;
>>     case ctos: sig_type = 'C'; break;
>>     case stos: sig_type = 'S'; break;
>>     case itos: sig_type = 'I'; break;
>>     case ftos: sig_type = 'F'; break;
>>     case atos: sig_type = 'L'; break;
>>     case ltos: sig_type = 'J'; break;
>>     case dtos: sig_type = 'D'; break;
>>     default:  ShouldNotReachHere(); return;
>>   }
>> 
>> Testing done: nsk.jvmti.testlist with fastdebug build.
>> 
>> Webrev: http://cr.openjdk.java.net/~rbackman/4965252/
>> 
>> Thanks
>> /R
>> 
>> 
>> 
>