Please review fix for 7196681: NPG: Some JSR 292 tests crash in Windows exception handler

[Vladimir Kozlov]

Can you use dummy/empty verify_method() and verify_klass() methods with TODO 
comment for this fix? It would be nice to not search for places where to add it 
in future for RFE fix.

Thanks,
Vladimir

Coleen Phillimore wrote:
> 
> I would like to have this but it's not as easy to do as verify_oop().  
> The Metaspace::contains() function does range checks for all the virtual 
> spaces that have been mmap'ed but this is not just one range like the oops.
> 
> Maybe we should have verify_method() do something like verify *ptr == 
> the vtable for Method, similarly for Klass.   I will file an RFE for this.
> Thanks,
> Coleen
> 
> On 9/11/2012 6:11 PM, Vladimir Kozlov wrote:
>> Coleen,
>>
>> Should we have verify_metadata() instead?
>>
>> Vladimir
>>
>> Coleen Phillimore wrote:
>>> Summary: There was a rogue os::breakpoint() call in log_dependency 
>>> left over from the jsr292 merge.  Also removed verify_oop() calls for 
>>> metadata.
>>>
>>> Tested CallSiteTest on windows 64 where this os::breakpoint() 
>>> actually sends a breakpoint signal.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/7196681/
>>> bug link at http://bugs.sun.com/view_bug.do?bug_id=7196681
>>>
>>> Thanks,
>>> Coleen
>>>
>>>