RFR(S): 8016131: nsk/sysdict/vm/stress/chain tests crash the VM in 'entry_frame_is_first()'

[Vladimir Kozlov]

On 7/15/13 11:59 PM, Rickard Bäckman wrote:
> Vladimir,
>
> I'll use the stack_base() method instead.

Good.

Thanks,
Vladimir

> Thanks
> /R
>
> On Jul 16, 2013, at 8:56 AM, Vladimir Kozlov wrote:
>
>> Rickard,
>>
>> One thing is_in_stack() has is check _stack_base == NULL. In your case it may not happen then you need to add assert to make sure it really does not happen. Or use stack_base() method which has the assert.
>>
>> Thanks,
>> Vladimir
>>
>> On 7/15/13 10:11 PM, Rickard Bäckman wrote:
>>> Vladimir & John,
>>>
>>> thanks for the suggestion. It makes sense.
>>> Updated webrev: http://cr.openjdk.java.net/~rbackman/8016131.u2/
>>>
>>> Thanks
>>> /R
>>>
>>> On Jul 16, 2013, at 4:26 AM, John Rose wrote:
>>>
>>>> On Jul 15, 2013, at 9:59 AM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>>>>
>>>>> There are several methods already in Thread class which do similar address cheacks: on_local_stackO, is_in_stack(). It would be nice to have new check (at least part of it) at the same place.
>>>>
>>>> Yes, that's a good idea!  Thread::is_in_usable_stack.  That way the computation that involves stack_guard_size, etc., goes in thread.cpp which is more logical than in frame.cpp.
>>>>
>>>> — John
>>>
>