Compressed oops failure with hs19 master

[Christian Thalinger]

On Dec 1, 2010, at 1:19 PM, Dr Andrew John Hughes wrote:
> On 12:35 Mon 29 Nov     , Tom Rodriguez wrote:
>> 
>> On Nov 29, 2010, at 11:37 AM, Dr Andrew John Hughes wrote:
>> 
>>> On 10:50 Mon 29 Nov     , Tom Rodriguez wrote:
>>>> Just FYI, I've got a test case and fix for this.  I'll be sending out a request for review soon but basically there's some code that skips unneeded zeroing for instances but it mistakenly gets applied to arrays because the array header is that same size as regular objects with compressed oops.
>>>> 
>>> 
>>> That's great to hear!  Looking forward to seeing the patch.
>>> 
>>> I guess that's why disabling the Array newInstance native worked
>>> around the bug.  It avoided accessing the array that should have been
>>> zeroed, but was skipped.
>> 
>> If it's not inlined then we don't statically know the size of the resulting array so we just fall back to the normal zeroing logic.
>> 
> 
> Ah, thanks.
> 
> Is this the fix? http://hg.openjdk.java.net/jdk7/hotspot-rt/hotspot/rev/0fc262af204f

For 7002666?  No, it's currently being reviewed:

http://cr.openjdk.java.net/~never/7002666/

-- Christian