RFR[S]: 8239014 -XX:-UseEmptySlotsInSupers sometime fails to reproduce the layout of the old code

[Frederic Parain]

Thank you David.

Fred


> On Feb 18, 2020, at 22:41, David Holmes <david.holmes at oracle.com> wrote:
> 
> Hi Fred,
> 
> Sorry for the delay getting back to this.
> 
> On 15/02/2020 4:34 am, Frederic Parain wrote:
>> Hi David,
>> Here’s a new webrev with a test case:
>> http://cr.openjdk.java.net/~fparain/8239014/webrev.01/
>> When -XX:-UseEmptySlotsInSupers is used, the field layout code
>> emulates the behavior of the old code by appending fields after
>> the rounded up end of the super class, adding primitive fields
>> by decreasing size, then oops.
>> The issue was that this strategy doesn’t fill the gap after the
>> header when the class pointer is 4 bytes and the first field is
>> either a long or a double. There was a special case in the old
>> code to handle this situation, but it was missing in the
>> emulation mode of the new code.
> 
> Thanks for clarifying.
> 
> Changes look good.
> 
> David
> 
>> Thank you,
>> Fred
>>> On Feb 13, 2020, at 17:53, David Holmes <david.holmes at oracle.com> wrote:
>>> 
>>> Hi Fred,
>>> 
>>> Can you add a testcase for this? I can't really evaluate the change just by looking at the code.
>>> 
>>> Thanks,
>>> David
>>> 
>>> On 14/02/2020 6:28 am, Frederic Parain wrote:
>>>> Please review this small change that fix the behavior of-XX: -UseEmptySlotsInSupers.
>>>> CR: https://bugs.openjdk.java.net/browse/JDK-8239014
>>>> Webrev: http://cr.openjdk.java.net/~fparain/8239014/webrev.00/index.html
>>>> Thank you,
>>>> Fred