RFR: Integration from jdk/jdk

[Roman Kennke]

Am 10.04.2018 um 17:29 schrieb Aleksey Shipilev:
> On 04/10/2018 05:10 PM, Roman Kennke wrote:
>> http://cr.openjdk.java.net/~rkennke/merge-2018-04-09/shenandoah-changes.patch
> 
> The T* -> char* change in shenandoahBarrierSet.inline.hpp -- is it needed in sh/jdk10?

Should not be, but shouldn't hurt either. I changed this to deal with a
compiler failure when T=void. It would not compile sizeof(void).

I believe the change makes the code more robust too: the computed offset
will always be in bytes (and not in the T type) and makes this fact
explicit. It's a small thing.


>> I've finished running the tests and fixed a bunch of failures. It's
>> clean now. Good to go?
> 
> OK.

Thanks!

Roman