[aarch64-port-dev ] RFC 8169201: Montgomery multiply intrinsic should use correct name

[Andrew Haley]

On 03/21/2018 03:42 PM, Andrew Hughes wrote:
> On 7 March 2018 at 15:37, Andrew Haley <aph at redhat.com> wrote:
>> On 02/03/18 10:16, Aleksey Shipilev wrote:
>>> We have recently pulled from aarch64-port/jdk8u to shenandoah/jdk8u, and got interesting bugs.
>>>
>>> In Shenandoah code, we *do* care about stub names for analysis/verification. In upstream, there is a
>>> fix for "Montgomery multiply intrinsic should use correct name", which is missing in current
>>> aarch64-port/jdk8u:
>>>   https://bugs.openjdk.java.net/browse/JDK-8169201
>>>
>>> Per repo status:
>>>  - aarch64-port/jdk8u DOES NOT have it
> 
> This fix isn't in any version of 8u. So the first thing to do is
> request a backport on
> the jdk8u-dev list.

Right.

>>>  - aarch64-port/sheandoah-jdk8u has it from our fixes
>>>  - shenandoah/jdk8u has it from our fixes
>>>  - shenandoah/jdk9 has it from upstream
>>>  - shenandoah/jdk10 has it from upstream
>>>  - shenandoah/jdk has it from upstream
>>>
>>> This is a landmine during merges, because we can always accidentally reintroduce the bug. Maybe it
>>> makes sense to pick up JDK-8169201 to aarch64-port/jdk8u?
>>
>> Absolutely, yes.  Everything from the relevant upstream repos should go in unless
>> there is some strong reason otherwise.
> 
> No, not everything. We work to specific release tags with jdk8u so we
> match proprietary JDK releases as
> much as possible, rather than random snapshots of the jdk8u tree.

Sorry, that's what I meant by relevant: everything from the appropriate
upstream repo at the appropriate tag, whatever that is, should go in.
-- 
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671