[aarch64-port-dev ] RFR: Fix post-CPU merge Shenandoah regressions

[Aleksey Shipilev]

On 01/25/2018 07:00 PM, Andrew Hughes wrote:
> On 25 January 2018 at 11:21, Aleksey Shipilev <shade at redhat.com> wrote:
>> http://cr.openjdk.java.net/~shade/shenandoah/aarch64-fix-1/webrev.01/
>>
>> This unbreaks Shenandoah after recent post-CPU merges. The patch goes to aarch64-port/shenandoah-8u.
>> We will retest after 8u161 update lands, but that update is not expected to cause any more troubles.
>>
>> Testing: hotspot_gc_shenandoah, specjvm with aggressive Shenandoah verification
>>
>> Thanks,
>> -Aleksey
>>
> 
> This looks good to me, based on what you said yesterday.
> 
> Sorry about the methodHandles.cpp regression; I seem to remember
> looking at upstream jdk8u and
> shenandoah OpenJDK 9 to work out which one was right; I guess I made
> the wrong choice :(

No problem, you did the best you could under the circumstances! I am actually impressed we haven't
broken many other things, considering the whole mess around this CPU.

Actually, we should just handle it better during the CPU. Shenandoah frequently fails like that,
especially after the bulk upstream merges, and this is why we have tests that capture these simple
omissions automatically. Just running hotspot_gc_shenandoah captures 99.5% of those, SPECjvm
captures around 0.4%, and the rest escapes for us to a few weeks later :)

-Aleksey