RFR: 8224861: Rework restrictions introduced by 8223029 for nullable arrays

[Tobias Hartmann]

Hi Roland,

I've executed some tests with webrev.01 and I'm hitting this intermittent issue when running
TestNullableArrays with "-XX:+EnableValhalla -XX:+UseParallelGC -XX:+UseParallelOldGC -XX:+UseNUMA
-XX:+IgnoreUnrecognizedVMOptions":

#  Internal Error
(/scratch/mesos/slaves/07fc96ef-bf4d-487f-b22f-a84e49f5f44a-S134314/frameworks/1735e8a2-a1db-478c-8104-60c8b0af87dd-0196/executors/ec10949b-d85c-45d4-a2bb-fd2b68f82d67/runs/323d5a78-5594-4d2e-8070-48d1eacc7efa/workspace/open/src/hotspot/share/oops/klass.inline.hpp:63),
pid=5508, tid=3335
#  assert(!is_null(v)) failed: narrow klass value can never be zero

Thanks,
Tobias

On 28.05.19 16:59, Tobias Hartmann wrote:
> Hi Roland,
> 
> this looks good to me and is much cleaner now.
> 
> Some minor comments:
> - In macroArrayCopy.cpp, you can move the 'array_klass' declarations in the else branches or even
> into the assert statements-
> - The indentation in TestNullableArrays.java:514 is wrong.
> - In type.cpp, please add comments to line 2370 and 3459 explaining that we don't set 'exact'
> because the runtime type might be [V due to [V <: [V?
> 
> No new webrev required.
> 
> Thanks,
> Tobias
> 
> On 28.05.19 15:22, Roland Westrelin wrote:
>>
>> New webrev:
>>
>> http://cr.openjdk.java.net/~roland/8224861/webrev.01/
>>
>> A couple conditions were not quite right in the previous one.
>>
>> Roland.
>>