RFR(S): 8142386: Octane crashes with assert(is_Load())

[Roland Westrelin]

> Looks good.

Thanks for the review, Vladimir.

> Regarding Unsafe usage in the test, in order to avoid boilerplate code to access the instance, you could either run the test from bootclasspath (@run main/bootclasspath) or use jdk.test.lib.Utils.getUnsafe() [1].
> 

I changed the test so it uses jdk.test.lib.Utils.getUnsafe(). Thanks for the suggestion.

Roland.

> Best regards,
> Vladimir Ivanov
> 
> [1] http://hg.openjdk.java.net/jdk9/hs-comp/hotspot/file/tip/test/testlibrary/jdk/test/lib/Utils.java#l353
> 
> On 11/16/15 12:28 PM, Roland Westrelin wrote:
>> Thanks for the review, Tobias.
>> 
>>> your fix looks good to me but I think you should add a comment to
>>> 
>>> 2541     if (bt == T_ARRAY) {
>>> 2542       bt = T_OBJECT;
>>> 2543     }
>> 
>> 
>> I’ll do that. Can I get another review for that change?
>> 
>> Roland.
>> 
>> 
>>> On 10.11.2015 15:24, Roland Westrelin wrote:
>>>> http://cr.openjdk.java.net/~roland/8142386/webrev.00/
>>>> 
>>>> There are 2 problems with my recent change to handle mismatched unsafe accesses:
>>>> 
>>>> - If the access is compiled as a LoadN+DecodeN then GraphKit::make_load() doesn’t get a load node from LoadNode::make() and can’t set the mismatched flag
>>>> 
>>>> - array fields are of type T_ARRAY which differs from T_OBJECT and so are treated as mismatched
>>>> 
>>>> Roland.
>>>> 
>>