[9] RFR (XXS): 8158546: C1 compilation fails with "Constant field loads are folded during parsing"

[Vladimir Ivanov]

Thanks, Vladimir.

Best regards,
Vladimir Ivanov

On 1/28/17 1:23 AM, Vladimir Kozlov wrote:
> Okay.
>
> thanks,
> Vladimir
>
> On 1/27/17 9:02 AM, Vladimir Ivanov wrote:
>> http://cr.openjdk.java.net/~vlivanov/8158546/webrev.00/
>> https://bugs.openjdk.java.net/browse/JDK-8158546
>>
>> The problematic assert checks that constant folding of field loads
>> happens only during parsing. There are some configurations through
>> when it's not the case: PatchALot, ScavengeRootsInCode < 2.
>>
>> But there's one case missing (for static fields): when holder class
>> can be in process of initialization during parsing, but it finishes by
>> the time canonicalizer kicks in.
>>
>> I asked to add the assert as part of JDK-8150102 [1] to raise our
>> awareness when LoadField case is needed in
>> Canonicalizer::do_ArrayLength(). IMO the assert did its job and it's
>> time to remove it.
>>
>> Thanks!
>>
>> Best regards,
>> Vladimir Ivanov
>>
>> [1] https://bugs.openjdk.java.net/browse/JDK-8150102