[lworld] RFR(S): 8220666: [lworld] Nullability information is lost at checkcast in ciTypeFlow

[Tobias Hartmann]

Found another problem when testing: In ciTypeFlow::StateVector::type_meet_internal(), we need to
unwrap the type first before handling the constant NULL cases in line 287. Otherwise, meeting
Value.val and NULL leads to Value.val instead of Value.box.

Test11 of TestNullableValueTypes triggered this. I've updated the webrev in place:
http://cr.openjdk.java.net/~thartmann/8220666/webrev.00/

Best regards,
Tobias

On 14.03.19 14:38, Tobias Hartmann wrote:
> Hi,
> 
> please review the following patch:
> https://bugs.openjdk.java.net/browse/JDK-8220666
> http://cr.openjdk.java.net/~thartmann/8220666/webrev.00/
> 
> When checkcasting from .box to .val and back .box we should keep track of the information that the
> value can never be null.
> 
> Thanks,
> Tobias
>