RFR: 8238494: Odd behavior when using VarHandle::toMethodHandle on a memory access handle

[Paul Sandoz]

Opps simple to do and hard to spot.  

Is it possible to add a cast or assert somewhere checking that the bound VH is not an instance of IndirectVarhandle? 

Paul.

> On Feb 4, 2020, at 2:57 PM, Maurizio Cimadamore <mcimadamore at openjdk.java.net> wrote:
> 
> Hi,
> this change fixes a bug introduced by the recently pushed VarHandle adaptation support. More specifically, when calling `VarHandle::toMethodHandle` on a memory access handle, some spurious index out of bound exception is generated. This is caused by the fact that when `IndirectVarHandle::toMethodHandle` generates the required method handle, it binds it against `this`, rather than against the leaf direct var handle. Which means we're calling a method handle which expects a certain kind of a VarHandle (a memory access handle) with a different VarHandle which has no associated offset information. In fact, the memory access var handle helper classes will try to read the `offset` field from a class that doesn't have it - resulting in a spurious read.
> 
> I've updated the adapter test to do the final get in the various tests through a method handle obtained by the corresponding adapted var handle.
> 
> Cheers
> Maurizio
> 
> -------------
> 
> Commits:
> - c06b8c46: Fix for OOB thrown when accessing memory access VarHandle using MH getter
> 
> Changes: https://git.openjdk.java.net/panama-foreign/pull/4/files
> Webrev: https://webrevs.openjdk.java.net/panama-foreign/4/webrev.00
>  Issue: https://bugs.openjdk.java.net/browse/JDK-8238494
>  Stats: 12 lines in 2 files changed: 0 ins; 0 del; 12 mod
>  Patch: https://git.openjdk.java.net/panama-foreign/pull/4.diff
>  Fetch: git fetch https://git.openjdk.java.net/panama-foreign pull/4/head:pull/4
> 
> PR: https://git.openjdk.java.net/panama-foreign/pull/4