[foreign-abi] RFR 8235258: Add support for empty segments and rebased addresses

[Jorn Vernee]

Hi Maurizio,

Changes look good to me.

All tests pass on my (Windows) machine.

Jorn

On 03/12/2019 15:51, Maurizio Cimadamore wrote:
> This patch implements the approach discussed in [1, 2] - which aims at 
> increasing the range of 'safe' operations provided by the memory 
> access API.
>
> Webrev:
>
> http://cr.openjdk.java.net/~mcimadamore/panama/8235258_v2/
>
> Changes compared to the patch in [2]:
>
> * added support for all ABIs (please double check :-))
> * refactored some instance methods in MemoryAddressImpl as static 
> methods in Utils (as per Jorn suggestion)
> * removed bound check from MemoryAddress::rebase (as per Jorn suggestion)
> * added a combinatorial test which checks several cases of rebase(), 
> more specifically:
>   - for overlapping segments, it tests that accessing elements in the 
> smaller segment works ok, regardless of whether the address used is 
> rebased or not
>   - for non-overlapping segments, if the segments have different kinds 
> (e.g. heap vs. off-heap) check that an exception is thrown when rebase 
> is called
>   - for non-overlapping segments of the same kind check that an 
> exception is thrown when the segment is accessed using the rebased 
> address (as that should be OOB)
>
> Maurizio
>
>
> [1] - 
> https://mail.openjdk.java.net/pipermail/panama-dev/2019-November/006746.html 
>
> [2] - 
> https://mail.openjdk.java.net/pipermail/panama-dev/2019-December/006780.html
>
>