[foreign] RFR: consolidate native vs. carrier conversions

[Sundararajan Athijegannathan]

Looks good

PS. Verified that all tests run fine on Mac

-Sundar

On 18/05/18, 7:25 PM, Maurizio Cimadamore wrote:
> Hi,
> as noted yesterday, we have several paths which perform native 2 java 
> conversion. The goal of this patch is to reduce the number of such 
> paths, and use LayoutType wherever possible. This resulted in removal 
> of a lot of duplicated code, in both UpcallHandler and NativeInvoker, 
> which I think is good.
>
> Given to the way we handle structs in native calls - that is spread 
> struct fields into registers, using LayoutType here is not possible as 
> structs have to be marshalled/unmarshalled bit by bit, according to 
> the ABI.
>
> In the process I realized that the binder was implicitly performing a 
> conversion between Java null and Pointer.nullPointer(). I removed that 
> for now, as I prefer things to err on the explicit-side of things.
>
> Webrev:
>
> http://cr.openjdk.java.net/~mcimadamore/panama/consolidate-load-store/
>
> P.S.
>
> This could probably use some testing in MacOS as well, since it 
> touches core parts of the binder, close to the system ABI.
>
> Cheers
> Maurizio
>
>