VTANativeAdjuster (Was: resolveConcreteDispatch() and resolveAbstractDispatch()_

["Stephen Cheng" <stephen.cheng@btinternet.com>]

> Of course, the NativeAdjuster design is not ideal.    However, it
> is fairly
> tricky to come up with something more portable,  and whenever new
> native methods
> are added, someone has to figure out their impact and somehow convey
> this information to VTA or other analyses that need to know what fields
> might be written and what objects might be returned.   It was already
> quite painful to build the one version of the NativeAdjuster.

As I see it the VTA should not attempt to analyse the internals of a
library, as libraries can independently change. Surely this will mean a
reduction in code optimisation, but the resulting optimised code will longer
be dependent on a particular version of library. This will eliminate the
necessity of VTANativeAdjuster.

Regards,
Stephen