[Soot-list] Empty points-to information

[Eric Bodden]

Hi.

> How safe is it to assume that if SPARK returns an empty points-to set for a
> local or field then it is actually correct? Does it take into consideration
> static initialisers?

Yes it does.

> I'm implementing some optimisations that would take
> advantage of this but wasn't sure about the soundness of it. I'm analysing
> against GNU Classpath 0.97.2p10.

It's as sound as your call graph is. If you can make sure that your
call graph is complete then you should be able to rely on the
invariant "if there is an empty points-to set for v then v will never
be assigned any value".

Eric