[Soot-list] simulating effects of native calls

[Khilan Gudka]

Dear Eric,

Thanks a lot for that useful information.

The simulation really only simulates the effects of these methods in the
> "cg" phase, i.e. it adds appropriate call edges and computes
> appropriate data flow for the points-to analysis. That's it.


We infer path expressions of the form x.f.g,but points-to information would
definitely help make our analysis more precise. For example:

   x.f = y;
  // access z.f below

Without points-into information we can only conservatively assume that z and
x may be aliases so have to record the fact that y could well be accessed
below. Currently, for this example or analysis would infer: { x, y, z.f  }.
Can native methods access Java objects?

Thanks,
Khilan


When you say "Our analysis infers what objects are being accessed in a
>


given block of code." it sounds to me like you should be using
> points-to info anyway. Aren't you?
>
> Eric
>
>
> On 02/03/2008, Khilan Gudka <khilan.gudka at imperial.ac.uk> wrote:
> > Hi all,
> >
> > Ondrej pointed out to me that soot can simulate the effects of native
> > methods in jdk 1.3. I wanted to ask is this only useful for pointer
> > analysis? Our analysis infers what objects are being accessed in a given
> > block of code. Also, how does one retrieve the simulated method body of
> a
> > native?
> >
> > Thanks,
> > Khilan
> >
>
> > _______________________________________________
> >  Soot-list mailing list
> >  Soot-list at sable.mcgill.ca
> >  http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
> >
> >
>
>
>
> --
> Eric Bodden
> Sable Research Group
> McGill University, Montréal, Canada
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.CS.McGill.CA/pipermail/soot-list/attachments/20080302/72da37f0/attachment.htm