[Soot-list] Cycle detection in pointer analysis

[Ondrej Lhotak]

On Thu, Nov 15, 2007 at 01:44:57PM -0600, Dimitrios Prountzos wrote:
> Hi,
>
> I was looking at the code that propagates points-to sets
> using on-line cycle detection, based on the Heintze and
> Tardieu algorithm (spark/solver/PropCycle.java) and I
> got the impression that in the current implementation
> there is no attempt to collapse cycles. In particular,
> in the computeP2Set() method, the only statement that
> tries to merge nodes is in comments.
> If this is indeed the case, would it be enough to just
> remove the comment from the specific statement, or does
> it require more work to support cycle elimination?

That file contains only a partial, unfinished implementation of Heintze 
and Tardieu's algorithm. I started implementing it at one point, but got
sidetracked by other things before finishing it. I don't remember how
far it is from being finished, but there is probably still substantial
work left to be done.

Ondrej

> Thank you,
>
> Dimitris
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>