[Soot-list] How to invalidate spark result?

[Ondrej Lhotak]

On Thu, Oct 14, 2004 at 03:36:04PM -0400, Chen Fu wrote:
> I reset the following fields in Singletons and G between two Spark,
> and it seems work now. maybe some of them are not necessary.
> 
> instanceSparkTransformer = null;
> instanceFieldTagger = null;
> instanceFullObjectSet = null;
> instanceFieldReadTagAggregator = null;
> instanceFieldWriteTagAggregator = null;
> instanceFieldTagAggregator = null;
> Parm_pairToElement.clear();
> MethodPAG_methodToPag.clear();

Thanks for the information, Chen. I think that really only the last two
are necessary. When I get a chance, I'll try to fix it so that they are
not necessary either.

Ondrej

> 
> 
> On Tue, 12 Oct 2004 22:39:10 -0400, Ondrej Lhotak
> <olhotak at sable.mcgill.ca> wrote:
> > On Tue, Oct 12, 2004 at 10:15:41PM -0400, Chen Fu wrote:
> > > Is there anyway to invalidate the points-to analysis result and rerun
> > > spark all over again. Suppose I defined two phase in the transformer,
> > > the first phase need points-to information so it's following a spark
> > > phase. But my first phase changed the code and then again my second
> > > phase also need points-to information. But the result of spark is not
> > > longer valid because of the modification in my first phase.
> > >
> > > I've tried to call
> > > Scene.v().releasePointsToAnalysis();
> > > then
> > > Sparktranformer.v().transform(...).
> > 
> > This is the correct way to do it. However, I don't normally run it
> > twice like this, so there may well be problems with it. I had a quick
> > look through the code looking for state that might be retained, but
> > didn't notice any. One thing that may be causing the problem is that
> > Spark will use a call graph from the Scene if there is one, so it may
> > be getting the out-of-date call graph from the first run. So, calling
> > releaseCallGraph on the Scene might help.
> > 
> > > But the second phase finished in no time and the result seems not right.
> > 
> > You don't say exactly how much time is no time, but it's not all that
> > unusual that it would run in a few seconds the second time. The reason
> > is that Spark itself is very quick; what takes most of the time is
> > reading in and converting to Jimple all the class files.
> > 
> > Ondrej
> > 
> > > What's the correct way to do it?
> > >
> > > Thanks!
> > > _______________________________________________
> > > Soot-list mailing list
> > > Soot-list at sable.mcgill.ca
> > > http://www.sable.mcgill.ca/mailman/listinfo/soot-list
> > >
> >
> _______________________________________________
> Soot-list mailing list
> Soot-list at sable.mcgill.ca
> http://www.sable.mcgill.ca/mailman/listinfo/soot-list
>