Re: newmultiarrays in Spark

[Manoj Plakal <plakal@cs.wisc.edu>]

Thanks for the quick fix, that worked.

A more general question: is there any
validation of the basic pointer
analysis results? How do you test that
Spark produces safe points-to sets? 

Apart from having a manual test suite,
maybe it would be good to have a
transformation phase (not just tagging) 
driven by the points-to information.
Bugs might get flushed out quicker.

Manoj


Ondrej Lhotak wrote (Mon, Jul 14, 2003 at 03:36:02PM -0400) :
> The following one-line fix should fix the bug. It will be included in
> future releases of Soot.
> 
> Ondrej
> 
> Index: src/soot/jimple/spark/builder/StandardParms.java
> ===================================================================
> --- src/soot/jimple/spark/builder/StandardParms.java    (revision 1120)
> +++ src/soot/jimple/spark/builder/StandardParms.java    (working copy)
> @@ -363,6 +363,7 @@
>          AllocNode prevAn = pag.makeAllocNode(
>              new Pair( nmae, new Integer( type.numDimensions ) ), type, currentMethod );
>          VarNode prevVn = pag.makeVarNode( prevAn, prevAn.getType(), currentMethod );
> +        addEdge( prevAn, prevVn );
>          setResult( prevAn );
>          while( true ) {
>              Type t = type.getElementType();