[Soot-list] Spark as default pointer analysis

[Eric Bodden]

Hello.

> For the non-expert Soot users, spark  seems to be highly resource-  and too
> time-consuming. Enabling Spark by default might scare them away from using
> Soot; For those who know well Soot, they should be willing to enable
> manually Spark I suppose.

I see your point. But on the other hand, it would be only enabled if
-w is enabled, in which case Soot normally takes its time anyway.

> Apparently, the -no-bodies-for-excluded and -allow-phahtom-classes go in
> this direction. I guess the analyses could be unsound then? An interesting
> question is really,  under what *exact* conditions, Spark with those options
> enabled still gives still sound results.   Then, it might be more
> appropriate to enable SOOT by default, in my humble opinion.

Yes, the analyses are generally unsound in this setting. Ondrej
Lhotak's group had a PLDI'12 paper on a better solution that is sound
but they are still in the process of porting it to Soot.

The reason for why I am proposing this change is that somehow I had it
in my mind that it already way enabled by default and then was totally
surprised by oddly imprecise benchmarks results, causing me to find
out that it was not.

Eric