[Soot-list] LocalMustAliasAnalysis inaccuracy

[Eric Bodden]

Hi all.

I just committed another version of the LocalMustAliasAnalysis.
Hopefully it is correct now. To implement this attempt, I had to add
an overloaded version of the merge function, which gets passed not
only the two in sets and the out set but also the "merge point", i.e.,
the successor unit of the merge. This allows us to assign to each
variable a unique number for each merge point. Because there are only
finitely many merge points and variables, the number of generated
numbers is bounded.

This approach is similar to generating "phi nodes" in SSA: there's one
phi node for each variable and merge point.

Arie, the analysis now seems to report the correct results (and
terminate!) on all the examples you provided and on all examples I
have in my personal set of test cases. I would appreciate if you could
try out the analysis and maybe also give it a quick code review.
Please report back if you find any more issues.

Eric
-- 
Eric Bodden
Software Technology Group, Technische Universität Darmstadt, Germany
Tel: +49 6151 16-5478    Fax: +49 6151 16-5410
Mailing Address: S2|02 A209, Hochschulstraße 10, 64289 Darmstadt