Bugfix: latest fix to policy iteration broke a previous regression test (rewpoliter.nm) - solved by doing inner solution using lowe level method which takes infinity-states as an argument (in any case, precomputation should not be needed for inner solution).

git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@7772 bbc10eb1-c90d-0410-af57-cb519fbb1720

prism/src/explicit/MDPModelChecker.java

@@ -1739,7 +1739,7 @@ public class MDPModelChecker extends ProbModelChecker
 			// Solve induced DTMC for strategy
 			dtmc = new DTMCFromMDPMemorylessAdversary(mdp, strat);
 			mcRewards = new MCRewardsFromMDPRewards(mdpRewards, strat);
-			res = mcDTMC.computeReachRewards(dtmc, mcRewards, target, reUseSoln ? soln : null, null);
+			res = mcDTMC.computeReachRewardsValIter(dtmc, mcRewards, target, inf, reUseSoln ? soln : null, null);
 			soln = res.soln;
 			totalIters += res.numIters;
 			// Check if optimal, improve non-optimal choices