THE USE OF MODEL CONSTRAINTS AS IMPRECISE SOFTWARE TEST ORACLES

Abstract

Many software test generation techniques target on generating software test data. Only a few of them provide automatic way to verify if software behaves correctly using generated test data. We propose a testing technique, which uses UML modeling language extension OCL as imprecise test oracle. Imprecise OCL constraints can be viewed as expressions which define expected results within some ranges of possible values. When software is executed using generated test data the output is verified against imprecise OCL constraints. If output invalidates imprecise OCL constraints, a tester can assume with some probability that software has bugs.