Logic representation of programs to detect arithmetic anomalies

Much interest, especially by banks and insurance companies is paid to detect arithmetic anomalies and inexactness of arithmetic expressions. Numerous examples in the past show that although mathematical methods for correct implementation of arithmetic expressions exist and are well understood, many programs contain arithmetic anomalies, impreciseness or faults. Software tests based on conventional coverage criteria (F. Belli, 1998) and functional tests are not well suited for detection of these faults. The detection of arithmetic anomalies by these methods strongly depends on the adequateness of test cases. The selection of effective test cases needs a lot of effort to detect context-sensitive arithmetic inexactness. The authors introduce a novel approach for detecting arithmetic anomalies. The method is based on the specification of fault classes combined with the transformation of the program under test into a predicate logic model. The number of potential context-sensitive faults is deployed as a criterion to precisely select modules in large software systems to increase the test effectiveness