Automation of Software Test, Second International Workshop on (2009)
Vancouver, BC Canada
May 18, 2009 to May 19, 2009
Gordon Fraser , Institute for Software Technology, Graz University of Technology, Inffeldgasse 16b/2, A-8010, Austria
Angelo Gargantini , Dip. di Ing. dell'Informazione e Metodi Mat., University of Bergamo, Viale Marconi 5, 24044 Dalmine, Italia
Many different techniques have been proposed to address the problem of automated test case generation, varying in a range of properties and resulting in very different test cases. In this paper we investigate the effects of the test case length on resulting test suites: Intuitively, longer test cases should serve to find more difficult faults but will reduce the number of test cases necessary to achieve the test objectives. On the other hand longer test cases have disadvantages such as higher computational costs and they are more difficult to interpret manually. Consequently, should one aim to generate many short test cases or fewer but longer test cases? We present the results of a set of experiments performed in a scenario of specification based testing for reactive systems. As expected, a long test case can achieve higher coverage and fault detecting capability than a short one, while giving preference to longer test cases in general can help reduce the size of test suites but can also have the opposite effect, for example, if minimization is applied.
formal specification, program testing, program verification, software quality
G. Fraser and A. Gargantini, "Experiments on the test case length in specification based test case generation," 2009 ICSE Workshop on Automation of Software Test (AST 2009)(AST), Vancouver, BC, 2009, pp. 18-26.