Despite the almost universal reliance on testing as the means of locating software errors and its long history of use, few criteria have been proposed for deciding when software has been thoroughly tested. As a basis for the development of usable notions of test data adequacy, an abstract definition is proposed and examined, and approximations to this definition are considered.

Harbhajan Singh, Mirko Conrad, Gottfried Egger, and Sadegh Sadeghipour Daimler-Benz AG Research and Technology, F3S/K Alt-Moabit 96a, D-10559 Berlin, Germany e-mail: fsinghjconradjeggerjsadeghg@DBresearch-berlin.de Abstract. This extended abstract describes an approach for generating test cases from Z specifications. The approach combines the classificationtree method with the disjunctive normal form approach as follows: A classification-tree is constructed from the formal specification and the resulting high level test cases are transformed into refined test cases by generating disjunctive normal form for them. Furthermore, such test cases contain information for evaluating the test results. The combination contains advantages of both methods, overcomes their limitations and can be supported by tools. 1 Introduction Testing is the primary method through which the software producer and the customer gain confidence that the software will work as intended or specified. The use of formal...

This paper presents a methodology to apply a software testing theory. The presentation of the methodology is accompanied by a central example and several supplementary examples. We also raise a number of issues related to our implementation methodology. A Methodology to Implement Gourlay's Mathematical Framework for the Investigation of TestingJune 16, 19932 1.0 Introduction Software testers find themselves faced with a vast array of test methods from which to choose. Of immediate concern is answering the question, "which test techniques are better than others?" Testing organizations have attempted to answer this question via empirical techniques, mostly based on establishing measures of performance based on organizational experience with different test techniques. Immediately, difficulties arise when attempts to normalize these performance measures (in order to apply results to predict future behavior) with respect to program size, complexity, experience of developers, and so on. ...