Testing Object-Oriented Software: A Survey

Component Completeness Checklng Formal Verlflca tlon Component Implementation T Fo rrn al Artifacts Informal Artifacts Behavior Figure 3. ADT artifacts cind verification techniques TESTING OBJECT-ORIENTED SOFTWARE 135 et al., 1981; Dauchy et al., 1993; Doong and Frankl, 1994). Relationships among axioms can be exploited to generate test sequences and test case data (Bougt et al., 1986; Choquet, 1986; Gaudel and Marre, 1988; Bemot et al., 1991; Dauchy et al., 1993; Doong and Frankl, 1994). Most ADT test approaches generate operation sequences which are hoped to be faultrevealing (Jalote and Caballero, 1988; Jalote, 1992; Zweben et al., 1992; Parrish et al., 1993b; Doong and Frankl, 1994). Tests using model-based specifications select sequences by analysing define-use relationships among exported operations (Zweben et al., 1992; Parrish et al., 1993b). A coverage measure may be derived from such a test suite. In contrast, tests using algebraic specifications make use of constraints on operations (axioms) to select paths by a substitution process. Automated support for model validation and consistency checking is discussed in Section 4. Product validation is not discussed here as it can be accomplished by established methods which are not specific to ADTs or object-oriented development. 2.2. Formal verification 2.2.1. Prover’s assistant An early presentation of the ADT approach argues that algebraic specification facilitates proofs by factoring complexity and allowing automated support. ‘Writing the axioms in a certain style allows them to be used as reduction rules so that proofs become largely symbol manipulation exercises.’ (Guttag et al., 1978, p. 1062). ‘Testing’ uses a direct implementation input to the ‘data type verification system’ (DTVS). A direct implementation for data type T is an ‘implementation whose representation part is a subset of the syntactic specification of T and whose program part is a subset of the semantic specification of T.’ (Guttag et al., 1978, p. 1059). The verification process has three steps: ( 1 ) input the specification and implementation to DTVS; (2) use DTVS to generate axioms and equality axioms for the specification (verification conditions); (3) the user attempts to prove the verification conditions, using an automated support for substitution of previously established results. This process is reported to be useful for discovering missing assumptions. The tool keeps track of proofs and does some syntactic checking.