Software Testing: Concepts and Operations

10Structural Criteria

Whereas the functional criteria discussed in Chapter 9 generate test data by analyzing the specification of the product, the techniques we discuss in this chapter do so by analyzing the source code of the product. Structural test generation criteria may sound counter-intuitive, since they are using the program to test the program. It is like recruiting the prime suspect in a crime to help in the crime investigation; in the same way that the criminal will bias the investigation away from himself, a faulty program may focus the attention of the tester away from its shortcomings. This is not an idle comparison: Consider that if the program fails to make provisions for half of the input data that it is supposed to process (a massive gap), no amount of structural testing will expose this fault; only by referring back to the specification can we uncover such a failure. More generally, when we endeavor to verify a program, our main resource is redundancy (redundancy between a program and its specification, redundancy between two presumably equivalent programs, redundancy between a program and an executable specification, redundancy between a program and a test oracle, etc.); if we use a program to test the same program, we have no redundancy to depend on. On the other hand, consider that in order to make any claim about the functional attributes of candidate programs, a test-generation strategy must focus on the domain of the specification as a starting point; to ...