And technical capability deploy the methods

TWO PRINCIPLES OF TEST-CASE SELECTION 9

and therefore become handy under situations when no existing method is applicable. For example, when a new software system is procured, the user organization often needs to test it to see if it is reliable enough to pay the vendor and release the system for production runs. If an operational profile is available, the obvious choice is to perform operational testing as described in Section 4.2. Otherwise, test-case selection becomes a problem, especially if the system is fairly large. Source code is generally not available to the user to make use of the methods presented in Chapter 2, and detailed design documents or specifications are not available to use the methods presented in Chapter 3. Even if they are available, a typical user organization simply does not have the time, resources, and technical capability to deploy the methods. In that event, the two principles can be utilized to select test cases. The components to be exercised could be the constituent subsystems, which can be recognized by reading the system user manual. Two inputs are weakly coupled computationally if they cause different subsystems to be executed in different sequences. Expert users should be able to apply the two principles readily to achieve a high probability of fault detection. In short, if one finds it difficult to use any existing method, use the two principles instead.

In the preceding section we said that a test case should be selected from a subdomain or a subset of inputs that causes a component to be exercised during the test. Is there a better choice if there is more than one? Is there any way to improve the fault-detection probability by using more than one test case from each subdomain? The answer depends on the types of faults the test is designed to reveal. What follows is a fault classification scheme that we use throughout the book.

In the abstract, the intended function of a program can be viewed as a function f of the nature f : X → Y. The definition of f is usually expressed as a set of subfunctions f1, f2, . . . , fm, where fi : Xi → Y (i.e., fi is a function f restricted to Xi for all 1 ≤ i ≤ m), X = X1 ∪ X2 ∪ · · · ∪ Xm, and fi ̸= f j if i ̸= j. We shall use f (x) to denote the value of f evaluated at x ∈ X, and suppose that each Xi can be described in the standard subset notation Xi = {x | x ∈ X ∧ Ci(x)}.

additional restrictions on the inputs. We shall use S(x) to denote the computation P2 ∨ · · · ∨ Pn, and P is in general equal to T (true) unless the programmer places

performed by an execution of S with x as the input.