Testing

TESTING
Testing is perhaps the oldest technique for detecting errors or problems in implemented software, hardware, or non-computer systems. It consists of executing or operating (in the case of a non-computer system) the system to be tested using a finite set of inputs and then checking to see if the corresponding outputs or behavior are correct with respect to the specifications. To test a real-time system, the values as well as the timing of the inputs are important. Similarly, both the output values and the time at which they are produced must be checked for correctness.

Many approaches have been developed for testing software, hardware, and noncomputer systems. The simplest technique is of course to perform an exhaustive test run of the system with every possible input and then to check if the corresponding output is correct. This approach is not practical except for small systems with limited input space. For larger systems, the time needed to test is prohibitively long. For systems with an infinite number of possible inputs, this approach is of course not viable. Since relatively little training is required on the part of the testing personnel, testing has been and will continue to be used extensively in industry.

There are three common techniques for software testing in the current state-ofthe- practice: functional testing, structural testing, and code reading. Functional testing uses a “black box” approach in which the programmer creates test data from the program specification using one or a combination of techniques such as boundary value analysis and equivalence partitioning. Then the program is executed using these test data as input, and the corresponding program behavior and output are compared with those described in the program specification. If the program behavior or output deviates from the specification, the programmer attempts to identify the erroneous part of the program and correct it.

Partition testing is a popular way to select test data, in which the program’s input domain is divided into subsets called subdomains, and one or more representatives from each subdomain are chosen as test input. Random testing is a degenerate form of partition testing since it has only one subdomain, the entire program.

Structural testing is a “white box” approach in which the programmer examines the source code of the program and then creates test data for program execution based on the percentage of the program’s statements executed.

Code reading by stepwise abstraction requires the programmer to identify major modules in the program, determine their functions, and compose these functions to determine a function for the whole program. The programmer then compares this derived function with the intended function as described by the program specification.