A comparative study of regression testing methods. (c1996)

Abstract

We present a comparative study of popular regressIOn testing algorithms. These algorithms include slicing, incremental, firewall, genetic, and simulated annealing algorithms. The study uses a variety of small-size and medium-size modules along with associated test cases tables, and is based on the following quantitative and qualitative criteria, efficiency, number of retests, precision, inclusiveness, user's parameter setting, global variables, type of maintenance, type of testing, level of testing, and type of approach. The comparison results show that the five algorithms are suitable for different requirements of regression testing. Slicing and adapted firewall algorithms detect the definition-use pairs that are affected by a change, and select the test cases for regression testing based on these definition-use pairs. Incremental algorithm selects the test cases whose outputs may be affected. Genetic and simulated annealing select the minimum number of test cases that provide full testing coverage. In terms of execution time for small-size modules, slicing, incremental, and adapted firewall algorithms exhibit a better behavior comparing to genetic and simulated annealing algorithms. For medium-size modules, the adapted firewall algorithm becomes the slowest. Genetic and simulated annealing algorithms produce the least number of retests, followed by incremental, slicing, and then adapted firewall.