Swarnendu Biswas and Rajib Mall
Software maintenance, regression testing, multi-objective optimiza-tion, genetic algorithms, embedded programs
Regression testing of evolving embedded programs is characterized by the large overheads incurred in terms of cost, time and effort. In addition, rapid release cycles and schedule constraints further exac- erbate the challenges in exhaustive regression testing. In this con- text, we propose a program model-based, multi-objective regression test suite optimization technique, named GA-TSO, for embedded programs. GA-TSO aims to minimize the cost of regression testing, maximize the reliability of the frequently executed non-critical func- tionalities, and removes redundant test cases during optimization. Given schedule constraints, GA-TSO also tries to ensure that the thoroughness of regression testing using the optimized test suite is not compromised by defining optimization constraints that do not allow to omit test cases that execute affected tasks and critical functionalities. Experimental results obtained using a prototype implementation of GA-TSO show that the test suites optimized by GA-TSO include more potentially fault-revealing test cases from the initial regression test suite as compared to existing optimization techniques, and at the same time achieve substantial savings in regression testing effort.
Important Links:
Go Back
