Limiting Transitive Closure for Static Regression Test Selection approaches

Abstract: In computer science and software development it is important to test software in order to ensure reliability. Regression testing in order to find potential faults introduced by software changes is key to assuring that the software is stable. This process may be time consuming, so in order to speed it up there are approaches which select a subset of relevant tests based on the software changes.In the field of regression test selection there are two main approaches, static approaches and dynamic approaches. These different kinds of approaches have different strengths and weaknesses, however the field is currently dominated by dynamic approaches as the performance of the static approaches lags behind severely. Ensuring that the correct approach types are used for the appropriate situations calls for improvement of static approaches.Regression test selection approaches uses transitive closure to select relevant tests. For any node in a directed graph, transitive closure is the set of all reachable nodes from the starting node. This thesis proposes a solution which attempts to lessen the performance gap by implementing a controlled limit to the transitive closure property of the main test selection algorithm. The aim of the limited transitive closure is to reduce the time taken to select tests, and to reduce the amount of superfluous tests selected.The results show that the limited transitivity property as implemented for this thesis did not improve the performance in a satisfactory way. Safety dropped severely. Since the runtime and precision improved, there is room for improvement in potential future research of the limited transitivity approach before it can readily be dismissed as an unviable approach.