Background

Regression testing is a complex and challenging activity and consumes a significant portion of software maintenance costs. Researchers are proposing various techniques to deal with the cost and complexity of regression testing. Yet, practitioners face various challenges when planning and executing regression testing. One of the main reasons is the disparity between research and practice perspectives on the goals and challenges of regression testing. In addition, it is difficult for practitioners to find techniques relevant to their context, needs, and goals because most proposed techniques lack contextual information.

Objective

This work aims to understand the challenges to regression testing practice and find ways to improve it. To fulfil this aim, we have the following objectives:

1) understanding the current state of regression testing practice, goals, and challenges,

2) finding ways to utilize regression testing research in practice, and

3) providing support in structuring and improving regression testing practice. 

Method

We have utilized various research methods, including literature reviews, workshops, focus groups, case studies, surveys, and experiments, to conduct the studies for this thesis.

Results

Research and practice stress different goals, and both follow their priorities. Researchers propose new regression testing techniques to increase the test suite's fault detection rate and maximise coverage. The practitioners consider test suite maintenance, controlled fault slippage, and confidence their priority goals. The practitioners rely on expert judgment instead of a well-defined regression testing process. They face various challenges in regression testing, such as time to test, test suit maintenance, lack of communication, lack of strategy, lack of assessment, and issues in test case selection and prioritization. 

We have proposed a GQM model representing research and practice perspectives on regression testing goals. The proposed model can help reduce disparities in research and practice perspectives and cope with the lack of assessment. 

We have created regression testing taxonomies to guide practitioners in finding techniques suitable to their product context, goals, and needs.  Further, based on the experiences of replicating a regression testing technique, we have provided guidelines for future replications and adoption of regression testing techniques.

Finally, we have designed regression testing checklists to support practitioners in decision-making while planning and performing regression testing. Practitioners who evaluated the checklists reported that the checklists covered essential aspects of regression testing and were useful and customizable to their context.

Conclusions

The thesis points out the gap in research and practice perspectives of regression testing. The regression testing challenges identified in this thesis are the evidence that either research does not consider these challenges or practitioners are unaware of how to replicate the regression testing research into their context. The GQM model presented in this thesis is a step toward reducing the research and practice gap in regression testing. Furthermore, the taxonomies and the replication experiment provide a way forward to adopting regression testing research. Finally, the checklists proposed in this thesis could help improve communication and regression test strategy. Moreover, the checklists will provide a basis for structuring and improving regression testing practice.