Assessing how developers review tests on Github

• Autor(a) principal: Victor Guerra Veloso
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/43919
• topic: Software testing; Code Review; Software repository mining; Software maintenance; Mutation Testing; Code quality; Computação – Teses; Engenharia de software – Teses; Software – Testes - Teses; Software – Controle de qualidade – Teses.

In modern software development, developers are expected to provide proper tests to cover their code changes. However, code contributions are not always attached to good tests. This way, reviewers may request test changes to the contributions. In practice, it is not clear what information is available to guide contributors in implementing the requested test methods during the code review. Therefore, better understanding test review practices would be important to guide both contributors and reviewers. In this dissertation, we propose two studies. First, we provide an empirical study to assess test reviews on GitHub. We find 11,836 test reviews in 5,421 open-source projects, suggesting that this is a common practice. We also detect eight broad categories of recommendations in test reviews: test scope, tool support, test scenarios, goal/purpose, refactoring, bad practices, fixtures, and miscellaneous. Lastly, we find that test reviews with more recommendations are more likely to be solved. In our second study, we propose a tool to assess the quality of individual test methods by relying on mutation testing. This tool extends a state-of-the-art mutation testing framework to analyze test methods and report mutation results at the method level. Finally, based on our results, we discussed implications for researchers and practitioners.