Towards an expertise-related metric to preprocessor-based configurable software systems

• Ano de defesa: 2022
• Autor(a) principal: Karolina Martins Milano Neves
• Orientador(a): Bruno Barbieri de Pontes Cafeo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: Code Authorship, Ownership, Expertise-related metrics, Configurable Software Systems, Preprocessor
• Link de acesso: https://repositorio.ufms.br/handle/123456789/5064

Context: Expertise-related metrics allow us to find the best developers for a target task in a file. Configurable systems use variability as a unit of abstraction to generate different members of a program family. This misalignment between files used by expertise-related metrics and variabilities used by configurable systems may make it impossible to use them together. Objective: The objective is twofold. The first is to explore how the work on mandatory and variable code is divided among developers and whether expertise-related metrics can indicate a developer with expertise for a task involving variable code. The second is to propose a variability-aware expertise related metric to indicate developers with expertise in variable code. Method: We investigate 49 preprocessor-based configurable systems. We analyzed how variabilities changes are divided between developers and whether these developers would be key developers indicated by expertise-related metrics. We use feature selection and multiple linear regression techniques to propose a variability-aware expertise-related metric. We validate our metric by comparing it with two well-known metrics. Results: Few developers are specialists in variable code. We also identified that only a few developers concentrate the majority of changes in variable code. The results also suggested that expertise-related metrics are not a good fit to indicate experts regarding variable code. We proposed a variability-aware expertise-related metric and showed that our proposed metric outperformed well-known expertise-related metrics. Conclusion: Even though the results show that a considerable number of developers touched variable code during the development history, such changes are only occasional. There is a concentration of work among a few developers when it comes to variable code. This uneven division may cause an unnecessary maintenance effort. We also conclude that variability-aware expertise related metrics may better support the identification of experts in configurable systems when compared to existing metrics.