Partially safe evolution of software product lines

• Ano de defesa: 2017
• Autor(a) principal: SAMPAIO, Gabriela Cunha
• Orientador(a): BORBA, Paulo Henrique Monteiro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Pernambuco
• Programa de Pós-Graduação: Programa de Pos Graduacao em Ciencia da Computacao
• País: Brasil
• Palavras-chave em Português: Engenharia de software; Linhas de produtos de software
• Link de acesso: https://repositorio.ufpe.br/handle/123456789/24883

Software product lines (SPLs) are sets of related systems that are built based on reusable artifacts. They have three elements: a variability model, that has feature declarations and dependencies among them; implementation artifacts and a configuration knowledge, that maps features to their implementation. SPLs provide several advantages, like software quality and reuse improvements, productivity gains and the capacity to customize a system depending on customers needs. There are several challenges in the SPL development context. To build customizable software and meet all customer needs, SPLs tend to increase over time. The larger a SPL becomes, the higher is the complexity to evolve it. Therefore, it is not trivial to predict which products are affected by a change, specially in large SPLs. One might need to check if products had their behaviour preserved to avoid inadvertently affecting existing users in an evolution scenario. In refactoring and conservative extension scenarios, we can avoid this problem by checking for behavior preservation, either by testing the generated products or by using formal theories. Product line refinement theories support that by requiring behavior preservation for all existing products. This happens in a number of situations, such as code refinements. For instance, in function renaming transformations, all existing products behave exactly as before the change, so we can say that this transformation is safe. Another example of SPL refinement would be changing a feature type from mandatory to optional. In this case, we increase variability, but preserving all products from the original SPL. Although several evolution scenarios are safe (or technically refinement), in many others, such as bug fixes or feature removals, there is a high chance that only some of the products are refined. In these scenarios, the existing theories would give no support, since we can not assume behaviour preservation holds for all products. To support developers in these and o ther non refinement situations, we define partially safe evolution for product lines, that is formalised through a theory of partial refinement that helps to precisely understand which products should not be affected by an evolution scenario. This provides a kind of impact analysis that could, for example, reduce test effort, since products not affected do not need to be tested. Additionally, we formally derive a catalog of partial refinement templates that capture evolution scenarios, and associated preconditions, not covered before. Finally, we evaluate the proposed templates by analyzing commits from two product line systems (Linux and Soletta) and we found evidence that those templates could cover a number of practical evolution scenarios.