Uma abordagem para seleção automática de bugs em projetos de software baseados em componentes

• Autor(a) principal: Lima, Georgenes de Araújo
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/123456789/13345
• topic: Engenharia de software baseada em buscas; Sistemas de rastreamento de bugs; Seleção de bugs; Componentes de software; Search based software engineering; Bug tracking systems; Bug selection; Software components; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Nowadays, technology is present in almost every work process and consequently the demand for technological resources is more and more frequent, leading to further changes in software systems, mainly due to additional maintenance requests, and, therefore, a greater number of bugs for correction. Taking into account that bug fixing is one of the factors that impacts on software quality improvement, but at the same time represents a high expenditure of resources in a software project, bug fixing processes represent a critical hurdle for software maintenance and development. Given that the number of existing reported bugs may be greater than the resources needed to correct them, and the fact that many software companies currently work with successive deliveries, called releases, which allow the delivery of parts of the software product, project managers may face the challenge of manually selecting the bugs to be fixed in the next software version, usually based on the criteria of priority and severity, indicated in the respective bug reports, as well as delivery time or cost. For this, project managers adopt the socalled Bug Tracking Systems (BTS), which establish communication channels between software producers and their customers. For large and complex systems, where a substantial number of bug reports are common, a bug selection approach based on the ad-hoc judgment of experts is not efficient and effective, representing a time-consuming, labor-intensive and errorprone strategy. In such a context, exploring Search Based Software Engineering (SBSE) techniques, this dissertation proposes an automated approach for selecting interdependent bugs in component-based software projects, adopting a multiclient model to select the most relevant bugs from the point of view of both the software producer and its customers, but with total cost limited by the budget available in the software project. In order to evaluate the proposed approach, two case studies were carried out, where the search strategy based on genetic algorithm is compared with other strategies based on both exhaustive and random algorithms, and, more realistically, also compared with the selection based on knowledge and experience of systems analysts. The expressive results obtained in the case studies express the feasibility, practicality and effectiveness of the proposed approach.