Documenting and managing self-admitted technical debt using issues

• Autor(a) principal: José Laerte Pires Xavier Júnior
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/51321; https://orcid.org/0000-0001-7925-4115
• topic: Technical debt; Documentation; Mining software repositories; Issue tracking systems; LTD framework; Computação – Teses; Dívida técnica – Teses; Mineração de repositórios de software – Teses; Framework (Programa de computador) – Teses; Software - Documentação – Teses

The Technical Debt (TD) metaphor refers to the unavoidable maintenance and evolution costs of the not-quite-right decisions commonly taken by software developers. The term was quickly adopted in industry and became the subject of various studies in the literature, mostly regarding techniques for TD identification, management and payment. In recent years, several studies emerged around the observation that developers explicitly document their debts, which the literature refers as Self-Admitted Technical Debt (SATD). Particularly, most previous studies investigated this practice by analyzing source code comments as indications of TD admission. In this thesis, we denote this form of SATD as SATD-C. However, there is a lack of knowledge about the adoption of different artifacts to admit TD. Particularly, few studies deeply investigated SATD in issue tracker systems, despite the increasing adoption of issues in software development. In order to contribute to this problem, we describe in this thesis an in depth study on the adoption of issues to document and manage Self-Admitted Technical Debt (which we refer as SATD-I). We organize the research in four major working units. We start by exploring the characteristics of SATD-I adoption, in terms of the types of TD commonly documented in issues, and the motivations of its insertion and payment. Next, we strengthen our initial results by investigating the interplay between SATD-C and SATD-I in a large-scale dataset. Based on the evidences that they have distinct natures, we assess the circumstances when each form is more suitable. Finally, we wrap-up the knowledge produced in our empirical studies by proposing and evaluating a light-weight framework to support developers to manage TD through the creation of issues. We call this framework as LTD: Less Technical Debt Framework. Overall, our results confirm that developers also use issues to admit TD in their projects. We also show that issues are more suitable to document high-level and high-priority debts, mostly regarding concerns that need visibility and discussions between developers. Finally, we achieved promising results after adopting LTD in two development teams in a large public company. For example, the teams could reduce TD and create a backlog of issues to manage TD during the execution of sprints.