Detecting and Mitigating Inconsistencies Between Code, Documentation and Tests

Abstract

Inconsistencies between different software artifacts, such as source code, documentation, and tests, are a common and long-standing problem in software engineering. These misalignments can degrade software quality, slow down development, and hinder maintenance. Each of these artifacts provides a distinct yet overlapping perspective of the same software behavior, forming a triangular relationship in which any one artifact can, in principle, be used to regenerate the others.In this PhD project, we aim to exploit these relationships through regeneration-based techniques to detect and ultimately mitigate inconsistencies across software artifacts. The approach focuses on building an SE-tool that leverages original and regenerated versions of artifacts to triangulate inconsistencies and improve reliability through cross-validation. A particular emphasis is placed on reducing false positives, ensuring that reported issues are trustworthy and actionable, especially when presented to developers in real-world projects.