Automated Test Generation from Program Documentation Encoded in Code Comments (2504.21161v1)
Abstract: Documenting the functionality of software units with code comments, e.g., Javadoc comments, is a common programmer best-practice in software engineering. This paper introduces a novel test generation technique that exploits the code-comment documentation constructively. We originally address those behaviors as test objectives, which we pursue in search-based fashion. We deliver test cases with names and oracles properly contextualized on the target behaviors. Our experiments against a benchmark of 118 Java classes indicate that the proposed approach successfully tests many software behaviors that may remain untested with coverage-driven test generation approaches, and distinctively detects unknown failures.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.