Year: 2025
Type: conference-paper
Abstract: Test smells indicate problems in automated tests that hinder the understanding, effectiveness, and maintenance of the test code. Due to their relevance, many tools have been created to detect them in xUnit frameworks, which has led to the emergence of test framework-agnostic strategies for detecting smells. However, these strategies are still little explored in the literature, and the available tools depend on obsolete components with considerable limitations. This study proposes the AgNose tool, which extends the strategy of agnostic test smell detection. The strategy is based on creating, from the test code in different frameworks, a simplified structure aimed solely at test smell detection. To implement the proposal, we (i) identify test smells analyzed simultaneously by different detection tools, as well as their identification steps; (ii) establish the minimum XML structure that represents the important information for detecting smells; (iii) implement the AgNose tool; (iv) validate the tool using real open-source projects from public repositories. As results, we map 8 test smells and their minimal XML structure. Also, the tool validation achieved a precision of 96.9%, a recall of 87.5%, and an f-measure of 92%. The presented strategy and tool extend current strategies for agnostic test smell detection, eliminating the perceived limitations of available tools.
Authors: André Cardoso, Denilson Vieira, Elvys Soares
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4413684735