Faculty of Science, Technology and Medicine
Faculty of Law, Economics and Finance
Faculty of Humanities, Education and Social Sciences
Interdisciplinary Centre for Security, Reliability and Trust
Luxembourg Centre for Systems Biomedicine
Luxembourg Centre for Contemporary and Digital History
Luxembourg Centre for European Law
Luxembourg Centre for Socio-Environmental Systems
The Doctoral School in Science and Engineering is happy to invite you to Hilal TAHA’s defence entitled
Maintenance of UI Tests in Evolving Web Applications
Supervisor: Assoc. Prof Michail PAPADAKIS
Web applications have become the primary interface for delivering online services, and their reliability is critical for user experience and business continuity. Automated End-to-End (E2E) testing is essential for validating complete workflows through the Graphical User Interface (GUI). However, the dynamic nature of modern web applications, which is driven by rapid development cycles, frequent User Interface (UI) changes, and evolving Document Object Model (DOM) structures, makes maintaining these tests a persistent challenge. Locator breakages, which stem from element identifiers becoming obsolete, account for the majority of test failures and significantly increase maintenance costs while eroding confidence in automated testing.
This dissertation addresses these challenges through three complementary contributions. First, we introduce a large-scale dataset of DOM evolution across 181 actively maintained websites, including up to 1,000 historical versions per site with ground truth annotations for DOM position stability and locator breakage detection. This dataset enables reproducible research on robust locator generation, test repair, and web evolution analysis—areas previously limited by the scarcity of representative, real-world data.
Second, we propose EAGL (Evolution-Aware Generation of robust Locators), a data-driven approach that predicts DOM position stability using six lightweight structural features and constructs XML Path Language (XPath) locators anchored at stable positions. Unlike heuristic-based methods, EAGL leverages these stability predictions to identify structurally persistent anchors, improving locator robustness without compromising precision. Comprehensive experiments on real-world websites demonstrate that EAGL outperforms both state-of-the-practice (Selenium) and state-of-the-art (Robula+) approaches in precision and robustness while maintaining practical efficiency. Third, we present an industrial replication study of the Correct Locator Recommender (COLOR) repair tool deployed in a privacy-sensitive Continuous Integration (CI) environment. This study analyses failure patterns across thousands of test executions, evaluates COLOR’s effectiveness in repairing broken locators, and identifies practical challenges in integrating automated repair into enterprise work-flows. Our findings demonstrate that while automated repair substantially reduces maintenance overhead, its effectiveness depends critically on attribute richness, locator design practices, and organizational constraints.
Collectively, these contributions advance understanding of test fragility and provide scalable, evolution-aware solutions for improving web application test maintainability. This work opens pathways for integrating stability prediction into self-healing frameworks, extending datasets to dynamic content, exploring semantic anchors for enhanced robustness, and bridging academic research with industrial practice to enable more resilient, cost-effective, and adaptive web testing strategies.