Automatic Systematic GUI Testing for Web Applications

Abstract

Automated testing is crucial for software development, providing efficiency, cost reduction, and repeatability. At the Graphical User Interface (GUI) level, it validates functionalities and detects faults in web applications. Tools that utilize generation and execution processes often aim to identify visible failures, such as crashes, error messages, and unexpected behaviors. In this context, we propose Cytestion, an automated and systematic GUI testing approach and tool for web applications, which applies a scriptless and progressive approach. Starting with an initial test case, it progressively explores elements using browser console errors, HTTP request status, and GUI failure messages for validation. In order to be effective, Cytestion had to face three primary challenges: automatic and unique discovery of actionable elements, robust synchronization with the application under test, and managing extended runtime in industrial web contexts. To address these, we introduced and evaluated the Unique Actionable Elements Search (UAES) approach, the NetworkWait mechanism, and the Iterative Deepening URL-Based Search (IDUBS) algorithm. These solutions advanced the field of automated testing. Our empirical studies using four open-source and twenty industrial web applications demonstrate Cytestion superior performance in fault detection and runtime efficiency compared to a state-of-the-art GUI testing tool.