Study On GUI Automation Test Model Based On Event-flow Graph

Nowadays Graphical user interfaces have become the most common means of interacting with software systems, and the investigation says that GUI testing can consistitute as much as 60% of the whole testing work of software testing. In the industry, C/R (capture and replay) and scripts-driven tools are widly used. These automation tools need largly manual work, with low adaptability to software variation, and no systematic management approach of test cases and test coverage.Event-flow graph is a directed graph with GUI events as the verteices and interactions between events as edges. Base on the Event-Flow Graph, a highly automated system model has been proposed. The proposed system using automation tool analyses the sample GUI build with reverse engineering apporch, and obtain a majority of event-flow graph and expected testing results.With the obtained graph model, the test case generation problem transforms to the one with the directed graph. Two strategies has been designed for two different test scenes: one for the daily smoke testing, ant colony optimization system has been used to tranverse the event-flow graph and find a path contains all event vertices as a test case; and the other one for the"deep"regression testing, using goal-directed searching approach to search through the spanning tree, and obtain the test cases of the target event vertex. During the automated test, an emergency policy, which is similar with the reverse engneer analyse process, has been adopted to avoid testing fail caused by the software variation.The implement of the prototype based on Microsoft UI Automation, which is the new accessiblility framework of Windows operation system. It addresses the needs of assistive technology products and automated test frameworks by providing programmatic access to information about the user interfaces, and represents the controls of GUI as a tree structure. A notepad application with varied controls has been used as the target program in the experiment. Several bugs have been injected in the GUI, and proves the detection capibility of the system.The results of the comparable experiments of the proposed model and traditional C/R tool illuminates that the event-flow graph and two strategies for test case generation is usable, and the system expose a helpful means for GUI test automation improvement.