This paper presents research about the automated testing of graphical user interfaces (GUIs). Although it is addressed to the designers of such tools, it can be useful to anyone who wishes to understand the nature of GUI testing.

The introduction explains the complexity of GUI testing and approaches to making it automatic. The testing is based on GUI events, which can be numerous and can occur in complex sequences. Two approaches for creating test cases are presented: script-based tools, where test cases are manually created to call GUI functions, and capture/replay tools, where the tool captures the user’s input and creates test cases that are then replayed to perform the automated test.

The authors describe their own tool, GUITAR, which is more complex than the above tools. GUITAR is model-based; it can be configured and it is modular. GUITAR is compared to other similar tools to show its advanced features. One section covers the components of GUITAR: the ripper, which analyzes the user interfaces and extracts all the possible events; a graph converter to represent the flow of user action; the test case generator; and the replayer, which executes the generated tests. The architecture of GUITAR is presented together with extensions.

The paper refers to the three phases of testing: build test cases, combine them in test suites, and then execute test cases and check if the results are as expected. The results of applying GUITAR to Java Foundation Classes (JFC) are shown with images and graphs.

The paper includes a wealth of references and the websites of different tools. It is missing a table of contents, which would provide readers with an overview of the structure of the paper. The paper is both broad and deep in its approach, and covers the huge domain of GUI applications that need testing.