Symbolic Execution for GUI Testing

A Graphical User Interface (GUI) is an abstraction providing users with a more natural way of interacting with computers. It consists of objects like buttons, text boxes, toolbars etc. The communication between users and GUIs is event driven. Users can modify the state of a GUI and trigger events that lead to the execution of different code fragments. Hence, in order to test a GUI one should execute event sequences simulating user behaviors. While the state of some GUI widgets is limited to a small number of values (the value of a radio button), others have a wide range of possible states (the value of a text box). Such widgets are used for data input from the user in the form of text (alphabetic or numeric). Since program execution may depend on the user input, it is a challenge to select suitable values in a way that allows thorough testing. We propose symbolic execution for obtaining these inputs. During symbolic execution, each branch of the program is visited and the constraints for control variables are resolved determining if it is reachable or not. Thus, by symbolically executing code that depends on user input, we can obtain values that ensure visiting each reachable branch in the program.