Using Code Instrumentation to Improve Text Input Generation for Scriptless GUI Testing

Abstract

Scriptless GUI testing aims to automatically test applications by traversing their GUI and detecting any error conditions that occur. It generally uses a black-box approach for generating test inputs for textwidgets, for example by using randomly generated text values. In an exploratory experiment, code instrumentation was used to dynamically generate potentially relevant input values for text widgets during scriptless GUI testing with the TESTAR tool.
Two Python-based web applications were instrumented with coverage measurement software. This instrumentation measured code coverage as a dependent variable and recorded a list of statements executed as a result of each test action. Relevant string literals in executed statements were extracted from the application source code and used to generate input values for text widgets.
Although no evidence was found that using code instrumentation to generate inputs for text widgets resulted in detection of more bugs or in structurally better code coverage, one issue was found that would not have been detected using the baseline method of generating random text inputs.

Date of Award	14 Dec 2022
Original language	English
Supervisor	Tanja Vos (Examiner) & Pekka Aho (Co-assessor)