Master Thesis Presentation:
Date: February 12, 2009 (Thursday) at 10:15
Per Heed och Alexander Westrup presenterar sitt examensarbete "Automated platform testing using input generation and code coverage"
Abstract
When using a Java platform it is important to test all aspects of it thoroughly. In this thesis an attempt to test the stability and overall function is presented. The method is to test the platform by running a large number of Java applications on the platform and see if anything goes wrong. To stress the platform as much as possible it is important to not only start the application but to attempt to explore it as well. This makes sure that as much of the platform as possible is tested. The focus of the thesis is to compare different ways to generate input to the application and try to find the most efficient method. Four different input generation methods are evaluated, random, adaptive and both with a constant startup sequence and compared to a manual reference. To compare the input generators performance, ten games were selected and run a number of times. During these runs the code coverage of the games was calculated. Factorial design was applied to the code coverage values to determine if there were any statistically significant differences between the input generators. The results show that the startup sequence give good code coverage values by quickly going through the menus and start the game. The adaptivity gives somewhat better code coverage in some games than simply random but requires code coverage to function which decreases the performance of the phone. From the results it can be concluded that no input generator is the best for all situations. However in the test environment used in the thesis the random with startup sequence was deemed the best due to too big performance loss when running in debug mode to enable code coverage calculation.
Room: E:2116
Last modified Dec 9, 2011 12:57 pm by Mikael.Antic@cs.lth.se
0176