Utveckling av testverktyg för mjukvara med fysiska komponenter

Sammanfattning: A lot of time is spent on manual software testing. When software can interact with exponentially more and more components the number of tests required will increase at the same rate. Automation of these manual tests has been previously shown that a time-saving can occur with test automation. When tests need to be done on software that interacts with physical components the results of testing automation are harder to predict. This thesis aims to explain the impact of physical components on the development and the result of test tools developed for use on automated tests on objects with physical components. In this work, a test tool for automatic regression testing for height adjustable table legs were developed. The method used to design the tool was Design Science Research where, for example, TESLA was adapted for use in automated tests with physical components. TESLA is a language for specifying test cases and allowing test automation with embedded systems that was implemented in the test tool's design which could be used to automatically generate scripts and automatically executes test cases. With the development and design of the testing tool, the thesis attempted to answer how a test tool should be designed when physical components are included in the test object and how the measurement accuracy of the tool may affect the test result. Experiments have been carried out during the development process where regression tests have been performed according to a test protocol. During these experiments a test protocol is executed. One manually by hand and one is done automatically with the testing tool. The results of the experiments show that the execution of the test protocol execution with the test tool gave a time saving of 35% compared to the manual tests. The physical components of the test object were shown to affect the automatic execution time negative and further analysis of previous research has shown that this result is insufficient to justify the large initial time that is required to automate tests.