STOPPING CRITERIA FOR REGRESSION TESTING IN GUI APPLICATION USING FAILURE INTENSITY AND FAILURE RELIABILITY

Abstract

This research proposes some criteria for GUI regression testing to determine the appropriate time to stop without wasting too much testing cost. This is essential for all software upgrades that can be released in a reasonably short time, yet still guarantees the product quality. One difficulty to achieve such a target depends on the sequence of test cases being input. The order of the input test case input sequence affects the number of failures found. As such, a test-stoppage model is proposed by determining factors that affect software reliability and the expected cost of continuing test. The procedure prioritizes the order of test cases into different sequences for the regression test input. When a failure is found, it is immediately edited before the test resumes. The test terminates when the failure intensity is within the predetermined threshold and the expected cost does not exceed the allotted budget limit. Performance of the proposed criteria encompasses three measures, namely, failure intensity, cost of testing and editing, and reliability. Failure intensity is a function of faults and fault detection rate. The costs are function of time spent on fixing errors. The reliability function incorporates Weibull distribution to better reflect the test data. The proposed model is tested using real GUI applications as test data. Performance shows satisfactory results on stopping criteria.