Spectrum-Based Test Suite Reduction And Fault Localization

Debugging is a complex and time-consuming part in software development and traditional debugging is usually carried out manually by programmers. In order to help relieving the debugging burden, various approaches have been proposed by researchers. Among them, spectrum-based fault localization approaches, which are considered to be the most effective approaches, help programmers locate faults by utilizing dynamic coverage information and calculating suspiciousness for each statement. However, there are some open problems in these approaches. Firstly, these approaches require a high-coverage test collection and its’ coverage information. Although high-coverage test inputs collection can be easily acquired by automated test case generation tools, but the results of test runs are unknown. Secondly, although these approaches could reduce the number of statements needs to be examined, for large-scale program, programmers still need to check a large number of statements in order to locate the faults.For the problems in spectrum-based fault localization, this paper does some thorough investigation and analysis, and proposes specific solutions in test suite reduction and fault localization efficiency improvement. The major contributions are listed as follows:(1)This paper proposes a novel dynamic-basic-block based coverage criterion and two dynamic-basic-block based strategies to reduce the size of test inputs. Compared to existing methods, methods in this paper make better use of the characteristics of dynamic basic blocks. Experiments prove that our approach can effectively reduce test inputs, and the fault-localization effectiveness after reduction almost does not decline.(2)This paper proposes a statement suspiciousness measurement method. Compared to existing measurement methods, this method utilizes full-coverage information to calculate suspiciousness of statements, takes the unexecuted statement information into account. Experiments prove the method produces good fault localization effectiveness.