Spectrum-Based Fault Localization in Regression Testing

Abstract

In maintenance, regression testing is used to revalidate modified software which is also an expensive and frequently executed process. In conventional regression testing approach, developers generally re-run all test suite or selectively run a sub-set of existing test cases on the modified version of program. After executing the test cases over program, developer may reveal the regression faults due to changes in code and use these non obsolete test cases from pre-existing test suite to explore and eradicate regression faults. This paper addresses the fundamental limitations of conventional regression testing approach. To overcome these limitations, this paper presents a spectrum-based fault localization strategy by which the stated limitations are resolved in effective manner. Spectrum-based fault localization strategy utilizes various program spectra to identify the behavioral differences between old and new version of the program under test. This comparison is also useful in pinpointing the cause of failures or errors and presence of difference in program spectra may indicate those test cases for which the construction of expected output or oracle or specification is not needed. It is very expensive to compute and verify expected output and some times it is impossible to compute expected outputs for non-trivial programs. This approach is used in regression testing to identify behavioral differences between old and new version of a program.