A Fault Localization Method Based on Dynamic Failed Execution Blocks

Software fault localization is the most tedious and time-consuming activity in program debugging. Spectrum-based software fault localization (SFL) is a typical light-weight automated diagnosis technique for boosting fault localization. In the previous work, the researchers has found that although there is no optimal formula, the accuracy of a specific SFL-based method may be affected by the used program spectra at different granularity levels, e.g., statements, methods, basic blocks. In this paper, therefore, we further explore the correlation between the spectra granularity and the fault localization accuracy, and then the dynamic failed execution block (DFEB), a novel granularity for program spectra, is introduced to improve the diagnostic accuracy of a specific SFL technique. Consequently, a new method for fault localization by using DFEBs is proposed. Finally, to validate the effectiveness of our proposed method, an experimental study on 8 classic SFL techniques and 11 well-known benchmark programs including Siemens suite, space, gzip, grep, and sed, is conducted in detail. Experimental results show that the new approach is more effective in most cases, in comparison with these subject SFL methods.