Precise Propagation of Fault-Failure Correlations in Program Flow Graphs

Abstract:Statistical fault localization techniques find  suspicious faulty program entities in programs by comparing passed and failed executions.  Existing studies show that such  techniques can be promising in locating program faults. However, coincidental correctness and execution crashes may make program entities indistinguishable in the execution spectra under study, or cause inaccurate counting, thus severely affecting the precision of existing fault localization techniques. In this paper, we propose  a BlockRank  technique, which  calculates,  contrasts, and propagates the mean edge profiles between passed and failed executions to alleviate the impact of coincidental correctness. To address  the issue of execution crashes, BlockRank identifies  suspicious basic  blocks by modeling how each basic  block contributes  to  failures by  apportioning their  fault relevance to surrounding  basic  blocks in  terms of the  rate of successful transition observed from passed and failed executions.  BlockRank is empirically shown to  be  more  effective than nine representative  techniques on  four real-life mediumsized programs.
Grants:NSFC 61003027, GRF 111410, GRF 717308, CityU 7002673
Links: PDF
Related Papers:-

Selected Tags

Tag Groups


ACM SigSoft
IEEE Software Engineering Online