Bug Signature Minimization and Fusion
Debugging is a time-consuming activity. To help in debugging, many approaches have been proposed to pinpoint the location of errors given labeled failures and correct executions. While such approaches have been shown to be accurate, at times the location alone is not sufficient in helping programmer...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Institutional Knowledge at Singapore Management University
2011
|
| Subjects: | |
| Online Access: | https://ink.library.smu.edu.sg/sis_research/1403 http://dx.doi.org/10.1109/HASE.2011.36 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Singapore Management University |
| Language: | English |
| Summary: | Debugging is a time-consuming activity. To help in debugging, many approaches have been proposed to pinpoint the location of errors given labeled failures and correct executions. While such approaches have been shown to be accurate, at times the location alone is not sufficient in helping programmers understand why the bug happens and how to fix it. Furthermore, a single location might not be powerful enough to discriminate failures from correct executions. To address the above challenges, there have been recent studies on extracting bug signatures which are composed of multiple locations appearing together in a particular order signifying an occurrence of a bug. The latest study on bug signatures by Cheng et al. models program executions as graphs. Two sets of graphs corresponding to failures and correct executions are then contrasted to extract the most discriminative connected sub graphs serving as bug signatures. However, there are two limitations: (1) returned signatures might not be minimal and (2) they can only capture localized bug context. In this work, we develop a signature minimization technique to capture minimal discriminative signatures. Also, we propose a technique of signature fusion to fuse disconnected sub graphs so that our method can capture bug contexts spanning multiple locations. Experimental study on Siemens and Space dataset shows the effectiveness of the proposed bug signature minimization and fusion techniques. Comparing with the state-of-the-art bug signature mining technique, we reduce the number of bugs missed by up to 57.7%, and reduce the average number of nodes traversed by up to 85.6%. |
|---|