Probabilistic fault localisation
Efficient fault localisation is becoming increasingly important as software grows in size and complexity. In this paper we present a new formal framework, denoted probabilistic fault localisation (pfl), and compare it to the established framework of spectrum based fault localisation (sbfl). We forma...
| Main Authors: | , , |
|---|---|
| Format: | Conference item |
| Published: |
Springer
2016
|
| _version_ | 1797104824112644096 |
|---|---|
| author | Landsberg, D Chockler, H Kroening, D |
| author_facet | Landsberg, D Chockler, H Kroening, D |
| author_sort | Landsberg, D |
| collection | OXFORD |
| description | Efficient fault localisation is becoming increasingly important as software grows in size and complexity. In this paper we present a new formal framework, denoted probabilistic fault localisation (pfl), and compare it to the established framework of spectrum based fault localisation (sbfl). We formally prove that pfl satisfies some desirable properties which sbfl does not, empirically demonstrate that pfl is signifuicantly more effective at finding faults than all known sbfl measures in large scale experimentation, and show pfl has comparable efficiency. Results show that the user investigates 37% more code (and finds a fault immediately in 27% fewer cases) when using the best performing sbfl measures, compared to the pfl framework. Furthermore, we show that it is theoretically impossible to design strictly rational sbfl measures that outperform pfl techniques on a large set of benchmarks. |
| first_indexed | 2024-03-07T06:38:58Z |
| format | Conference item |
| id | oxford-uuid:f89f78f4-bf8f-4ee5-b6a3-66e5f38e3bec |
| institution | University of Oxford |
| last_indexed | 2024-03-07T06:38:58Z |
| publishDate | 2016 |
| publisher | Springer |
| record_format | dspace |
| spelling | oxford-uuid:f89f78f4-bf8f-4ee5-b6a3-66e5f38e3bec2022-03-27T12:51:46ZProbabilistic fault localisationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:f89f78f4-bf8f-4ee5-b6a3-66e5f38e3becSymplectic Elements at OxfordSpringer2016Landsberg, DChockler, HKroening, DEfficient fault localisation is becoming increasingly important as software grows in size and complexity. In this paper we present a new formal framework, denoted probabilistic fault localisation (pfl), and compare it to the established framework of spectrum based fault localisation (sbfl). We formally prove that pfl satisfies some desirable properties which sbfl does not, empirically demonstrate that pfl is signifuicantly more effective at finding faults than all known sbfl measures in large scale experimentation, and show pfl has comparable efficiency. Results show that the user investigates 37% more code (and finds a fault immediately in 27% fewer cases) when using the best performing sbfl measures, compared to the pfl framework. Furthermore, we show that it is theoretically impossible to design strictly rational sbfl measures that outperform pfl techniques on a large set of benchmarks. |
| spellingShingle | Landsberg, D Chockler, H Kroening, D Probabilistic fault localisation |
| title | Probabilistic fault localisation |
| title_full | Probabilistic fault localisation |
| title_fullStr | Probabilistic fault localisation |
| title_full_unstemmed | Probabilistic fault localisation |
| title_short | Probabilistic fault localisation |
| title_sort | probabilistic fault localisation |
| work_keys_str_mv | AT landsbergd probabilisticfaultlocalisation AT chocklerh probabilisticfaultlocalisation AT kroeningd probabilisticfaultlocalisation |