Probabilistic Fault Localisation

David Landsberg; Hana Chockler; Daniel Kroening

doi:10.1007/978-3-319-49052-6_5

Probabilistic Fault Localisation

David Landsberg^*, Hana Chockler, Daniel Kroening

^*Corresponding author for this work

University of Oxford

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

2 Citations (Scopus)

222 Downloads (Pure)

Abstract

Efficient fault localisation is becoming increasingly impor- tant 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 sig- nificantly 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.

Original language	English
Title of host publication	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Subtitle of host publication	Hardware and Software: Verification and Testing
Publisher	Springer‐Verlag Berlin Heidelberg
Pages	65-81
Number of pages	17
Volume	10028 LNCS
ISBN (Print)	9783319490519
DOIs	https://doi.org/10.1007/978-3-319-49052-6_5
Publication status	Published - 1 Nov 2016
Event	12th International Haifa Verification Conference, HVC 2016 - Haifa, Israel Duration: 14 Nov 2016 → 17 Nov 2016

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10028 LNCS
ISSN (Print)	03029743
ISSN (Electronic)	16113349

Conference

Conference	12th International Haifa Verification Conference, HVC 2016
Country/Territory	Israel
City	Haifa
Period	14/11/2016 → 17/11/2016

Keywords

Fault localisation
Spectrum based fault localisation
Triage and debug technologies

Access to Document

10.1007/978-3-319-49052-6_5Licence: Other

Probabilistic Fault Localisation_LANDSBERG_Firstonline1November2016_GREEN AAM
"The final publication is available at Springer via http://dx.doi.org10.1007/978-3-319-49052-6_5".
Accepted author manuscript, 834 KB

Cite this

Landsberg, D., Chockler, H., & Kroening, D. (2016). Probabilistic Fault Localisation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Hardware and Software: Verification and Testing (Vol. 10028 LNCS, pp. 65-81). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10028 LNCS). Springer‐Verlag Berlin Heidelberg. https://doi.org/10.1007/978-3-319-49052-6_5

Landsberg, David ; Chockler, Hana ; Kroening, Daniel. / Probabilistic Fault Localisation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Hardware and Software: Verification and Testing. Vol. 10028 LNCS Springer‐Verlag Berlin Heidelberg, 2016. pp. 65-81 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inbook{c18bac0d184f48cd839895c2e9de9bbe,

title = "Probabilistic Fault Localisation",

abstract = "Efficient fault localisation is becoming increasingly impor- tant 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 sig- nificantly 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.",

keywords = "Fault localisation, Spectrum based fault localisation, Triage and debug technologies",

author = "David Landsberg and Hana Chockler and Daniel Kroening",

year = "2016",

month = nov,

day = "1",

doi = "10.1007/978-3-319-49052-6_5",

language = "English",

isbn = "9783319490519",

volume = "10028 LNCS",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer‐Verlag Berlin Heidelberg",

pages = "65--81",

booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

note = "12th International Haifa Verification Conference, HVC 2016 ; Conference date: 14-11-2016 Through 17-11-2016",

}

Landsberg, D, Chockler, H & Kroening, D 2016, Probabilistic Fault Localisation. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Hardware and Software: Verification and Testing. vol. 10028 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10028 LNCS, Springer‐Verlag Berlin Heidelberg, pp. 65-81, 12th International Haifa Verification Conference, HVC 2016, Haifa, Israel, 14/11/2016. https://doi.org/10.1007/978-3-319-49052-6_5

Probabilistic Fault Localisation. / Landsberg, David; Chockler, Hana; Kroening, Daniel.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Hardware and Software: Verification and Testing. Vol. 10028 LNCS Springer‐Verlag Berlin Heidelberg, 2016. p. 65-81 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10028 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

TY - CHAP

T1 - Probabilistic Fault Localisation

AU - Landsberg, David

AU - Chockler, Hana

AU - Kroening, Daniel

PY - 2016/11/1

Y1 - 2016/11/1

N2 - Efficient fault localisation is becoming increasingly impor- tant 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 sig- nificantly 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.

AB - Efficient fault localisation is becoming increasingly impor- tant 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 sig- nificantly 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.

KW - Fault localisation

KW - Spectrum based fault localisation

KW - Triage and debug technologies

UR - http://www.scopus.com/inward/record.url?scp=84995380264&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-49052-6_5

DO - 10.1007/978-3-319-49052-6_5

M3 - Chapter

AN - SCOPUS:84995380264

SN - 9783319490519

VL - 10028 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 65

EP - 81

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer‐Verlag Berlin Heidelberg

T2 - 12th International Haifa Verification Conference, HVC 2016

Y2 - 14 November 2016 through 17 November 2016

ER -

Landsberg D, Chockler H, Kroening D. Probabilistic Fault Localisation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Hardware and Software: Verification and Testing. Vol. 10028 LNCS. Springer‐Verlag Berlin Heidelberg. 2016. p. 65-81. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-49052-6_5

Probabilistic Fault Localisation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this