Keshmesh: Bringing Advanced Static Analysis to Concurrency Bug Pattern Detectors

Vakilian, Mohsen; Negara, Stas; Tasharofi, Samira; Johnson, Ralph E.

Permalink

https://hdl.handle.net/2142/44837 Copy

Description

Title

Keshmesh: Bringing Advanced Static Analysis to Concurrency Bug Pattern Detectors

Author(s)

• Vakilian, Mohsen
• Negara, Stas
• Tasharofi, Samira
• Johnson, Ralph E.

Issue Date

2013-05

Keyword(s)

• bug
• pattern
• detector
• fixer
• concurrency
• parallelism
• static analysis
• program analysis

Abstract

Bug patterns are coding idioms that may make the code less maintainable or turn into bugs in future. The state-of-the-art tools for detecting concurrency bug patterns (CBPs) perform simple, intraprocedural analyses. While this simplicity makes the analysis fast, it does not provide protection against CBPs that involve aliasing or multiple methods. This paper introduces a practical and extensible framework, Keshmesh, which employs advanced static analysis for detecting CBPs. Keshmesh builds upon the points-to analysis of WALA, a static analysis framework, and the user interface of FindBugs, a popular bug pattern detection tool. Keshmesh detects five CBPs using interprocedural analyses and fixes two of them. The challenges in automatic detection of these CBPs include reducing the rate of false positives, scaling to large projects, and accurate propagation of unsafe accesses along the call graph up to synchronization constructs. Keshmesh found 40 previously unknown CBP instances and only 12 false positives in six open-source projects. Programmers fixed 11 of the 20 issues we reported. These results show that Keshmesh is applicable to large projects, finds CBPs that programmers want to fix, and reports few false positives.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/44837

Copyright and License Information

Copyright is held by the author/owner(s).

Owning Collections