DynaMine
Benjamin Livshits, Thomas Zimmermann · ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE) · 2005
A great deal of attention has lately been given to addressing software bugs such as errors in operating system drivers or security bugs. However, there are many other lesser known errors specific to individual applications or APIs and these violations of application-specific coding rules are responsible for a multitude of errors. In this paper we propose DynaMine, a tool that analyzes source code check-ins to find highly correlated method calls as well as common bug fixes in order to automatically discover application-specific coding patterns. Potential patterns discovered through mining are passed to a dynamic analysis tool for validation; finally, the results of dynamic analysis are presented to the user.The combination of revision history mining and dynamic analysis techniques leveraged in DynaMine proves effective for both discovering new application-specific patterns and for finding errors when applied to very large applications with many man-years of development and debugging effort behind them. We have analyzed Eclipse and jEdit, two widely-used, mature, highly extensible applications consisting of more than 3,600,000 lines of code combined. By mining revision histories, we have discovered 56 previously unknown, highly application-specific patterns. Out of these, 21 were dynamically confirmed as very likely valid patterns and a total of 263 pattern violations were found.
Cited by 330 works. doi:10.1145/1081706.1081754
Authors
References
- Fast Algorithms for Mining Association Rules in Large Databases
- Fast algorithms for mining association rules
- Efficient algorithms for discovering association rules
- Eraser
- A First Step Towards Automated Detection of Buffer Overrun Vulnerabilities.
- Checking system rules using system-specific, programmer-written compiler extensions
- A static analyzer for finding dynamic programming errors
- Bugs as deviant behavior
- Dynamically discovering likely program invariants to support program evolution
- Detecting format string vulnerabilities with type qualifiers