Build Predictor: More Accurate Missed Dependency Prediction in Build Configuration Files
Conference Proceeding Article
Software build system (e.g., Make) plays an important role in compiling human-readable source code into an executable program. One feature of build system such as make-based system is that it would use a build configuration file (e.g., Make file) to record the dependencies among different target and source code files. However, sometimes important dependencies would be missed in a build configuration file, which would cause additional debugging effort to fix it. In this paper, we propose a novel algorithm named Build Predictor to mine the missed dependncies. We first analyze dependencies in a build configuration file (e.g., Make file), and establish a dependency graph which captures various dependencies in the build configuration file. Next, considering that a build configuration file is constructed based on the source code dependency relationship, we establish a code dependency graph (code graph). Build Predictor is a composite model, which combines both dependency graph and code graph, to achieve a high prediction performance. We collected 7 build configuration files from various open source projects, which are Zlib, putty, vim, Apache Portable Runtime (APR), memcached, nginx, and Tengine, to evaluate the effectiveness of our algorithm. The experiment results show that compared with the state-of-the-art link prediction algorithms used by Xia et al., our Build Predictor achieves the best performance in predicting the missed dependencies.
Software and Cyber-Physical Systems
IEEE 38th Annual International Computer Software and Applications Conference: Proceedings: 27-29 July 2014, Västerås, Sweden
IEEE Computer Society
City or Country
Los Alamitos, CA
Zhou, Bo; Xin, Xia; LO, David; and Wang, Xinyu.
Build Predictor: More Accurate Missed Dependency Prediction in Build Configuration Files. (2014). IEEE 38th Annual International Computer Software and Applications Conference: Proceedings: 27-29 July 2014, Västerås, Sweden. 53-58. Research Collection School Of Information Systems.
Available at: http://ink.library.smu.edu.sg/sis_research/2417