ACM SIGSOFT Distinguished Paper Award for the 28th ACM SIGSOFT  International  Symposium on Software Testing and Analysis

History-driven Build Failure Fixing: How Far Are We?

Yiling Lou, Junjie Chen, Lingming Zhang, Dan Hao, Lu Zhang

Build systems break frequently during software evolution, which makes automated build fixing techniques in huge demand. In this paper, we propose a lightweight build-failure fixing technique, which does not rely on history data but instead simply utilizes the present information of the build code, build log and external build-related resources. The proposed technique includes two phases: fault localization and patch generation. In particular, the fault-localization phase analyzes error logs of the given build failures to extract error information, and then localizes the possible buggy locations via inter-procedural data-flow analysis; the patch-generation phase generates patch candidates based on search-based strategy. The proposed technique successfully fixes 2X more reproducible build failures than the state-of-art technique.

2019/7/18

The 28th ACM SIGSOFT International Symposium on Software Testing and Analysis(ISSTA 2019)

https://conf.researchr.org/home/issta-2019

https://conf.researchr.org/details/issta-2019/issta-2019-Technical-Papers/3/History-driven-Build-Failure-Fixing-How-Far-Are-We-