Bug Mining

Mining reproducible bugs from real-world projects is difficult. Differing build systems, explicit and implicit dependencies, different versions of the underlying programming language, etc. all present challenges.

Defects4J is a collection of reproducible Java bugs from real-world projects which is widely used in evaluations in the software engineering research community for applications such as automatic program repair. I have been improving the bug-mining infrastructure in Defects4J to utilize a project’s original build system to reduce loss of bugs due to build issues as well as adding more support for rapid and automatic resolution of build issues. Additionally, I have been migrating Defects4J away from ant to native java and javac build and execution targets for mined projects to support more customized set-ups for software engineering researchers using Defects4J.