Comparing developer-provided to user-provided tests for fault localization and automated program repair
We have provided a self-contained virtual machine (VirtualBox). Once the VM is booted, it will automatically login, and the reviewers will find the artifact on the Desktop. Note that we ported the artifact to GNU Linux; the experiments reported in the paper ran on OSX.
Working with the artifact requires the use of the command line (open a terminal in the VM).
The provided artifact is reusable: * The extracted user-provided tests, corresponding annotations, and support scripts will be included in an upcoming Defects4J release (https://github.com/rjust/defects4j).
The infrastructures for performing fault localization and automated program repair experiments are publicly available (https://bitbucket.org/rjust/fault-localization-data and https://github.com/chrisparnin/repair-cluster) and maintained.
All scripts within the artifact are modular, with core functionality extracted into dedicated modules. This allows others to adapt specific analyses or implement new ones.