AN IMPLEMENTATION OF SAVANT, BUG LOCALIZATION TECHNIQUE FOR EXPERIMENT REPRODUCTION PURPOSES

• Main Author: Erdiza Yogatama, Kevin
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/48278
• Institution: Institut Teknologi Bandung

There is a number of proposed approaches to solve a bug localization problem, and one of which is a technique named Savant. The experiment done to Savant shows that the technique outperforms several other bug localization techniques. The publication that proposes Savant provides the source code for Savant so the experiment results can be reproduced. However, several users have pointed out that the provided code and data have problems. Since the reproducibility of an experiment is an important aspect in determining whether the experiment’s result is acceptable or not, an attempt to implement a solution to support the reproduction of Savant’s experiment is done in this final year project. To identify the problem, Artifact Review dan Badging , a policy from ACM, is used. Savant’s source code then as an artifact is evaluated with 4 criteria that make a functional artifact determined by Artifact Review dan Badging . Those 4 criteria are documented , consistent , complete , and exercisable . After that, Resavant, a new reimplementation of Savant, is developed with an aim to achieve those 4 criteria. Resavant that has been developed has only achieved 3 criteria of a functional artifact: documented , complete , and exercisable . Resavant didn’t achieve the consistent criteria because of several technical difficulties: very long execution time for some inputs and unresolved incompatibilities between Resavant’s program component. However, Resavant has successfully reproduced one experiment scenario and produce a result similar to the original Savant’s experiment result.