An in-depth study of software library upgrade dependency issues

• Main Author: Leow, Wei Jie
• Other Authors: Li Yi (SCSE)
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/181284
• Institution: Nanyang Technological University

With the rapid advancements in Artificial Intelligence and Large Language Models, the potential to leverage cutting-edge AI for addressing code-related issues is continually growing. This study explores the potential of utilizing AI and evaluate its effectiveness in resolving software dependency upgrade incompatibility issues. We developed an Automated Repair Program SLUDI and evaluated the effectiveness of three different large language models against a dataset of 30 upgrade incompatibility issues. Results show that the models achieved an average of 38.9% identification rate and 41.1% correctness rate. The evaluation results show that given only the exception information, source code of the method, context of the library upgraded, the large language models are not very effective in resolving software dependency upgrade incompatibility issues. Future research recommendations include extracting and providing more crucial information to the AI, enabling them to gain a deeper understanding of the project, which could improve their ability to identify and fix the incompatibility issue with greater accuracy. Exploring automatic repair methods could also be implemented to enhance efficiency and reduce the potential for human error.