A Bot for Patch Linkage Detection In Modern Code Review Platforms

Code review tools are essential in software quality assurance for projects like Android and OpenStack, and companies such as Google and Microsoft. OpenStack alone has over 100,000 contributors and 600+ repositories, where parallel patch submissions often lead to redundant work. Studies indicate that identifying duplicate patches takes, on average, 2.5 reviewers and 5.2 comments per review, delaying processes and raising maintenance costs.

Detecting and linking patches early helps highlight dependencies, broader context, or alternative solutions, reducing redundant efforts and streamlining reviews. This capstone project aims to develop a tool that automates the detection of patch linkages using advanced machine learning techniques. It will evaluate the performance of large language models (LLMs), pre-trained transformers, and basic ML models in detecting patch linkages in real-world open-source projects like OpenStack and Eclipse, which use Gerrit for code review.

Faculty Supervisor:

Moataz Chouchen

Student:

Partner:

École nationale des sciences de l'informatique

Discipline:

Computer science

Sector:

Education

University:

Concordia University

Program: