Abstract:Collaborative software development significantly enhances development productivity by enabling multiple contributors to work concurrently on different branches. Despite these advantages, such collaboration often increases the likelihood of causing conflicts. Resolving these conflicts brings huge challenges, primarily due to the necessity of comprehending the differences between conflicting versions. Researchers have explored various automatic conflict resolution techniques, including unstructured, structured, and learning-based approaches. However, these techniques are mostly heuristic-based or black-box in nature, which means they do not attempt to solve the root cause of the conflicts, i.e., the existence of different program behaviors exhibited by the conflicting versions. In this work, we propose sMerge, a novel conflict resolution approach based on the semantics of program behaviors. We first give the formal definition of the merge conflict problem as well as the specific conditions under which conflicts happen and the criteria employed to select certain version as the resolution. Based on the definition, we propose to resolve the conflicts from the perspective of program behaviors. In particular, we argue that the key to resolving conflicts is identifying different program behaviors, and thus can be solved through targeted test generation. We conduct an extensive evaluation of sMerge using a comprehensive dataset of conflicts sourced from various projects. Our results show that sMerge can effectively solve the merge problem by employing different test generation techniques, including search-based, GPT-based, and manual testing. We remark that sMerge provides a way to understand the program behavior differences through testing, which not only allows us to solve the merge problem soundly but also enables the detection of incorrect ground truths provided by developers, thereby enhancing the reliability of the merge process.

Git Merge Conflict Resolution Leveraging Strategy Classification and LLM

Merge Conflict Resolution: Classification or Generation?

Early Prediction of Merged Code Changes to Prioritize Reviewing Tasks

Revisiting the Conflict-Resolving Problem from a Semantic Perspective

Predicting Merge Conflicts in Collaborative Software Development

Can Pre-trained Language Models be Used to Resolve Textual and Semantic Merge Conflicts?

Automatic Detection and Resolution of Software Merge Conflicts: Are We There Yet?

A Characterization Study of Merge Conflicts in Java Projects

CONGRA: Benchmarking Automatic Conflict Resolution

Conflict Resolution for Structured Merge Via Version Space Algebra

WizardMerge -- Save Us From Merging Without Any Clues

Evaluation of Version Control Merge Tools

SoManyConflicts: Resolve Many Merge Conflicts Interactively and Systematically

Can Program Synthesis be Used to Learn Merge Conflict Resolutions? An Empirical Analysis

IntelliMerge: a Refactoring-Aware Software Merging Technique

Automated Commit Message Generation with Large Language Models: An Empirical Study and Beyond

Conflicts Classification and Solving for Collaborative Feature Modeling.

Composite-based Conflict Resolution in Merging Versions of UML Models

MAGIS: LLM-Based Multi-Agent Framework for GitHub Issue Resolution

Arcee's MergeKit: A Toolkit for Merging Large Language Models

GitAgent: Facilitating Autonomous Agent with GitHub by Tool Extension