Leonardo Murta

Heleno de Souza Campos Junior, Leonardo Gresta Paulino Murta

Context: The resolution of software merge conflicts is being reshaped by two competing paradigms: generative approaches based on Large Language Models (LLMs) and optimization approaches from Search-Based Software Engineering (SBSE). While tools from both paradigms have shown promise, their relative strengths, weaknesses, and trade-offs are not yet well understood. Objective: This paper presents the first in-depth empirical study directly comparing these paradigms to identify their capabilities and limitations in real-world scenarios. Method: We evaluated MergeGen, a state-of-the-art LLM-based tool, against SBCR, a novel SBSE approach employing a Random Restart Hill Climbing (RRHC) algorithm. The comparison used thousands of real-world conflicts from open-source projects written in Java, C#, JavaScript, and TypeScript. Results: Our findings reveal fundamental trade-offs. The LLM paradigm excels at resolving conflicts with imbalanced content by leveraging learned patterns. However, it struggles with non-English content and large inputs, which can lead to truncated or empty resolutions. Conversely, the SBSE paradigm demonstrates superior generalization across datasets and performs best on balanced conflicts, highlighting its potential as a robust, data-independent alternative. Conclusions: Neither paradigm is a silver bullet. Our findings highlight context-dependent strengths and motivate the development of hybrid systems that combine the complementary capabilities of LLM and SBSE approaches to create more robust and reliable merge conflict resolution tools.

Felipe Curty, Troy Kohwalter, Vanessa Braganholo et al.

Nowadays, quickly evolving and delivering software through a continuous delivery process is a competitive advantage and a way to keep software updated in response to the frequent changes in customers' requirements. However, the faster the software release cycle, the more challenging to track software evolution. In this paper, we propose Releasy, a tool that aims at supporting projects that use continuous delivery by generating and reporting their release provenance. The provenance generated by Releasy allows graphical visualization of the software evolution and supports queries to discover implicit information, such as the implemented features of each release and the involved developers. We also show in this paper a preliminary evaluation of Releasy in action, generating the changelog of an open source project with the provenance collected by our tool.

Erica Mourão, João Felipe Pimentel, Leonardo Murta et al.

Context: When conducting a Systematic Literature Review (SLR), researchers usually face the challenge of designing a search strategy that appropriately balances result quality and review effort. Using digital library (or database) searches or snowballing alone may not be enough to achieve high-quality results. On the other hand, using both digital library searches and snowballing together may increase the overall review effort. Objective: The goal of this research is to propose and evaluate hybrid search strategies that selectively combine database searches with snowballing. Method: We propose four hybrid search strategies combining database searches in digital libraries with iterative, parallel, or sequential backward and forward snowballing. We simulated the strategies over three existing SLRs in SE that adopted both database searches and snowballing. We compared the outcome of digital library searches, snowballing, and hybrid strategies using precision, recall, and F-measure to investigate the performance of each strategy. Results: Our results show that, for the analyzed SLRs, combining database searches from the Scopus digital library with parallel or sequential snowballing achieved the most appropriate balance of precision and recall. Conclusion: We put forward that, depending on the goals of the SLR and the available resources, using a hybrid search strategy involving a representative digital library and parallel or sequential snowballing tends to represent an appropriate alternative to be used when searching for evidence in SLRs.

Tayane Silva Fernandes de Moura, Leonardo Gresta Paulino Murta

Developers that use version control systems can work in parallel with other developers and merge their versions afterwards. Sometimes these merges fail, demanding manual intervention to resolve conflicts. Some studies aim at analyzing the merges failures, however, there is a lack of tool support in the literature to measure the merge effort, jeopardizing such kind of analyses. In this article, we propose a technique and its companion tool for analyzing Git repositories and providing metrics related to the merge effort. We evaluated our tool over five projects, showing that rework and wasted work happens in, approximately, 10% to 30% of the projects. Moreover, the average of actions of these efforts is almost the same.