We present SwingArena, an adversarial evaluation framework for Large Language Models (LLMs) that approximates real-world software development workflows. Unlike traditional static benchmarks, SwingArena models the collaborative process of software iteration by pairing LLMs as submitters, who generate patches, and reviewers, who create test cases and verify the patches through continuous integration (CI) pipelines. To support these interactive evaluations, we introduce a retrieval-augmented code generation (RACG) module that handles long-context challenges by providing relevant code snippets from large codebases across multiple programming languages (C++, Python, Rust, and Go). Our adversarial evaluation can surface limitations that are often overlooked by traditional evaluation settings. Our experiments, using over 400 high-quality real-world GitHub issues selected from a pool of 2,300 issues, indicate differing behavioral tendencies across models in patch generation versus validation. SwingArena offers a scalable and extensible approach to evaluating LLMs in CI-driven software development settings.