Fray - An Efficient General-Purpose Concurrency Testing Platform for the JVM

Explore a groundbreaking conference presentation from OOPSLA 2025 that introduces an innovative solution to one of software engineering's most persistent challenges: testing concurrent Java applications. Learn about the fundamental limitations of existing controlled concurrency testing (CCT) approaches when applied to real-world JVM programs, including issues with concurrency mocking, VM modifications, and OS-level scheduling that restrict their practical applicability. Discover how researchers from Carnegie Mellon University developed a first-principles approach to identify the core requirements and design considerations necessary for effective concurrency testing on arbitrary JVM applications. Understand the novel "shadow locking" mechanism that enables faithful orchestration of thread interleavings without replacing existing concurrency primitives, allowing for comprehensive testing of program behaviors beyond simple data races. Examine the platform's support for multiple sophisticated search algorithms including random walk, PCT (Probabilistic Concurrency Testing), POS (Partial-Order Sampling), and SURW (Selectively Uniform Random Walk). Analyze empirical results demonstrating significant improvements in bug detection rates, with 70% more bugs found compared to JPF and 77% more than RR's chaos mode across 53 benchmark programs. Review real-world validation through testing of 2,664 tests from major Apache projects (Kafka, Lucene) and Google Guava, resulting in the discovery of 18 previously unknown concurrency bugs affecting 371 existing tests. Gain insights into how this platform bridges the gap between academic research and industrial practice, providing developers with advanced testing capabilities while offering researchers a robust evaluation framework for large-scale concurrency testing studies.