Answers are grounded in this paper and cite their sources.
Ensuring correctness in multi-threaded programs requires effective detection of concurrency bugs, which is challenging due to the nondeterminism of thread scheduling. Recently, there have been advancements in dynamic detectors based on various consistency models, assumptions, and algorithms. These detectors significantly improve concurrency coverage. However, due to limitations in the model assumptions, they are confined to the space obtained by reordering observed traces ( S ob ). It is important to investigate how different model assumptions impact detection results. Additionally, since each detector is evaluated on different sets of benchmark programs, it is unclear which detectors and algorithms are more effective and efficient. In this paper, we provide a systematic analysis of the reordering of observed traces by establishing a unified framework of four trace-restricted consistency models, whose detection spaces are strictly confined to S ob . We evaluate 14 recently published dynamic data race detectors and one classic happens-before-based detector against these models and answer seven research questions through theoretical analysis and experimental evaluation. This offers a new perspective for understanding and improving existing methodologies. Finally, we conceptualize a new model that extends beyond S ob by incorporating program semantics, serving as a direction for future research to bridge the gap between observed traces and the full program execution space.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: