David Airlie, a maintainer of the graphics driver stack in the Linux kernel, announced the implementation of an auxiliary automated patch review system, which utilizes the AI platform Claude Opus 4.6 to check changes. The system is introduced as an experiment to evaluate the feasibility of using AI to optimize the workflows of Red Hat employees.
AI reviews cover changes submitted for inclusion in the DRM (Direct Rendering Manager) subsystem. The publication of results is organized using the mailing lists used in the Linux kernel development. To avoid cluttering the main mailing list, which manages the development of the DRM subsystem, a separate mailing list for AI reports, drm-ai-reviews, was created.
It is expected that patch authors can use AI reviews as an initial feedback. The analysis of such reviews is not enforced, and authors are free to consider AI reviews at their discretion. However, if AI reviews identify regressions overlooked by patch authors, it is possible that the analysis of AI reviews will be mandated more insistently in the future.
Initially, to organize the review in Claude, Dave attempted to use a set of prompts proposed by Chris Mason, which clarify the specifics of subsystem and protocol operations and allow the AI assistant to better understand the context during reviews. Chris's set was intended for identifying regressions and analyzing patches when applied sequentially to the current tree of the kernel code and for final regression testing in the latest patch.
Ultimately, Dave had to refine and expand Chris's prompts, as it was necessary for the AI to review a series of patches as a whole, followed by a deep analysis of each patch individually. In addition to the current kernel code tree, the review was extended to the drm-next branch. Dave also prepared tools for integration with the Claude AI service and the organization of the mailing list.
Forrás: opennet.ru
