Meeting 2025-10-07

[jonathanpallant]

This is the agenda for the next meeting. Please add anything you'd like to discuss below!

• Meetings: Tuesday 8pm Europe/Berlin time
• Join the Matrix chat
• IRC/Matrix logs

[jonathanpallant]

Arm have officially become maintainers of many of the 32-bit Arm targets (although not all).

In addition, David Wood has applied to become a member of the Rust Embedded Working Group's Arm Team, on Arm's behalf. See #873 for the T-Arm voting page.

[romancardenas]

I would like to discuss two points:

• First, just a quick review of cortex-m-types crate: standard traits and result handling for Cortex-M targets cortex-m#560
• Then, I would like to discuss/get some feedback regarding the WG and AI-assisted contributions (see @rmsyn comments in Add mtopi CSR riscv#347). I am not completely against to using AI tools, but I think it is important for us to discuss how we want to proceed in these cases and apply a uniform policy. Also, there are millions of ways of using AI for coding, and maybe some of them should be banned/controlled at a WG level.

[jonathanpallant]

Shout out to the proposed LLVM AI policy.

https://discourse.llvm.org/t/rfc-llvm-ai-tool-policy-start-small-no-slop/88476

[orhun]

FWIW at @ratatui we have recently updated our contribution guidelines about AI usage: https://github.com/ratatui/ratatui/blob/main/CONTRIBUTING.md#ai-generated-content

Somewhat related, we also created copilot instructions for better instructing the copilot integration on GitHub.

[rmsyn]

I'm glad to see LLVM and Ratatui take a clear stance on AI, though I wish they were stronger. For instance, the last point in Ratatui's guidance is unenforceable. These LLMs are basically the perfect tool for circumventing plagiarism of licensed material, there has already been a number of lawsuits about it. It requires an inordinate amount of effort to verify the provenance of LLM output, even for "open" models.

There is a huge difference between classic IDE tools which may use machine learning techniques to perform auto-complete / auto-suggestion based on scanning the project locally, and these large-scale LLMs which have been trained on the entire corpus of human output on the internet, and reside in datacenters that collectively eat up the energy budget of nations. These are anti-human, anti-expert technologies.

[adamgreig]

On rust-embedded/cortex-m#560, we concluded:

• use arch-types for the error/traits crate
• use arch-peripherals for the register definitions of core peripherals
• use arch for the high level "hal" methods, asm etc
• use arch-rt for the runtime
• riscv-pac will move to riscv-types, cortex-m-types will start to exist

[jannic]

What's the status of fixedvec?
The same question came up 5 years ago, rust-embedded/fixedvec-rs#20, but I don't see any activity after that date. Does the crate provide anything that's not also part of heapless?