Meeting 2026-02-03

[eldruin]

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

[BartMassey]

The REWG Bookshelf Archaeology meeting will be 28 February 8:00-10:00PM Berlin Time (2000-2200 CET). Please do plan to attend if you'd like to help us excavate the stacks! Looks like we have four right now, but can use as many more as want to participate: anyone with knowledge and interest is welcome, not just REWG members.

Let's try Jitsi this time: https://meet.jit.si/REWG-Bookshelf-Archaeology-2026 . As a backup we can use Google Meet https://meet.google.com/wdo-nmhb-msa

[BartMassey]

The Rust Embedded Google Summer of Code is on! Join us at #re-gsoc:matrix.org to join the discussion.

If you want to propose a project idea, please PR https://github.com/rust-embedded/google-summer-of-code after reading carefully. If you want to help mentor a project (your own or others) please let us know on the RE-GSOC Matrix channel.

Hoping to merge and PR our fork to the Rust Project in the next couple of days.

[jonathanpallant]

#[unsafe(naked)] doesn't play nice with #[instruction_set], causing Thumb code to execute in Arm mode (or vice-versa), which is very bad.

rust-lang/rust#151946

[jonathanpallant]

I've got three AArch32 PRs that need review: https://github.com/rust-embedded/aarch32/pulls

[jonathanpallant]

I want to nudge this Cortex-M API for MRC/MCR system register reads: rust-embedded/cortex-m#531

• Is it OK to feature-gate something that requires an MSRV (I think so?)
• Is it OK to let people safely read any system register (can they have side-effects?) or write any system register (that seems bad?)

I note that CP8 to CP15 are reserved for Arm's use, and only CP10 and CP11 are defined (for the floating point unit), which don't need MRC/MCR calls. So this API is only for vendor-defined CP0 to CP7 extensions. For example, the Raspberry Pi RP2350 lets you read 64-bits of GPIO with an MRRC on CP0.

[jonathanpallant]

cortex-m has an MSRV of 1.61 and now the dependencies don't build (proc-macro2 needs 1.68). Maybe our MSRV should just be "stable minus 3", to give people 18 weeks to update to latest stable? Anyone who needs an old version of the compiler can just pin to an old version of the crate (and handle hotfixes themselves).

[jamesmunns]

I think the last decision was "update it whenever we have a good reason too, no more recent than stable". That's just from memory tho.

[therealprof]

I totally agree, but that's not how we started. 😅

[jannic]

Or use resolver.incompatible-rust-versions=fallback with an MSRV of at least 1.84?

[BartMassey]

@jannic Holy undiscoverable power features, Batman! That is really convenient, and I've never heard of it.

That said, I'm with @jonathanpallant in thinking that stable - 3 should be plenty. I would also like to see rolling releases every six months minimum, updating at least the dependencies at each release. I am willing to help with this.

[jannic]

Holy undiscoverable power features, Batman!

The nice thing about this feature is that you don't need to discover it: With edition = "2024", it's the default setting.

[thejpster]

Just a reminder that the rust-embedded-wg matrix channel is a great place to post notifications of PRs that you've opened and would like reviewing, and a great place to be if you want to be notified when there are PRs waiting for your attention.

[jannic]

Triage

• PeripheralSpec svd2rust#946 needs a reviewer
• disable respace svd2rust#879 is approved but wasn't merged. Why? By now, it has merge conflicts. Should these be fixed or is there another reason why this shouldn't be merged?