Murdock results

✔️ PASSED

5d7bff8 treewide: Update riot-sys

Artifacts

• Documentation preview

CI said no; apparently, SPI clocks are odd somewhere.

That's a lot of dependencies.

Yes it is. It makes sense to take a more fine-grained viewpoint, though:

• Many of those in Cargo.lock don't even show up in the Cargo tree (ie. they have no way of influencing the build): For example, windows-sys gets pulled in through libc and errno, but only when building on Windows.
• Many that do show up in the cargo tree are build tools, eg. bindgen. In other environments than cargo, we might just require a CLI tool to be installed and have the build system call out to it. On Rust, it is common for build.rs to just pull in a crate for that – both making it convenient for the user to not have to install an extra tool (like c2rust where sadly that all doesn't work) and for us to have good control of when things may break, but on the downside those are all listed as dependencies.
• A cargo tree --edges normal still shows a lot of crates, but of a more managable size. Many of those are interface-only crates (embedded-graphics, embedded-hal, coap-message, switch-hal, mutex-trait) – we could make those optional, but given they are rather cheap for all other aspects than the list, that'd just make maintenance harder.
  • Besides riot-wrappers, also coap-handler-implementations pulls in such a swath of (partially old, even dangerous) interfaces (eg. multiple versions of minicbor). They don't do anything (no code emitted from there) unless someone else comes with "hey I have data from that library, please send it", but once more, making all them optional would put an otherwise useless load on maintenance.

This can be unblocked by RIOT-OS/rust-riot-wrappers#144, which needs to be merged before I can unblock this PR.

The last patch updates riot-sys only to address the previous build failures.