Implement IoPin

[olback]

Implements the IoPin trait from embedded-hal.

[eldruin]

I see here the problem that if setting the state fails, the pin is dropped.

[olback]

That is something I did not think about. I mostly copied all the implementations from stm32-rs/stm32f4xx-hal where they use into_push_pull_output_in_state which would solve this I think.

[richardeoin]

The set_low and set_high methods implemented here are infallible (use the Never type as an error), so set_state would also be infallible.

Maybe we should implement our own infallible set_state (as stm32f4xx-hal does) and use that?

[richardeoin]

I'd not seen the IoPin trait yet, but it's certainly interesting.

This PR should also bump the embedded-hal version in Cargo.toml to 0.2.6, since 0.2.{4,5} would no longer be supported.

[olback]

This PR should also bump the embedded-hal...

Should the other outdated dependencies also be updated?

[richardeoin]

No, the issue is that with this change the crate would no longer compile with embedded-hal 0.2.4. Therefore it's necessary to increase the minimum patch version (3rd number) so Cargo explicitly knows it can't use 0.2.4 but has to download 0.2.6 (or 0.2.7, 0.2.8 ...). For other dependencies that have a new major or minor version, it's good to update them but probably in a separate PR.

If there's a good solution to avoid ever returning Err from the into_output_pin methods, then I think we can get this merged

[olback]

Would that really be necessary though? set_state never fails, it's just the trait impl that requires a Result.

[richardeoin]

Hmm, yes there is the type Error = Never; in the implementation. I would still replace Self::Error with Never for consistency with every other implementation in this file, and even replace the ? with an .unwrap() to make completely clear that this never fails.

[olback]

Somewhat unrelated, is there a reason for using enum Never {} over core::convert::Infallible?

[richardeoin]

Looks good!

My best guess is that enum Never was the solution before convert::Infalliable was stabilised, and there was an expectation that convert::Infalliable would soon be replaced with the actual ! never type. However stabilisation of ! seems to have hit some roadblocks, and enum Never works ok, so it's not been changed.

[richardeoin]

bors r+

[bors]

Build succeeded:

• ci (1.52.0, stm32h735)
• ci (1.52.0, stm32h743)
• ci (1.52.0, stm32h743v)
• ci (1.52.0, stm32h747cm7)
• ci (1.52.0, stm32h753)
• ci (1.52.0, stm32h753v)
• ci (1.52.0, stm32h7b0)
• ci (1.52.0, stm32h7b3)
• ci (stable, stm32h735)
• ci (stable, stm32h743)
• ci (stable, stm32h743v)
• ci (stable, stm32h747cm7)
• ci (stable, stm32h753)
• ci (stable, stm32h753v)
• ci (stable, stm32h7b0)
• ci (stable, stm32h7b3)
• Rustfmt