Implement UART for sercom

[zklapow]

This adds structs and corresponding serial::{Read, Write} impls for a UART on each sercom.

This works as is but needs some improvement to fully support all the features available on the samd21. Currently this is:

• Only asynchronous
• Only Arithmetic baud mode is supported
• Only 8N1
• Both TX/RX are always on

@wez

[wez]

Many thanks for this! I'll have some time to properly review on saturday.
Hmm, looks like the travis CI bits might need updating for more recent changes in rust.

[zklapow]

👌 thanks @wez I'll probably update this a bit this weekend as well so no rush, at the very least I want to add a UART example but I need to clean up some of my code.

[zklapow]

I also pinned the nightly version to 2018-05-16 (which was the version I was developing on) to fix some the upstream build issues. I was able to fix the panic_implementation errors by upgrading panic-abort but then got a bunch of proc-macro issues which seem to need upstream changes to get merged in cortex-m-rtfm so I just went with rolling back nightly.

[zklapow]

@wez any chance you've had some time to take a look at this?

[wez]

@zklapow sorry for not getting back to you already; I'll look at this tomorrow!

[djmcgill]

I've been using this branch, seems to be okay on a samd21g18a. Is there a reason that core::fmt::Write isn't implemented for the blocking variant*? Without it I was having a pretty tough time printing numbers and hex values and so on.

The version which I'm using is (the struct is only there to get around the orphan issue):

struct BlockingWrite<W>(W) where W: hal::blocking::serial::Write<u8>;
impl<W: hal::blocking::serial::Write<u8>> core::fmt::Write for BlockingWrite<W> {
    fn write_str(&mut self, s: &str) -> core::fmt::Result {
        match self.0.bwrite_all(s.as_bytes()) {
            Ok(()) => Ok(()),
            Err(_) => Err(core::fmt::Error),
        }
    }
}

*: since core::fmt::Result can't contain WouldBlock I don't think it's possible to do a non-blocking version of this.

[zklapow]

@djmcgill glad its working for you! I actually had a commit with implemented the core::fmt side of things that I just forgot to push, should be there now. I think I managed to avoid the intermediate struct by moving some imports around.

[wez]

Sorry for the long lag in reviewing; this looks great and I've pushed it to master! Many thanks!