automatically truncate larger integer types passed to bitfields

[Lotterleben]

closes #102

[japaric]

Looks good! Left some comments

[japaric]

🎉

[japaric]

   --> /tmp/mdbook-FMDBPy/bitfields.md:25:1
    |
4   | binfmt::trace!("first two bits: {0:0..3}", 254);
    | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `binfmt::export::sealed::Truncate<u8>` is not implemented for `i32`

this book test is failing due to type inference (as mentioned in the issue description). you'll need to specify the type of the literal, e.g. 254u8

[japaric]

could we move this one further down? so that all Truncate<u16> impls are next to each other

[Lotterleben]

sure, done

[Lotterleben]

   --> /tmp/mdbook-FMDBPy/bitfields.md:25:1
    |
4   | binfmt::trace!("first two bits: {0:0..3}", 254);
    | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `binfmt::export::sealed::Truncate<u8>` is not implemented for `i32`

this book test is failing due to type inference (as mentioned in the issue description). you'll need to specify the type of the literal, e.g. 254u8

So are we happy with always having to specify the type for now or should this be fixed while we're at it? if the former, I'll add some doc pointing this out

[japaric]

So are we happy with always having to specify the type for now or should this be fixed while we're at it?

In practice most of the time you'll be passing something fully typed, e.g. my_register.read() which returns a u32 for example.
The type inference error happens when you use (untyped) literals as the argument and that scenario mainly occurs in tests and sample programs.
Would be good to document that a fully typed value is expected: the type must be one of u{8,16,32,64}

[japaric]

should this be fixed while we're at it?

addendum: we can "fix" this particular test by implementing Truncate<u8> for i32 but I think allowing signed integers makes the feature more confusing: what happens with the sign bit? is the bit 31 of i32 the sign bit? should bit 15 return the sign bit if the arg is i32 but then casted to u16? This feature is meant for viewing registers and those are not signed.

[jonas-schievink]

Yeah I don't think this need to be fixed if it's documented right.

[Lotterleben]

done!