add basic derive macro for Pod, Zeroable and TransparentWrapper for structs

[icewind1991]

Adds derive macros for Zeroable and Pod for structs.

The macro is exported behind the derive feature flag (same way as serde does)

• The Zeroable macro checks that all fields are Zeroable.
• The Pod macro checks that all fields are Pod, there is no padding bytes and that the struct is #[repr(C)] or #[repr(packed)]

This currently only supports deriving for structs, deriving Zeroable would be possible in the future.

Fixes #4

[Shnatsel]

Now that rgb crate requires these trait bounds, it'd be nice to have a safe way to derive them.

@danielhenrymantilla do I recall correctly that you've come up with a way to derive these via macros-by-example only?

[Lokathor]

I'm very interested in the idea of this PR, but I won't have the time to look until later today, and I'm also a little nervous about putting proc-macros near unsafe code since I don't really use them much, so I'm going to ask around for extra eyes to look at this to make sure it's all good.

[icewind1991]

added a macro for deriving TransparentWrapper

it checks that the struct is #[repr(transparent)] and contains the a field with the type specified in the trait bounds.

If the struct only contains a single field it will automatically use the type of that field for the train bound, if there is more than one field (extra zero-sized fields) you'll need to specify the type using #[transparent(T)].

For example:

#[derive(TransparentWrapper)]
#[repr(transparent)]
struct TransparentSingle {
  a: u16,
}

#[derive(TransparentWrapper)]
#[repr(transparent)]
#[transparent(u16)]
struct TransparentWithZeroSized {
  a: u16,
  b: (),
}

note that this macro is more strict that the requirements of the trait, since the trait allows for "nested" transparent wrappers:

#[repr(transparent)]
struct A(u16);

#[repr(transparent)]
struct B(A);

unsafe impl TransparentWrapper<u16> for B{};

the macro currently rejects this since I can't think of a good way to verify the constraints while allowing this

[Lokathor]

it looks like some of the output code is calling std::mem::size_of ? That at least can be core::mem::size_of.

[Shnatsel]

https://crates.io/crates/zerocopy-derive solves largely the same problem, perhaps some techniques from it can be adopted here.

[icewind1991]

it looks like some of the output code is calling std::mem::size_of ? That at least can be core::mem::size_of.

Fixed

[icewind1991]

running

cargo expand --test basic

from the derive folder is probably the easiest way to check what checks are being generated by the macro

[LPGhatguy]

Looks great to me, I really want this! 😅