Support bincode 2

[MavethGH]

This PR adds a new feature (bincode2) that when enabled, derives Encode and Decode for Uuid. See this section of the migration guide for bincode 2.

[KodrAus]

Hi @MavethGH 👋 I’m not familiar with bincode 2 yet. If bincode 2.x is expected to produce compatible bytes as 1.x then we should make sure we serialize to the same binary when using serde + bincode 1.x, and when using bincode 2.x.

[MavethGH]

If bincode 2.x is expected to produce compatible bytes as 1.x then we should make sure we serialize to the same binary when using serde + bincode 1.x, and when using bincode 2.x.

The bytes will not be the same, since serde includes the length of the array, while bincode 2.x doesn't since it's a fixed-size array. However, bincode 2.x can still deserialize a Uuid serialized that way by using Compat and config::legacy().

[KodrAus]

It looks like bincode 2.x does include the length of the array if we serialize it as a &[u8] instead of as [u8; 16], which is how we do it with serde and bincode 1.x just because that's the API serde gives us.

cc @ZoeyR @VictorKoenders who look like they might be maintaining bincode now, is there a general expectation that if you already support bincode 1.x that you'll maintain an equivalent encoding in bincode 2.x? Or should users expect there to be differences?

[ZoeyR]

If at all possible the encodings should be equivalent.

[KodrAus]

Thanks @ZoeyR. In that case, I think we should implement Encode and Decode manually, and add a test to ensure equivalence with serde + bincode 1.x. We don't necessarily need to depend on serde and bincode 1.x to do it, we can just use a pre-serialized buffer.

[francisbr]

I worked a a simple manual implementation of Encode & Decode
#821

[MavethGH]

@francisbr Thanks!

[MavethGH]

Going to close this now in that case.