Implement Write trait for MaybeUninit<u8>

[SUPERCILEX]

This PR makes it possible to write something like the following:

letmut buf = [MaybeUninit::<u8>::uninit();20];let written = {let slice = buf.as_mut_slice();write!(slice,"{}", u64_int).unwrap();
buf.len() - slice.len()};let buf:[u8] = unsafe{ mem::transmute(buf[..written])};

[rust-highfive]

r? @m-ou-se

(rust-highfive has picked a reviewer for you, use r? to override)

[SUPERCILEX]

Fixed the benchmark and added one for MaybeUninit:

test io::impls::tests::bench_write_maybeuninit ... bench: 24 ns/iter (+/- 0)
test io::impls::tests::bench_write_slice ... bench: 25 ns/iter (+/- 6)
test io::impls::tests::bench_write_vec ... bench: 28 ns/iter (+/- 0)

[m-ou-se]

I'm not sure if this is the right kind of API we want to provide for this. The guarantee that the written part is properly initialized and usable as &[u8] is not expressed through this API, and still requires the user to use unsafe code. An abstraction/type for partially initialized buffers is probably a lot nicer to use, and could be entirely safe to use.

[SUPERCILEX]

Would it make sense to implement Write for ReadBuf then?

[JohnCSimon]

@SUPERCILEX
Ping from triage: Can you please post the status of this PR? Maybe ping any commenters for additional feedback?
Thank you.

[SUPERCILEX]

I'm waiting on a reply to my above comment.

[JohnCSimon]

@m-ou-se ☝️ ?

[bors]

☔ The latest upstream changes (presumably #97910) made this pull request unmergeable. Please resolve the merge conflicts.

[JohnCSimon]

I'm waiting on a reply to my above comment.

@m-ou-se - can you comment on this?

[SUPERCILEX]

This has been implemented in #97015. You can now do this:

letmut buf = [MaybeUninit::<u8>::uninit();20];letmut buf:BorrowedBuf = buf.as_mut_slice().into();write!(buf.unfilled(),"{}",42).unwrap();let buf:&[u8] = buf.filled();