Should IoSliceMut::advance() be used when impl std::io::Read::read_vectored()?

[zh-jq-b]

As of io_slice_advance is stable since 1.81, the following asserting may won't work if the users use IoSliceMut::advance in their impl of Read trait.

fn what_to_read<R>(reader: R)
where
    R: Read,
{
    let mut b1 = [0u8; 16];
    let mut b2 = [0u8; 32];

    let mut iov = [IoSliceMut::new(&mut b1), IoSliceMut::new(&mut b2)];
    let nr = reader.read_vectored(&mut iov).unwrap();

    if nr > 0 {
        assert_eq!(b1[0], iov[0][0]);
    }
    if nr >= b1.len() {
        assert_eq!(iov[0].len(), b1.len());
    }
}

It is unclear that whether the returned size of read_vectored is the real read data size, or the valid buffer size in the passed &mut [IoSliceMut<'_>].
Please clarify that so Read implementations can do the correct thing.

[zh-jq-b]

Use IoSliceMut::advance() inside read_vectored() can avoid copy when do HTTP chunks decode reading, or something encoded like that. But this will break code that always use the original buffer as the read address of the received data.