Clippy errors and says to use Read::read_exact inside implementations of Read::read_exact that don't directly use the output of Read::read

[slightlyoutofphase]

Given the following read_exact implementation:

#[inline(always)]
fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> {
  //Our implementation of `read` always returns `Ok(read_length)`, so we can unwrap safely here.
  self.read(buf).unwrap();
  Ok(())
}

Clippy gives the following error:

error: handle read amount returned or use `Read::read_exact` instead
    --> src/lib.rs:1331:7
     |
1331 |       self.read(buf).unwrap();
     |       ^^^^^^^^^^^^^^^^^^^^^^^
     |
     = note: `#[deny(clippy::unused_io_amount)]` on by default
     = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unused_io_amount

[flip1995]

Hmm. Suggesting to use read_exact inside read_exact is definitely wrong. But I would rewrite this function as:

#[inline(always)]
fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> {
  //Our implementation of `read` always returns `Ok(read_length)`, so we can unwrap safely here.
  let buf_len = buf.len();
  let bytes_read = self.read(buf).unwrap();
  assert_eq!(bytes_read, buf_len); // Or `debug_assert_eq!` or return an `Err`
  Ok(())
}

So I think the lint triggerring here is correct, it shouldn't suggest read_exact though.

[slightlyoutofphase]

Good point! I adjusted it on my end.

On a general note I don't really get why read_exact doesn't just return io::Result<usize> like read does instead of io::Result<()> anyways, but obviously that's kind of a moot point.

[flip1995]

I guess this is because by definition it would always return Ok(buf.len()), panic (not recommended) or return an Err.