Signaling WouldBlock with empty data? about mio-serial HOT 6 CLOSED

The mio docs state that for edge-triggered read operations the client should read until a WouldBlock error. The default match case where b <= 0 should never be triggered because b will always have a size of at least one. However, we still need the guard on the match case or rust will throw an error at compile time. The logic could be more clear if I changed the println! to something like panic!.

For the second part: It does look like you can just catch the error on read, and then check what type of error was raised .

from mio-serial.

Ok, in that case, is the guard on bytes count necessary at all? Also, according to MIO docs, with edge-triggered events we ought loop reading until a WouldBlock is received. Maybe this kind of example (replacing the part I've quoted in the first post) would be more correct? Or is there anything special about mio-serial MIO semantics?

loop {
    // With edge triggered events, we must perform reading until we receive a WouldBlock
    match rx.read(&mut rx_buf) {
        Ok(b) => println!("{:?}", String::from_utf8_lossy(&rx_buf[..b])),
        Err(ref e) if e.kind() == io::ErrorKind::WouldBlock => {
            break;
        }
        Err(e) => {
            println!("Error:  {}", e)
            break;
        },
    }
}

from mio-serial.

Hi, @berkowski ,

if you find time, what is your opinion on my last comment? We're implementing communication with measurement devices that needs to be pretty reliable, so I need to be sure about the correct way to use mio-serial. I've had issues before with TCP communication freeze when I didn't loop until WouldBlock, that's why I've raised the question with the example I think would be a proper way to handle MIO compatbile Evented read stream..

from mio-serial.

@dvtomas

Sorry about the delay. I'm on a ship at the moment and internet access is pretty..... slow....

Your example code seems like it would work (minus the small missing ';' after the second prinln!). I swapped in your method in the example and didn't see any change using the signals I can easily generate here at the moment.

My own knowledge of the internals of mio are limited to what their docs were about a year ago and it looks like they've put in a lot more work there since my last read. I don't think there is much difference between the two methods in practice, but your version certainly reads better and seems more idomatic.

from mio-serial.

Cool, thank you for your investigation. Should I update the docs with my example and make a PR? Maybe even replacing println!("Error..") with panic! would better show the intent..

Enjoy your sail!

from mio-serial.

I'd welcome a PR with those changes :) Otherwise I'll get around to it when I'm back on shore in December.

from mio-serial.