More robust peer banning

[quentinlesceller]

Handle the errors and propagate them instead of just logging.

[antiochp]

👍 yes!

[hashmap]

Here and below - perhaps just return error chain::ErrorKind::Other(format!("ban peer error :{:?}", e)).into() instaead of defining a temp var and returnning it?

[quentinlesceller]

I have no choice otherwise Rust complains (cannot infer type thing).

[quentinlesceller]

Unfortunately cannot be done. Rust need type annotations for that one.

[hashmap]

I'd prefer to return on the previous line, I remember @antiochp like to be more explicit though

[quentinlesceller]

👍

[antiochp]

[totally off-topic]

My concern is mainly around multiple calls in a block of code -

// inconsistent across multiple lines (easy to miss that `three()` returns a Result)
{
    one()?;
    two()?;
    three()
}

// more consistency, less line noise if we introduce additional lines or reorder lines etc.
{
    one()?;
    two()?;
    three()?;
    Ok(())
}

Also we definitely have cases where you need the ? to translate errors appropriately.
Not sure this is an issue here though.

I'm definitely coming around to single line blocks of code returning on that single line as you suggest @hashmap. As long as the errors convert correctly.

// Also an option if necessary
{
    Ok(one()?)
}

[hashmap]

Yeah, but requires more lines, may require to introduce a temp var and last but not least - we unwrap Result on line 3 and then wrap it on line 4, instead directly returning it as is. Same with the last approach. I hope the compiler produces the same code though.

[quentinlesceller]

Will merge once tests pass.