bug about socket

[despiegk]

has been reported before, made new test script

module main

import net

fn do()?{
	for i in 0 .. 10000{
		println(i)
		mut socket := net.dial_tcp("localhost:6379")?
	}
}



fn main() {
	do() or { panic(err) }
}

on osx m.1

0
1
2
3
V panic: dial_tcp failed for address localhost:6379
v hash: 5498a6c
0   redis_connect_error                 0x0000000104c53b40 main__main + 100
1   redis_connect_error                 0x0000000104c54d2c main + 84
2   dyld                                0x0000000104df50f4 start + 520

always at the exact same iteration

has nothing to do with time, because added some sleep time (1sec even)

[despiegk]

is to a redis server

[despiegk]

when 127.0.0.1 went further

i've seen in other code when localhost, was different then with 127.0.0.1

[spytheman]

Try ulimit -n 30000, then in the same shell: v -d net_blocking_sockets run redis_connect_error.v

[emily33901]

on osx m.1

I don't have an m1 mac so i cant test this necessarily. From the error message it

1. successfully resolves the address to something(s).

	addrs := resolve_addrs_fuzzy(address, .tcp) or {
		return error('$err.msg(); could not resolve address $address in dial_tcp')
	}

2. Goes through those addresses, creating sockets, trying to connect and failing to connect (for whatever reason, we could do with a better way of passing the why down to the end user).

	for addr in addrs {
		mut s := new_tcp_socket(addr.family()) or {
			return error('$err.msg(); could not create new tcp socket in dial_tcp')
		}
		s.connect(addr) or {
			// Connection failed
			s.close() or { continue }
			continue
		}

4. Returns an error because none of the sockets it created pointing to those addresses successfully connected.

	}
	// failed
	return error('dial_tcp failed for address $address')

I think i mentioned to @medvednikov one time that it could be alignment related (given m1 is arm and not x64), however I have never been able to test this. All I know is that there is no issue with this on windows, or linux on x64

[emily33901]

If you add a socket.close() at the end then this goes away.

module main

import net

fn do()?{
        for i in 0 .. 10000{
                println(i)
                mut socket := net.dial_tcp("localhost:6379")?
                println("Connected successfully")
                socket.close()?
        }
}

fn main() {
        do() or { panic(err) }
}

Could I recommend closing your sockets after you have finished using them (atleast if they are still open)? My other thoughts are that it might be trying to re-use a local port for the same socket that is still open (because you havent closed it. No idea how to check that though...)