net: add failed addresses + details on connect errors, make connect more robust in the default non blocking mode (#15364)

examples/net_failconnect.v

@@ -0,0 +1,5 @@
+import net
+
+conn := net.dial_tcp('[::1]:57000')?
+peer_addr := conn.peer_addr()?
+println('$peer_addr')

vlib/net/errors.v

@@ -12,12 +12,13 @@ pub const (
 		errors_base + 2)
 	err_option_wrong_type   = error_with_code('net: set_option_xxx option wrong type',
 		errors_base + 3)
-	err_port_out_of_range   = error_with_code('', errors_base + 5)
+	err_port_out_of_range   = error_with_code('net: port out of range', errors_base + 5)
-	err_no_udp_remote       = error_with_code('', errors_base + 6)
+	err_no_udp_remote       = error_with_code('net: no udp remote', errors_base + 6)
 	err_connect_failed      = error_with_code('net: connect failed', errors_base + 7)
 	err_connect_timed_out   = error_with_code('net: connect timed out', errors_base + 8)
 	err_timed_out           = error_with_code('net: op timed out', errors_base + 9)
 	err_timed_out_code      = errors_base + 9
+	err_connection_refused  = error_with_code('net: connection refused', errors_base + 10)
 )
 
 pub fn socket_error_message(potential_code int, s string) ?int {
@@ -43,13 +44,13 @@ pub fn socket_error(potential_code int) ?int {
 }
 
 pub fn wrap_error(error_code int) ? {
+	if error_code == 0 {
+		return
+	}
 	$if windows {
 		enum_error := wsa_error(error_code)
 		return error_with_code('net: socket error: $enum_error', error_code)
 	} $else {
-		if error_code == 0 {
-			return
-		}
 		return error_with_code('net: socket error: $error_code', error_code)
 	}
 }

vlib/net/net_nix.c.v

@@ -10,6 +10,8 @@ module net
 
 #flag solaris -lsocket
 
+const is_windows = false
+
 fn error_code() int {
 	return C.errno
 }
@@ -23,4 +25,5 @@ pub const (
 
 const (
 	error_ewouldblock = C.EWOULDBLOCK
+	error_einprogress = C.EINPROGRESS
 )

vlib/net/net_windows.c.v

@@ -1,5 +1,7 @@
 module net
 
+const is_windows = true
+
 // WsaError is all of the socket errors that WSA provides from WSAGetLastError
 pub enum WsaError {
 	//
@@ -739,6 +741,7 @@ pub fn wsa_error(code int) WsaError {
 
 const (
 	error_ewouldblock = WsaError.wsaewouldblock
+	error_einprogress = WsaError.wsaeinprogress
 )
 
 // Link to Winsock library

vlib/net/tcp.v

@@ -1,6 +1,7 @@
 module net
 
 import time
+import strings
 
 const (
 	tcp_default_read_timeout  = 30 * time.second
@@ -24,12 +25,16 @@ pub fn dial_tcp(address string) ?&TcpConn {
 		return error('$err.msg(); could not resolve address $address in dial_tcp')
 	}
 
+	// Keep track of dialing errors that take place
+	mut errs := []IError{}
+
 	// Very simple dialer
 	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 {
+			errs << err
 			// Connection failed
 			s.close() or { continue }
 			continue
@@ -41,8 +46,20 @@ pub fn dial_tcp(address string) ?&TcpConn {
 			write_timeout: net.tcp_default_write_timeout
 		}
 	}
+
+	// Once we've failed now try and explain why we failed to connect
+	// to any of these addresses
+	mut err_builder := strings.new_builder(1024)
+	err_builder.write_string('dial_tcp failed for address $address\n')
+	err_builder.write_string('tried addrs:\n')
+	for i := 0; i < errs.len; i++ {
+		addr := addrs[i]
+		why := errs[i]
+		err_builder.write_string('\t$addr: $why\n')
+	}
+
 	// failed
-	return error('dial_tcp failed for address $address')
+	return error(err_builder.str())
 }
 
 // bind local address and dail.
@@ -431,7 +448,11 @@ fn (mut s TcpSocket) connect(a Addr) ? {
 		if res == 0 {
 			return
 		}
-
+		ecode := error_code()
+		// On nix non-blocking sockets we expect einprogress
+		// On windows we expect res == -1 && error_code() == ewouldblock
+		if (is_windows && ecode == int(error_ewouldblock))
+			|| (!is_windows && res == -1 && ecode == int(error_einprogress)) {
 			// The  socket  is  nonblocking and the connection cannot be completed
 			// immediately.  (UNIX domain sockets failed with EAGAIN instead.)
 			// It is possible to select(2) or poll(2) for completion by selecting
@@ -441,24 +462,25 @@ fn (mut s TcpSocket) connect(a Addr) ? {
 			// unsuccessfully (SO_ERROR is one of the usual error codes  listed  here,
 			// ex‐ plaining the reason for the failure).
 			write_result := s.@select(.write, net.connect_timeout)?
-		if write_result {
 			err := 0
 			len := sizeof(err)
-			socket_error(C.getsockopt(s.handle, C.SOL_SOCKET, C.SO_ERROR, &err, &len))?
+			xyz := C.getsockopt(s.handle, C.SOL_SOCKET, C.SO_ERROR, &err, &len)
-
+			if xyz == 0 && err == 0 {
-			if err != 0 {
-				return wrap_error(err)
-			}
-			// Succeeded
 				return
 			}
-
+			if write_result {
-		// Get the error
+				if xyz == 0 {
-		socket_error(C.connect(s.handle, voidptr(&a), a.len()))?
+					wrap_error(err)?
-
+					return
-		// otherwise we timed out
+				}
-		return err_connect_timed_out
+				return
+			}
+			return err_timed_out
+		}
+		wrap_error(ecode)?
+		return
 	} $else {
-		socket_error(C.connect(s.handle, voidptr(&a), a.len()))?
+		x := C.connect(s.handle, voidptr(&a), a.len())
+		socket_error(x)?
 	}
 }

vlib/net/tcp_self_dial_from_many_clients_test.v

@@ -0,0 +1,114 @@
+module main
+
+import net
+import time
+
+const xport = 15523
+
+struct Context {
+mut:
+	ok_client_dials   int
+	fail_client_dials int
+	//
+	ok_client_close   int
+	fail_client_close int
+	////
+	ok_server_accepts   int
+	fail_server_accepts int
+	//
+	ok_server_close   int
+	fail_server_close int
+	//
+	received []int
+}
+
+fn elog(msg string) {
+	eprintln('$time.now().format_ss_micro() | $msg')
+}
+
+fn receive_data(mut con net.TcpConn, mut ctx Context) {
+	mut buf := []u8{len: 5}
+	for {
+		bytes := con.read(mut buf) or { -1 }
+		if bytes < 0 {
+			break
+		}
+		ctx.received << buf[0]
+	}
+	con.close() or {
+		ctx.fail_server_close++
+		return
+	}
+	ctx.ok_server_close++
+}
+
+fn start_server(schannel chan int, mut ctx Context) {
+	elog('server: start_server')
+	mut tcp_listener := net.listen_tcp(net.AddrFamily.ip, ':$xport') or {
+		elog('server: start server error $err')
+		return
+	}
+	elog('server: server started listening at port :$xport')
+	schannel <- 0
+
+	for {
+		mut tcp_con := tcp_listener.accept() or {
+			elog('server: accept error: $err')
+			ctx.fail_server_accepts++
+			continue
+		}
+		go receive_data(mut tcp_con, mut ctx)
+		ctx.ok_server_accepts++
+		elog('server: new tcp connection con.sock.handle: $tcp_con.sock.handle')
+		continue
+	}
+}
+
+fn start_client(i int, mut ctx Context) {
+	elog('client [$i]: start')
+	mut tcp_con := net.dial_tcp('127.0.0.1:$xport') or {
+		elog('client [$i]: net.dial_tcp err $err')
+		ctx.fail_client_dials++
+		return
+	}
+	ctx.ok_client_dials++
+	elog('client [$i]: conn is connected, con.sock.handle: $tcp_con.sock.handle')
+	tcp_con.write([u8(i)]) or { elog('client [$i]: write failed, err: $err') }
+	time.sleep(1 * time.second)
+	elog('client [$i]: closing connection...')
+	tcp_con.close() or {
+		elog('client [$i]: close failed, err: $err')
+		ctx.fail_client_close++
+		return
+	}
+	ctx.ok_client_close++
+}
+
+fn test_tcp_self_dialing() {
+	elog('>>> start')
+	start_time := time.now()
+	mut ctx := &Context{}
+	mut server_channel := chan int{cap: 1}
+	go start_server(server_channel, mut ctx)
+	svalue := <-server_channel
+	elog('>>> server was started: ${svalue}. Starting clients:')
+	for i := int(0); i < 20; i++ {
+		go start_client(i, mut ctx)
+		elog('>>> started client $i')
+		// time.sleep(2 * time.millisecond)
+	}
+	max_dt := 5 * time.second
+	for {
+		t := time.now()
+		dt := t - start_time
+		if dt > max_dt {
+			elog('>>> exiting after $dt.milliseconds() ms ...')
+			dump(ctx)
+			assert ctx.fail_client_dials < 2, 'allowed failed client dials, from $ctx.ok_server_accepts connections'
+			assert ctx.received.len > ctx.ok_server_accepts / 2, 'at least half the clients sent some data, that was later received by the server'
+			elog('>>> goodbye')
+			exit(0)
+		}
+		time.sleep(10 * time.millisecond)
+	}
+}

vlib/net/websocket/websocket_test.v

@@ -26,9 +26,13 @@ fn test_ws_ipv6() {
 		return
 	}
 	port := 30000 + rand.intn(1024) or { 0 }
+	eprintln('> port ipv6: $port')
 	go start_server(.ip6, port)
-	time.sleep(500 * time.millisecond)
+	time.sleep(1500 * time.millisecond)
-	ws_test(.ip6, 'ws://localhost:$port') or { assert false }
+	ws_test(.ip6, 'ws://localhost:$port') or {
+		eprintln('> error while connecting .ip6, err: $err')
+		assert false
+	}
 }
 
 // tests with internal ws servers
@@ -37,9 +41,13 @@ fn test_ws_ipv4() {
 		return
 	}
 	port := 30000 + rand.intn(1024) or { 0 }
+	eprintln('> port ipv4: $port')
 	go start_server(.ip, port)
-	time.sleep(500 * time.millisecond)
+	time.sleep(1500 * time.millisecond)
-	ws_test(.ip, 'ws://localhost:$port') or { assert false }
+	ws_test(.ip, 'ws://localhost:$port') or {
+		eprintln('> error while connecting .ip, err: $err')
+		assert false
+	}
 }
 
 fn start_server(family net.AddrFamily, listen_port int) ? {
@@ -58,15 +66,15 @@ fn start_server(family net.AddrFamily, listen_port int) ? {
 	})?
 	s.on_message(fn (mut ws websocket.Client, msg &websocket.Message) ? {
 		match msg.opcode {
-			.pong { ws.write_string('pong') or { panic(err) } }
+			.pong { ws.write_string('pong')? }
-			else { ws.write(msg.payload, msg.opcode) or { panic(err) } }
+			else { ws.write(msg.payload, msg.opcode)? }
 		}
 	})
 
 	s.on_close(fn (mut ws websocket.Client, code int, reason string) ? {
 		// not used
 	})
-	s.listen() or { panic('websocket server could not listen') }
+	s.listen() or { panic('websocket server could not listen, err: $err') }
 }
 
 // ws_test tests connect to the websocket server from websocket client
@@ -104,11 +112,11 @@ fn ws_test(family net.AddrFamily, uri string) ? {
 			println('Binary message: $msg')
 		}
 	}, test_results)
-	ws.connect() or { panic('fail to connect') }
+	ws.connect() or { panic('fail to connect, err: $err') }
 	go ws.listen()
 	text := ['a'].repeat(2)
 	for msg in text {
-		ws.write(msg.bytes(), .text_frame) or { panic('fail to write to websocket') }
+		ws.write(msg.bytes(), .text_frame) or { panic('fail to write to websocket, err: $err') }
 		// sleep to give time to recieve response before send a new one
 		time.sleep(100 * time.millisecond)
 	}

vlib/vweb/tests/vweb_test.v

@@ -269,6 +269,7 @@ fn simple_tcp_client(config SimpleTcpClientConfig) ?string {
 	mut tries := 0
 	for tries < config.retries {
 		tries++
+		eprintln('> client retries: $tries')
 		client = net.dial_tcp(localserver) or {
 			if tries > config.retries {
 				return err
@@ -278,6 +279,10 @@ fn simple_tcp_client(config SimpleTcpClientConfig) ?string {
 		}
 		break
 	}
+	if client == unsafe { nil } {
+		eprintln('coult not create a tcp client connection to $localserver after $config.retries retries')
+		exit(1)
+	}
 	client.set_read_timeout(tcp_r_timeout)
 	client.set_write_timeout(tcp_w_timeout)
 	defer {

vlib/vweb/tests/vweb_test_server.v

@@ -21,7 +21,7 @@ struct Config {
 
 fn exit_after_timeout(timeout_in_ms int) {
 	time.sleep(timeout_in_ms * time.millisecond)
-	// eprintln('webserver is exiting ...')
+	println('>> webserver: pid: $os.getpid(), exiting ...')
 	exit(0)
 }
 
@@ -43,7 +43,7 @@ fn main() {
 		timeout: timeout
 		global_config: config
 	}
-	eprintln('>> webserver: started on http://localhost:$app.port/ , with maximum runtime of $app.timeout milliseconds.')
+	eprintln('>> webserver: pid: $os.getpid(), started on http://localhost:$app.port/ , with maximum runtime of $app.timeout milliseconds.')
 	vweb.run_at(app, host: 'localhost', port: http_port, family: .ip)?
 }