MQTT client should wait for PUBLISH before sending PUBREC

[dborovcanin]

Here is the snippet:

package main

import (
	"flag"
	"fmt"
	"os"
	"os/signal"
	"syscall"
	"time"

	mqtt "github.com/eclipse/paho.mqtt.golang"
)

const server = "tcp://localhost:1883"

const username = "36e1f808-ec19-43fd-9068-710c2dc8c597"
const password = "36f72eb5-4f13-4a87-83d4-bd60c2551529"
const clientID = "test"
const topic = "channels/3b04078c-5217-4bed-9334-9fca6a142b1f/messages/request"

func main() {
	off := flag.Bool("off", true, "Exit on message received.")
	flag.Parse()
	opts := mqtt.NewClientOptions().AddBroker(server)
	opts.SetUsername(username)
	opts.SetPassword(password)
	opts.SetClientID(clientID)
	opts.SetAutoReconnect(false)
	opts.SetMaxReconnectInterval(time.Second * 55)
	opts.SetCleanSession(false)
	opts.SetProtocolVersion(4)
	opts.SetStore(mqtt.NewFileStore("./store"))
	mc := mqtt.NewClient(opts)
	token := mc.Connect()
	token.Wait()
	if token.Error() != nil {
		fmt.Printf("UH OH %s", token.Error())
		return
	}
	mc.Subscribe(topic, 2, makeHandler(*off))
	fmt.Printf("subscribed to %s", topic)
	errs := make(chan error, 2)
	go func() {
		c := make(chan os.Signal)
		signal.Notify(c, syscall.SIGINT)
		errs <- fmt.Errorf("%s", <-c)
	}()
	err := <-errs
	fmt.Printf("terminated %s", err)
}

func makeHandler(off bool) func(client mqtt.Client, msg mqtt.Message) {
	return func(client mqtt.Client, msg mqtt.Message) {
		fmt.Println("received msg", string(msg.Payload()))
		if off {
			os.Exit(1)
		}
	}
}

Running it with the off flag set to false (the flag is here just for convenience, you can simply comment os.Exit(1)) means that the client will exit after PUBLISH receive, but before PUBREC. Once the client is restarted, it will continue where it stopped: sending PUBREC. Though this is not against the protocol specification (at least I could not find anything that confronts this behavior), according to the spec, the client should be waiting for the broker to loop back PUBLISH with DUP=1 in this scenario. It looks that something like removing *packets.PublishPacket from here will solve this issue. I'm willing to send a PR, but I didn't want to do it on my own, before opening this discussion.

[alsm]

Thanks for noticing this, while I don't think it should cause any issues it's better to be correct.

[alsm]

Read the linked issues, turns out it does cause issues, even more reason to be correct.