amqp.Channel's method InspectQueue blocks when connection is lost

[PeterEFinch]

A colleague and I noticed that amqp.Channel's method InspectQueue is sometimes blocking when the connection is lost. I have included a minimal example of the code used below.

I tested this code under two different circumstances:

1. AMQP_URL had the address of a local RabbitMQ server running in docker (rabbitmq:3-management). Once the connection was established I stopped the docker container. The program stopped as expected (see expected_output.txt).
2. AMQP_URL had the address of our production RabbitMQ server. Once the connection was established I disconnected the internet. The program never stopped and appeared to be stuck at the method channel.QueueInspect (see
   blocking_output.txt).

Additional information: I ran this code inside GoLand with go version go1.16.2 darwin/amd64 . I removed the preamble but left in the exit message .

Hopefully, this problem should be reproducible.

package main

import (
	"fmt"
	"os"
	"time"

	"github.com/streadway/amqp"
)

func main() {
	const queue = "bug_testing_queue"

	url, ok := os.LookupEnv("AMQP_URL")
	if !ok {
		panic("no environment variable AMQP_URL")
	}

	connection, err := amqp.Dial(url)
	if err != nil {
		panic(err)
	}

	channel, err := connection.Channel()
	if err != nil {
		panic(err)
	}

	_, err = channel.QueueDeclare(queue, true, false, false, false, nil)
	if err != nil {
		panic(err)
	}

	ticker := time.NewTicker(time.Second)
	defer ticker.Stop()

	for {
		fmt.Println("waiting")
		select {
		case <-ticker.C:
			fmt.Println("pre queue inspect")
			_, err := channel.QueueInspect(queue)
			fmt.Println("post queue inspect")
			if err != nil {
				fmt.Printf("unable to inspect queue: %v\n", err)
				return
			}
		}
	}
}

---

Finally, adding

notifyCh := channel.NotifyClose(make(chan *amqp.Error))

before the for loop and

case <-notifyCh:
fmt.Println("channel closed unexpectedly")
return

to the select statement doesn't prevent the blocking output.

[PeterEFinch]

@michaelklishin I noticed you have said that Team RabbitMQ plans on forking this library (#497 (comment)) and that @streadway lacks time (#497 (comment)). Is there anyway we can move such issues to this planned library?