umqtt.robust: check_msg() blocks after reconnect

[puuu]

By testing the issue described in #186 I found that check_msg() did not work as expected in case of reconnection. This isssue may be related to #102 ...
I can reproduce it with the ESP8266 and unix port. Broker is mosquitto 1.3.4 with persistence true.

import utime as time
from umqtt.robust import MQTTClient

def sub_cb(topic, msg):
    print((topic, msg))

c = MQTTClient("umqtt_client", SERVER)
c.DEBUG = True
c.set_callback(sub_cb)

c.connect(clean_session=False)
c.subscribe(b"foo_topic")

Now you can call c.check_msg() multiple times and see the non-blocking behaviour and if you publish a message then sub_cb() is called, as expected.

But if you restart the broker and call c.check_msg() again, you get:

>>> c.check_msg()
mqtt: OSError(-1,)

and check_msg() blocks until a new message is published.

This behaviour is obviously, because during reconnect, connect() creates a new socket and do not set non-blocking mode.
IMHO blocking during reconnection is fine, since delay() can be overriden, but check_msg() should not block, without pending message.

[turutupa]

You found a solution? Is it fixed? What is the solution to a robust connection??

[EddieParis]

can you try to pull #385 and see if it fixes your issue ?

[dpgeorge]

Fixed by 4dc2d5e