Memory leaks through missing callback deregistration in Bluez

[JoeHut]

• bleak version: master
• Python version: 3.7.3
• Operating System: Linux
• Bluez version: 5.50

Description

In my use case I have a long running application that is looking for devices with certain properties (like a name prefix, RSSI etc.) and connects to them. To accomplish this I have an infinite loop that calls discover and checks the list of devices until the proper one is available. Something like:

while True:
    devices = await discover(timeout=1)
    for d in devices:
        if d.name.startswith("PREFIX"):
            return d

I observed that if it is running for a while, the discover takes not only around a second but up to multiple minutes.

Explanation

The issue is that the DBus callbacks in discovery() for InterfacesAdded, InterfacesRemoved and PropertiesChanged are never deleted again. Therefore they pile up and the parse_msg callback is called thousands of times after a couple of minutes.

The issue is therefore similar to #83 where the notifications are sent multiple times after a reconnection. I think it makes sense to go through all Bluez callback registrations and make sure they don't leak.

[JoeHut]

@hbldh I will add a PR for this. Please let me know if there are other ways to handle those callback de-registrations. I am not an expert for DBus/Bluez

[hbldh]

This is a very good catch. Very bad thing to miss this. Will integrate it asap.