Writing to characteristic values from outside the event loop

[its2loud]

I'm not sure how to make use of the event loop properly:
I would like to use the GATT SDK inside a flask server. When a http request is sent to the server I would like to write a value to a characteristic (by handle). But how can I access the DeviceManager or even start / stop it once it is running forever?

[swojo]

Run manager.run() in another thread or process, then you can stop it with manager.stop().
If you run it in a thread, it should be easy to access the device class members so you'll be able to use the services/characteristics once they're resolved.

[harbaum]

Did you actually try that? I did. It works but has some odd side-effects like the thread running the manager then receiving exceptions like ctrl-c and not the main thread anymore.

I think a more detailed example would sure be helpful.

[its2loud]

i tried running it in another thread as well:
https://gist.github.com/its2loud/93f65c27c74df8a934067d7c330cef18

unfortunately i wasn't able to stop it then. any idea, what i am doing wrong?
an example would be very useful of course.

[swojo]

Yeah, I tried it and didn't have any problems.
@its2loud, I edited your example a little to show you how I did it. The manager.run() function is the blocking one so that's the only thing I ran in the thread. Try this?

https://gist.github.com/swojo/2b9b45cdd3bc744e28f7a2fbe885eace

[harbaum]

Thanks a lot. But as commented in your example i still see the manager thread receiving the ctrl-c and not the main thread.

[harbaum]

I see this on two different ubuntu x86 machines and on a raspberry pi. On what kind of machin does this work for you?

[its2loud]

As commented below the gist I can't even stop it with Ctrl+C. It takes a few seconds and throws another exception..

[harbaum]

If i replace the call to manager.run() by something trivial like a time.sleep() then the exception handling works as expected. So manager.run() somehow influences all this.

[harbaum]

The closest thing i could come up with is this:
https://gist.github.com/harbaum/db0a0928458fefc0793df66a5d377b16

But it e.g. requires the main thread to actively check for the manager thread to be still alive.

Here someone describes exactly the same problem:
https://stackoverflow.com/questions/36680667/terminate-gobject-mainloop-threads-together-with-main/36778891
Interestingly he accepts the answer although it doesn't help in this case as the signal handler of the main thread never receives the signal.

[harbaum]

This one also describes the same problem:
https://stackoverflow.com/questions/28465611/python-dbusgmainloop-inside-thread-and-try-and-catch-block/28476630
The solution presented there also doesn't really solve the problem but rather works around it.

[Snevzor]

@harbaum, can you confirm #28 fixes your issue?