Create FileWatcher when ConfigManagingActor runs

[daniel-zullo-frequenz]

The ConfigManagingActor was unable to process events after being restarted because the FileWatcher cannot be reused when the ConfigManagingActor is restarted, so we need to create a new instance every time the actor starts running.

[llucax]

Other than that, LGTM.

[llucax]

I think you should add a del file_watcher when the run finishes (including if an exception is raised), otherwise the file watcher internal task will keep running. Sadly it doesn't provide a stop() yet, I think we should add one, as using del is weird.

[daniel-zullo-frequenz]

I think it's no needed. The reference to the local file_watcher variable is gone once exiting the _run() method in the actor which will cause the file watcher internal task to be stopped when the object is collected. See here for more details

[llucax]

Finalization is complicated. It is very likely that __del__ is called, but not guaranteed (if there is a cycle for example). This is why ideally we should probably have a stop() method and be more explicit about it.

I'm fine with leaving this as is, although I think it is always safer to do things explicitly to avoid surprises.

[daniel-zullo-frequenz]

I agree it's safer to explicitly delete the file_watcher then, will do it. I presume I'll need capture the error, delete file_watcher and re-raise the error

[daniel-zullo-frequenz]

Updated to explicitly delete file_watcher when exiting _run(). I needed to capture BaseException to be able to delete file_watcher when the actor is stopped (by calling stop()).

[llucax]

You are manually writing a finally 😆

Suggested change

	except (Exception, BaseException):
	del file_watcher
	raise
	finally:
	del file_watcher

You could also use a ExitStack. Or we could make the FileWatcher a context manager (look ma, here it comes BackgroundService/Service again, if FileWatcher were a Service it would automatically be a context manager 😉), but I would leave that for the future as it requires a lot of work (changing FileWatcher, releasing etc.) and it should come automatically when we migrate channels to use core (and Service).

[daniel-zullo-frequenz]

I really forgot finally exists in python, somehow I only thought of try/except/else...sorry 👼

You could also use a ExitStack. Or we could make the FileWatcher a context manager (look ma, here it comes BackgroundService/Service again, if FileWatcher were a Service it would automatically be a context manager 😉), but I would leave that for the future as it requires a lot of work (changing FileWatcher, releasing etc.) and it should come automatically when we migrate channels to use core (and Service).

Agreed