fix: emit events more reliable when starting and stopping io #5097

[Septias]

Send EventType::ConnectivityChanged when using the context methods start_io and stop_io.

close #5097

[link2xt]

Needs a test that 1. Stopping i/o emits an event and connectivity status is actually disconnected/offline when the event arrives. 2. Starting i/o back emits event and connectivity status is at least connecting or so.

This is for displaying online status in account manager, if connectivity status does not actually change or changes after emitting the event, emitting events is not useful.

[Septias]

For me it looks like the test you mentioned in the comment (test_connectivity) actually does cover the cases you pointed out @link2xt

[link2xt]

For me it looks like the test you mentioned in the comment (test_connectivity) actually does cover the cases you pointed out

This test passes even without the fix currently. Maybe need to run consume_events() before calling start_io() and stop_io().

[adbenitez]

I am using this, works like a charm, would be nice to finish the test thing and merge it

[Septias]

Even with consume_events the test runs without a problem.
The logic for sending ConnectivityChanged is spread more or less randomly across the core, most occurrences in ConnectivityStore and all its methods which are executed in SmtpConnectionState and ImapConnectionState. As soon as any connections of these is established or errored out, the event is emitted. Startup should thus be pretty reliable, but teardown not so much. The tests suggest differently though...
I suggest that we merge this without a specific test as it should improve reliability of the event.