Inform ourselves of presence info.

[MF1-MS]

Treat resources belonging to the logged in user as being in the roster, so clients are informed about presence changes on the same account.

This MR adds the above function and two UTs to check the existing behaviour of the Roster.contains(), and verify the new function.

[Flowdalic]

I am skeptical about this one, as this makes the roster behave differently depending on if the user's JID is known or not.

I can see that such a functionality would be desirable though. I wonder if we shouldn't take a different approach on the design though. I would probably experiment with an design where there are dedicated callbacks for presence changes from other devices of the user, that can be requested from the Roster. Then, let Roster hook into the connection authenticated callback (if not already), and register the stanzas listeners with the now known JID of the user.

I am not yet decided if those "own presence changed" callbacks are also invoked for presence changes of the local connection. From an UI perspective it may be beneficial? That said, I am not sure if the protocol does reflect the presence back to self. Need to look that up in the spec.

[MF1-MS]

In what circumstances do we not know the logged in user's JID?

I can't find anything specifically on the matter in the base XMPP RFCs or XEP-0162. But that means I can't fid anything that argues this would be spec-uncompliant either.

[Flowdalic]

In what circumstances do we not know the logged in user's JID?

That depends on your definition of "logged in". In XMPP you do not know your JID after you authenticated (cf. RFC 6120 § 9.1.2). Only after you bound a resource the client becomes aware of its own JID (cf. RFC 6120 § 9.1.3). Note that Smack's ConnectionListener.authetnicated(), despite its name, runs after the resource has been bound. So we can use that callback, as I wrote in my previous comment.

[MF1-MS]

So to rephrase what you suggested for my understanding, create a ConnectionListener for Roster then when triggered for authenticated() constructs a StanzaListener containing the now known Jid and adds it to the connection. But wouldn't having a separate StanzaListener like this lead to a lot of code duplication with the existing PresencePacketListener?