interfaces: bluez: fix dbus rules

[ssweeny]

This fixes a few issues with the bluez interface, specifically making the dbus policy permanent, fixing the label to match what we see in snappy, and allowing obexd to claim its dbus name.

[zyga]

Looks good

[ssweeny]

Looks like the autopkgtest is hitting some issues downloading snaps. I don't think this is something my branch could have caused, but you're the expert @zyga 😄

[morphis]

Tested this with the snap from https://code.launchpad.net/~ssweeny/+snap/bluez/+build/601 in KVM instance having a BT dongle connect. Works fine, see https://paste.ubuntu.com/15931839/ for details.

[niemeyer]

This is hard-coding an interface to be used by a specifically named snap, which doesn't seem to make sense for an interface in a platform.

@zyga can you please work with him to ensure we have the proper content in this template? I thought we had already discussed this last week?

[morphis]

From what I got the agreement was that we go with this until we have a proper mechanism in place (which @zyga and @jdstrand are working on) to replace the peer label with the correct label from the connected plug.

[ssweeny]

This was my understanding as well. With this fix we can have a working bluez snap in the store today (as opposed to the broken one that's in there now) while we wait on the proper fix.

[jdstrand]

I agree with Gustavo but in the interest of time if people want to fix bluez (and network-manager) quickly, I would be ok with this but please add a comment:

FIXME: adjust to use 'snap..' from the connecting slot

[niemeyer]

That doesn't seem to reflect the conversation we had last week about this problem.

[zyga]

I'm working on improving interface details now

[morphis]

@jdstrand @zyga @niemeyer So what is the decision? Do you want automatic peer label be fixed before we merge this? I would like to get this merged now and fix those bits afterwards so that we get a working snap back in the store.

[ssweeny]

@morphis @zyga sent me a patch to stop hardcoding the snap name. I think he's waiting on some input from @jdstrand before giving me the go-ahead to pull it in.

[morphis]

@ssweeny Thanks. Once we have this here I will do the same for the networkmanager interface

[mvo5]

retest this please

[jdstrand]

I think we are deadlocked. I gave @zyga a patch that starts the non-hardcoding, but it required snapd changes to implement correctly which @zyga is working on. But then people said they are waiting on me-- can someone say specifically what this is still waiting on?

[jdstrand]

I'm told that the PR in question was zyga/snapd@16228ad which I didn't see in my email. I commented on that just now.

[zyga]

I'm closing this pull request. I've improved on this branch in #1078