[Bug 69585] New: MC regression in 5.15.1: account-manager/connectivity.py fails

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=69585

          Priority: medium
            Bug ID: 69585
                CC: rishi.is at lostca.se
          Assignee: simon.mcvittie at collabora.co.uk
           Summary: MC regression in 5.15.1:
                    account-manager/connectivity.py fails
        QA Contact: telepathy-bugs at lists.freedesktop.org
          Severity: major
    Classification: Unclassified
                OS: All
          Reporter: simon.mcvittie at collabora.co.uk
          Hardware: Other
            Status: ASSIGNED
           Version: git master
         Component: mission-control
           Product: Telepathy

Here's a fun regression in the new GNetworkMonitor-based connectivity
monitoring.

1. We're connected to the Internet. MC has a live Connection.
   The connectivity state is (awake, up, stable, running), all of which
   are needed.

2. We go to an indeterminate state (DISCONNECTING).

(process:4081): mcd-DEBUG: connectivity_monitor_nm_state_change_cb: New
NetworkManager network state 30 (unstable state)
(process:4081): mcd-DEBUG: connectivity_monitor_change_states: awake: 1 -> 1;
up: 2 -> 2; stable: 4 -> 0; running: 8 -> 8

3. We finish disconnecting. MC tells us that it has disconnected before the
GNetworkMonitor does.

(process:4081): mcd-DEBUG: connectivity_monitor_nm_state_change_cb: New
NetworkManager network state 20 (stable state)
(process:4081): mcd-DEBUG: connectivity_monitor_change_states: awake: 1 -> 1;
up: 2 -> 2; stable: 0 -> 4; running: 8 -> 8

... spot the error: we're back in a stable state, we haven't yet been told by
the GNetworkMonitor that we've been disconnected (which would be indicated by
"up: 0"), so we try to reconnect!

A moment later we're told by the GNetworkMonitor that the connection has gone
down

(process:4081): mcd-DEBUG: connectivity_monitor_network_changed:
GNetworkMonitor (FakeNetworkMonitor) says we are offline
(process:4081): mcd-DEBUG: connectivity_monitor_change_states: awake: 1 -> 1;
up: 2 -> 0; stable: 4 -> 4; running: 8 -> 8

and try to disconnect the connection again.

Open question: can this happen in real life, or is it just a deficiency in our
simulated connectivity? Will NM ever signal DISCONNECTED before netlink/etc.
says this has happened? (I suspect the answer is "it depends who reads which
sockets in which order")

Once the system has "settled", if NM says we are disconnected, but the
GNetworkMonitor says we're OK, we do probably still want to give it a try, so
ideally, we shouldn't treat DISCONNECTED as "definitely offline" - but I'm
willing to compromise on that if necessary.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.