Comment 30 for bug 1277589

Thanks Elias, this is very helpful information. I think this aligns with my thinking that we don't properly handle login-protected networks. But there's likely more: once we have a problem due to the login network, we don't recover as easily when switched to an open network (or the login is provided, thus giving us unencumbered downloads). This is a case I need to cover in the test suite.

Here's a question though: did you test it after a single reboot? Or after, say a 30m wait? I ask because the system-image-dbus process has a lifetime of 10 minutes of inactivity, meaning if you attempt no updates and s-i-dbus sees no other activity for 10m, it will exit. Of course a reboot will also exit the process. So the question is whether subsequent errors are due to some state being retained in the running process, or some extra-process state hanging around on the file system.

Anyway, both are interesting situations to explore.