Comment 13 for bug 1570901

>takes reeeally long to get to the live system [...] and the load is at 6.

Ignore this part, that was just because I forgot "-enable-kvm".

> Also, keyboard and mouse are broken in the live session

I'm still trying to figure out this. Annoyingly it works if I use the serial console (QEMU option "-serial stdio" and kernel option "console=ttyS0"). In one boot I had X.org use VT1, which would collide with getty on VT1 that we always start (as in Debian/Ubuntu we usually want the first X.org on VT7).

Interestingly, when I do that ubiquity.service actually does start, but fails to connect to X:

● ubiquity.service - Ubuntu live CD installer
   Loaded: loaded (/lib/systemd/system/ubiquity.service; enabled; vendor preset: enabled)
   Active: inactive (dead) since Mon 2016-04-18 10:33:09 UTC; 4min 23s ago
  Process: 1238 ExecStart=/usr/share/ubiquity/start-ubiquity-dm (code=exited, status=0/SUCCESS)
 Main PID: 1238 (code=exited, status=0/SUCCESS)

Apr 18 10:32:57 ubuntu systemd[1]: Starting Ubuntu live CD installer...
Apr 18 10:32:59 ubuntu python3[1242]: pam_unix(su:session): session opened for user ubuntu by (uid=0)
Apr 18 10:33:09 ubuntu start-ubiquity-dm[1238]: Gtk-Message: Failed to load module "overlay-scrollbar"
Apr 18 10:33:09 ubuntu start-ubiquity-dm[1238]: Failed to connect to Mir: Failed to connect to server socket: No such file or directory
Apr 18 10:33:09 ubuntu start-ubiquity-dm[1238]: Unable to init server: Could not connect: Connection refused
Apr 18 10:33:09 ubuntu start-ubiquity-dm[1238]: (zenity:1705): Gtk-WARNING **: cannot open display: :0
Apr 18 10:33:09 ubuntu systemd[1]: Started Ubuntu live CD installer.

Apr 18 10:33:03 ubuntu ubiquity[1556]: Ubiquity 2.21.57
Apr 18 10:33:04 ubuntu ubiquity[1556]: Exception in GTK frontend (invoking crash handler):
Apr 18 10:33:04 ubuntu ubiquity[1556]: Traceback (most recent call last):
Apr 18 10:33:04 ubuntu ubiquity[1556]: File "/usr/lib/ubiquity/bin/ubiquity", line 649, in <module>
Apr 18 10:33:04 ubuntu ubiquity[1556]: main(oem_config)
Apr 18 10:33:04 ubuntu ubiquity[1556]: File "/usr/lib/ubiquity/bin/ubiquity", line 635, in main
Apr 18 10:33:04 ubuntu ubiquity[1556]: install(query=options.query)
Apr 18 10:33:04 ubuntu ubiquity[1556]: File "/usr/lib/ubiquity/bin/ubiquity", line 264, in install
Apr 18 10:33:04 ubuntu ubiquity[1556]: wizard = ui.Wizard(distro)
Apr 18 10:33:04 ubuntu ubiquity[1556]: File "/usr/lib/ubiquity/ubiquity/frontend/gtk_ui.py", line 239, in __init__
Apr 18 10:33:04 ubuntu ubiquity[1556]: self.watch = Gdk.Cursor.new(Gdk.CursorType.WATCH)
Apr 18 10:33:04 ubuntu ubiquity[1556]: TypeError: constructor returned NULL

But in this boot, the dependency cycle got broken at plymouth-quit-wait.service instead of ubiquity.service, so this might all just be weird fallout from this.

Indeed, if I boot with "maybe-ubiquity break=casper-bottom systemd.debug-shell" and remove /<email address hidden>" in initramfs before booting, then:

 - the menu always appears
 - ubiquity-dm's X runs at VT1 ! (This explains keyboard/mouse failure, failure to start X and ubiquity-dm)
 - Corresponding to the above, there is an "xorg" process with an explicit "vt1" CLI argument, so this is not about auto-detection.
 - the dependency cycle is gone

So we either

 - need to teach ubiquity-dm to run X on VT7 instead of VT1, to not get in the way of gettys, or
 - Disable <email address hidden> if it is *intended* to not have any gettys in ubiquity-dm. (But I wouldn't see a reason for that)