[ubuntu-x] gnome-settings-daemon: the new Xserver?

[Tormod Volden]

On Fri, Oct 1, 2010 at 4:28 PM, Sebastien Bacher wrote:
> It was not just before the freeze and the issue was that the upstream

Launchpad says "Uploaded by: Sebastien Bacher on 2010-09-13" so it was
three days before... Never mind.

> default to turn off external screens lead to people have no monitor
> turned on in some configurations

You wrote this in bug 640807 comment 7 also, but still without any
reference to bug reports.

> The change which created the issue seems to be
> http://git.gnome.org/browse/gnome-settings-daemon/commit/?id=67958ef6faab5797d5c5ad939db36f393706984a
>
> Not sure what would be the right policy though, maybe the default should
> be to let xorg set the screen if no configuration file is on disk and
> have a key to activate those options to force monitors to be on

I understand the explanation that xorg should "just work" and leave
policy to the desktop, thus g-s-d here. BTW Sebastian, this was in the
thread on the ML which you might not have seen - maybe you should
subscribe to it :)

But what I find backwards here is that xorg has figured out a
preferred mode for the screen, and then g-s-d goes on to select a
sub-optimal mode. And it seems from the bug reports that it is
happening because of the activation of external screens. Of course
since the 9.10 gdm, this also happens in the login screen, which can
make it more difficult if something goes wrong.

In your "docked laptop with lid closed" example, I think xorg
activates both screens (clone). So IMO g-s-d should just leave it as
it is or read out correctly what xorg has set up. Until the user
selects something else of course. The autodetection is not perfect for
all video drivers, so g-s-d should be able to override it, but it
should profit from the driver intelligence when there is one. From
looking at above git commit, it looks like g-s-d must turn external
screens on or off, with no option to leave it as it is, or check with
xrandr what the current status is. And I think both true and false are
bad defaults, like "there is no correct answer because the question is
wrong" :)

I do not fully understand what goes wrong in these bug reports, maybe
the new default only exposes other bugs in g-s-d and its way of
parsing and choosing optimal modes.

Tormod