[Intel-gfx] [PATCH] change initial modesetting if outputs are aligned in 1 dimension

[Florian Mickler]

On Wed, 01 Jun 2011 15:30:03 -0400
Adam Jackson <ajax at redhat.com> wrote:

> On 6/1/11 6:06 AM, Florian Mickler wrote:
> > Recently the kernel started reporting my outputs in a different ordering due to
> >
> >      commit cb0953d734
> >      (drm/i915: Initialize LVDS and eDP outputs before anything else)
> >
> > Which made X choose a "wrong" resolution for my VGA display. Since they are
> > aligned horizontally, I wish them to be aligned in vertical
> > Resolution only.
> >
> > Before this patch, the sum of squared distances would force my VGA display
> > (1680x1050 native resolution) to 1280x1024 (non-native) due to my internal
> > display beeing considered first and 1400x1050 as native resolution.
> >
> > This was not an issue the other way around (VGA beeing first) because 1400x1050
> > is nearest to 1680x1050 anyway.
> >
> > This patch changes the heuristic to only align resolution vertically if the
> > displays are horizontally aligned, or vice versa.
> >
> > Signed-off-by: Florian Mickler<florian at mickler.org>
> > ---
> >
> > Ok, Adam... seems I lost the staring contest... :)
> > What about something like this?
> 
> At this point, given the near-unity overlap of "RANDRful drivers" and 
> "usable KMS support", I think I'd prefer something more like:
> 
> http://pkgs.fedoraproject.org/gitweb/?p=xorg-x11-server.git;a=blob_plain;f=xserver-1.6.99-right-of.patch;h=a0c9e7f64b98a091f64faaf5a8a432e2122e25f9;hb=HEAD
> 
> particularly once per-crtc pixmaps land.
> 
> - ajax

I think using horizontal spanning as a default is a good idea. 

Also bringing all outputs up in their preferred mode could be the right
move.

[That commit in question wouldn't help my case, since I have a "Right Of" in the
xorg.conf and thus it would use a different code-path (right now).]


When there is no preferred mode, xf86ClosestMode (probably enhanced to
respect horizontal or vertical relation in some form) would still be
needed... choosing of mode and positioning of outputs
relative to each other should probably be decoupled...

Regards,
Flo