945GM + Xrandr 1.2 + modesetting: color corruption on LVDS

[Gregory Maxwell]

On 2/12/07, Keith Packard <keithp at keithp.com> wrote:
> On Sun, 2007-02-11 at 15:59 +0100, HENRY David wrote:
> > I have a color corruption on my LVDS panel when I run xrandr. The
> > screens looks like if color depth were at 8 bits (I'm running at 24
> > bits). If I switch to another screen mode and come back to my "normal"
> > desktop (like a VT switch, or running Quake 3 fullscreen and exiting,
> > etc.), the problem disappears. As soon as I run xrandr (even just for a
> > query), the colors are degraded again.
>
> There are several subtle timing-related issues in setting modes on the
> LVDS output. We've tested the current code on many displays and have
> gotten it working well after some tweaking (and, after constant
> reference to the specifications). I don't know what version of the
> driver you're running, but newer is generally better, so please try
> building the current version from the modesetting branch rather than
> relying on pre-built drivers at this point; we haven't made any formal
> releases, so you're just getting Daniel's 'best guess' as to which
> version to ship.
>
> And, if you do have problems with the latest code, please include as
> much information about the machine you're using as possible. We may be
> able to get access to a matching machine and debug the problem directly.

Hi Keith, I'm seeing the same thing on my 945GM.
Normally my screen looks like it's in 16 bit mode (which I brought up
on the list a few days ago, and you had me try a git version of the
modesetting driver which I'm still on), if I run the xrandr command..
even without arguments,  the quality clearly degrades further ...
switching to a VT and switching back fixes it until I run the command
again.

This give me a little hope that my 16 bit issue is driver related. :)

In any case, I'd be glad to conduct any testing or try any code.