xf86-video-intel 2.0.0 rotation problems

[Barry Scott]

Let me summarize where  my experiments with  rotation and intel
driver have got to.

I believe I have the configuration file with the right options to setup
rotation at start up: Option rotate in the monitor section.

I can force the intel driver to use the mode I want at start up by adding
Option PreferredMode to the monitor section. But this only works
in normal rotation. In rotation left I get 800x600 not 1440x900 that
is configured.

Intel driver prints warning that these options are not used but the
behaviour shows they have an effect.

(WW) intel(0): Option "PreferredMode" is not used

The 99.9% CPU use by Xorg is fixed by Keith's patch to remove the
call to  i830WaitSync(pScrn). I have not been able to check for tearing
problems given the other problems.

If I use XAA acceleration on a 965 the code SEGV because of an logic
error. I've mailed details to this list under subject
"exa_965_state used when in XAA mode".

If I use EXA acceleration on a 965 it does not crash but the display is
tiled wrong. Using:

    /usr/X11R6/bin/xwd -silent -root -screen | \
        /usr/bin/xwdtopnm 2>/dev/null | \
            /usr/bin/pnmtopng 2>/dev/null >$1

I see that the frame buffer contents are as I expect. But he image
on the screen is in about 3 bands that are like wallpaper that is
not lined up. I see duplicated pieces of the screen. There is also
sections of the screen with wide black and white strips.

I cannot allow the Xorg server to reset itself as the intel driver
loses all the modes on the VGA output. Why does the Xorg server
reset when the last client disconnects? What would be the impact of
patching Xorg server to stop doing this? In my use of Xorg this
behaviour is leading to the exposure of more bugs and seems to
have no benefit to me.

Both 2.0.0 and git drivers have these problems.

What more information and help can I provide to get
rotation working?

Barry