Bugs in xf86-video-intel 1.9.91 (2.0 RC1)

[Ken Mandelberg]

Keith Packard wrote:
> On Wed, 2007-03-07 at 12:41 -0500, Ken Mandelberg wrote:
> 
>> Do you have any ideas why I m killing the Xserver with the subsequent 
>> git, or the 1.9.91 (2.0 RC1) tarball release, with or without the extra 
>> monitor section? It must be related to one of the last code changes.
> 
> I don't have any idea; can you build the server and driver for debugging
> and get a more complete stack trace under gdb? And, can you send the
> xorg.conf file which causes the segfault?
> 

Well I did use gdb and figured out the reason for the segfault. My 
xorg.conf did not specify a load of dri.

I830ScreenInit tests with I830CheckDRIAvailable and sets 
pI830->directRenderingDisabled. This leaves  pI830->XvEnabled false, and 
causes i830_allocate_overlay to skip over initializing pI830->xaa_linear.

Later in I830ScreenInit pI830->XvEnabled is reset to !pI830->XvDisabled 
and finally it goes on to reference a null pointer at

  pI830->xaa_linear->offset / pI830->cpp


The logic was quite a bit different in the February git code, and the 
issue didn't turn up with the same xorg.conf sans dri.

I added the load dri to my xorg.conf, and now the 1.9.91 driver works.

I did run into some issue with xvideo corruption, that was related to my 
specification of memory allocation params in xorg.conf. In the February 
code this was necessary, but with 1.9.91 the auto allocation code is 
working properly, and the manual specs interfered with it.

So I seem to have a working driver, with the only xorg.conf tweak needed 
is one to manually specify horizontal and vertical refresh limits in the 
special vga monitor section. Thats needed to not rule out the 
interesting modes.