Comment 217 for bug 395932

Well, there was a set of discrete Xv fixes which went in after 2.9.0 was cut but before the massive refactoring work for removing UMS hit. It comprised 17 patches total, which I've unified into the patch attached here.

$ diffstat debian/patches/102_xv-fix-8xx-video.patch
 drmmode_display.c | 9
 i830.h | 3
 i830_driver.c | 2
 i830_video.c | 1513 +++++++++++++++++++++++++-----------------------------
 i830_video.h | 20
 i915_video.c | 33 -
 i965_video.c | 48 -
 7 files changed, 755 insertions(+), 873 deletions(-)

This is mostly confined to the i8xx code, but there are some changes to the 9xx video code, which alter the algorithmic logic, so I can't say with 100% certainty that the changes are going to be regression-free for 9xx, which makes me reticent to pull this in for Lucid. There is also a lot of refactoring going on, which probably accounts for most of the size of the patch, and unfortunately also makes it tough to review in detail.

But... Since this backport was pretty straightforward, I've gone ahead and packaged and set up a PPA of this here:
https://edge.launchpad.net/~bryceharrington/+archive/white

If 8xx users can test this and say that it gives a perfect fix to Xv support (i.e. no secondary issues), AND we can get some 9xx users to verify it causes no regressions, then we can look at including this in Lucid. Otherwise, 8xx users can use this ppa or