On Tue, February 22, 2011 04:25, Indan Zupancic wrote:

Hi,

On Sun, February 20, 2011 14:21, Daniel Vetter wrote:

...

Well, don't start jumping around, yet. These patches are just to rule out some theories. Now: Is it fixed with just the 2nd patch alone or do you need both patches? This is very important, so please test extensively whether there are really no corruptions with just the 2nd patch.

I managed to create some corruption with an xterm above xpdf. It looks different than the original corruption, so I think it's safe to say it's a different bug:

http://img593.imageshack.us/i/ss1298340823.png http://img203.imageshack.us/i/ss1298340776.png

This was without xcompmgr running, I don't think I would have seen it otherwise. Actually, it turns out it's really easy to trigger as well, so I can more easily test patches now. Unfortunately I wasn't smart enough to store the bisecting kernels, gah!

I tried with both your patches, as well as the HIC poking patch, but I still get it.

I'll try to pinpoint more exactly when this started to happen.

Okay, I tried older kernels and I get it even with 2.6.34, so I think it's a bug in userspace, like intel-dri, or perhaps the intel X driver (though I get the same with an older version).

Greetings,

Indan