Testing mesa 7.6 RC on Ubuntu karmic

[tbpringle]

Okay now I see that some of my posts do not need moderator review but others do. very confusing. I guess the original replies I made had links in them and they need to be reviewed.

[highlandsun]

I'm getting bits of corruption with the latest xorg-edgers stuff and a 2.6.32-rc4 kernel on xubuntu karmic beta. This is on an Asus M6Ne with Radeon Mobility 9700. The corruption occurs one glyph at a time. For example, mousing over the minimize/maximize buttons of a window titlebar, the buttons get replaced with hash instead of highlighting. That's benign because moving away restores the original button.

The worse problem is that individual characters of the default font start to be corrupted. It starts with a single character, apparently at random, and the longer I run the session the more characters become corrupted. So for example a lower case 'c' becomes corrupted - it is drawn as garbage (identically) everywhere it occurs on every window of every app. It also occurs gradually to various icons in Seamonkey.

My worry here is that these random corruptions aren't just hitting graphics/font memory, and may be corrupting my actual data. Anyone seen this problem already?

[jntesteves]

Can you please tell which glitches? Are they reported somewhere? Since the new 7.6 is "on trial" it is useful to have documented what it fixes.

Hi, with Ubuntu 9.10 beta with compiz on an Intel notebook (don't know the exact chip) when I clicked the clock on the gnome-panel the shadow didn't disappear after I clicked it again and the calendar disappeared. moving other windows over the calendar shadow made it disappear. I didn't report the bug as I immediately downloaded the daily build with the new mesa and the problem was gone.

[tormod]

I'm getting bits of corruption with the latest xorg-edgers stuff and a 2.6.32-rc4 kernel on xubuntu karmic beta.

Note that xorg-edgers tracks master trunk (mesa 7.7-to-be) whereas x-updates tracks the 7.6 branch. Stock karmic has the 7.6.0 release with some patches (for crashes and corruptions which are not yet fixed in the 7.6 branch).

[highlandsun]

Note that xorg-edgers tracks master trunk (mesa 7.7-to-be) whereas x-updates tracks the 7.6 branch. Stock karmic has the 7.6.0 release with some patches (for crashes and corruptions which are not yet fixed in the 7.6 branch).

Are you suggesting I should have gone with stock Karmic, or maybe x-updates? Stock karmic was freezing randomly on me.

[legume]

The worse problem is that individual characters of the default font start to be corrupted. It starts with a single character, apparently at random, and the longer I run the session the more characters become corrupted. So for example a lower case 'c' becomes corrupted - it is drawn as garbage (identically) everywhere it occurs on every window of every app. It also occurs gradually to various icons in Seamonkey.

My worry here is that these random corruptions aren't just hitting graphics/font memory, and may be corrupting my actual data. Anyone seen this problem already?

Is kms is enabled? I don't use ubuntu so don't know what's what.

I have seen what you describe and it never affected any data. For me it is be caused by accelerated download from screen and can be exposed by an xserver/driver change.

If you aren't using kms then using kms may actually fix it - if you are using kms then a more upto date kernel may help.

Chances are that

Option "ExaNoDownloadFromScreen"

or possibly

Option "ExaNoUploadToScreen"

in the Device section of your xorg.conf will fix it.

[tormod]

Are you suggesting I should have gone with stock Karmic, or maybe x-updates? Stock karmic was freezing randomly on me.

Just hinting you might be bit off-topic unless you have tested 7.6 as well Make sure you get 7.6.0-1ubuntu4 from stock karmic. As I said, it has fixes for some crashers. Please file a bug if it is still freezing.

Of course very interesting to know if there are freezes or crashes in 7.6 that are fixed in trunk.

[highlandsun]

Is kms is enabled? I don't use ubuntu so don't know what's what.

I have seen what you describe and it never affected any data. For me it is be caused by accelerated download from screen and can be exposed by an xserver/driver change.

If you aren't using kms then using kms may actually fix it - if you are using kms then a more upto date kernel may help.

Chances are that

Option "ExaNoDownloadFromScreen"

or possibly

Option "ExaNoUploadToScreen"

in the Device section of your xorg.conf will fix it.

Thanks, will give those a try. I'm using KMS on 2.6.32-rc4, there is no more up to date kernel at the moment...

[highlandsun]

Just hinting you might be bit off-topic unless you have tested 7.6 as well Make sure you get 7.6.0-1ubuntu4 from stock karmic. As I said, it has fixes for some crashers. Please file a bug if it is still freezing.

Of course very interesting to know if there are freezes or crashes in 7.6 that are fixed in trunk.

I've so far had no freezes or crashes with the xorg-edgers packages. Just the corrupted glyphs.

Also, with stock Karmic, text consoles didn't work at all, all I saw was a hash of green lines on the top half of the screen. Apparently switching to a VT didn't actually drop out of graphics mode, and the green hash is how the text bytes got rendered on a graphical frame buffer. Note that KMS was not enabled on the stock Karmic build. When I enabled KMS and loaded fbcon, this problem went away. (But loading fbcon without KMS didn't fix it either.)

[legume]

Thanks, will give those a try. I'm using KMS on 2.6.32-rc4, there is no more up to date kernel at the moment...

Ahh, it does look like that may have a now reverted commit that caused some corruption - but then it was more images than fonts, though you never know, next rc may be better for you.

I am using the drm-next branch which has the fix but it doesn't look like linus has pulled it into his tree yet.

http://git.kernel.org/?p=linux/kerne...log;h=drm-next

http://git.kernel.org/?p=linux/kerne....git;a=summary

(putting drm in the search for linus shows last drm 7 days ago and the revert was 3 days ago, rc-4 was set 4 days ago)