Created attachment 26666 [details]
screenshot of googleearth
On G45 64bit, The image of googleearth will be covered by kinda "fog" when zoom in, the closer of zoom the thicker of the "fog". Please see attachment for screenshot.
It was tested on:
OSD: Fedora release 8 (Werewolf)
with KMS and UMS, and we tested both with and without compiz.
Created attachment 26667 [details]
Created attachment 26668 [details]
*** Bug 22257 has been marked as a duplicate of this bug. ***
Yifei, any other machines also impacted? like 945?
Same issue here with 965GM.
We've also tested on 915GM and GM45 32bit, 915GM works fine but GM45 has the same problem. I'll test more platform later today.
Sorry for one mistake, we used both googleearth 4.x and 5.x in the previous test. After notice this, we retested on several platforms, all those platforms has this problem with googleearth 5.x, and 4.x works fine. Tested platforms include 915GM, GM965, G35, G45 and 945GM.
Beware that some versions of Google Earth may work around this bug based on criteria encoded in their drivers.ini files.
Hi Michel, could you be more specific? we've tweaked with googleearth's drivers.ini a little bit, but have not figure out whether it's because of drivers.ini that make 4.x works fine.
Just wanted to point out that some versions of Google Earth will use different athmosphere shaders when they detect certain drivers, in case you didn't know that yet.
Yifei, please file a separate bug for i915.
We've tested googeearth again, now it works fine on 915GM and 945GM, but the same problem still be here on GM965 G35 and G45.
I looked at this bug a bit. Google conveniently provides the source to their vertex and fragment shaders (in the shaders/ directory). In the fragment shader, fragment.texcoord.w is used as a fog factor. I played with forcing this to a constant value, and correct results were produced. I infer that the problem is the vertex shader is always setting this to a result.texcoord.w to 0.0.
The next step is to force setting result.texcoord.w to a non-zero constant in the vertex shader.
There's no such fog if "Atmosphere" was closed from View --> Atmosphere.
decreasing priority since there's a workaround by disabling Atmosphere.
btw, bug#17397 says Atmosphere causes slowness. Yifei, can you validate this on G45 and 945GM?
I have the same problem on my Foxconn G45M-S board with Ubuntu Karmic 64bit + xorg-edgers ppa.
I was able to work around it by enabling weather.
Mass version move, cvs -> git
*** Bug 23754 has been marked as a duplicate of this bug. ***
It works well with following commits:
I tested following commits:
libdrm (more recent):
mesa: (this is git master, more recent I guess, the mesa_7_6_branch pulled into very recently)
xserver (git master)
and kernel is quite outdated (2.6.31-rc5)
While I confirm that GE works great (consumes only 250 MB of memory, quite fast, streetview works nicely, but this shader bug exists)
I'll test with your revisions as well soon.
And I have G965 (intel DG965RY motherboard)
Beware that latest GE beta doesn't reproduce that bug.
I think they have made a workaround.
However normal stable 5.0 version has this bug.
Author: Eric Anholt <email@example.com>
Date: Tue Nov 10 10:54:15 2009 -0800
i965: Unalias src/dst registers for SGE and friends.
Fixes piglit vp-sge-alias test, and the googleearth ground shader. \o/
Verified with currently commit.