Does anybody know about any setup option (gnome/gtk/radeon driver) that may cause corruptions as shown in this bug:
https://bugzilla.redhat.com/show_bug.cgi?id=679579
PNG: https://bugzilla.redhat.com/attachment.cgi?id=480275
I tried to use various video modes, color depths etc, but there is no difference.
Thanks for any hint.
Adam Pribyl
On Wed, 2011-04-06 at 17:00 +0200, Adam Pribyl wrote:
Does anybody know about any setup option (gnome/gtk/radeon driver) that may cause corruptions as shown in this bug:
https://bugzilla.redhat.com/show_bug.cgi?id=679579
PNG: https://bugzilla.redhat.com/attachment.cgi?id=480275
I tried to use various video modes, color depths etc, but there is no difference.
Thanks for any hint.
Adam Pribyl
I'm experiencing the same issue here, using radeon's drivers on an ATI Radeon 9200 Pro.
In my opinion it's related to the OpenGL version we're running; probably the Gnome Shell (specifically Mutter) requires some OpenGL extension that were introduced with OpenGL 1.4 that we don't have, since we're stuck with OpenGL 1.3.
You may find which OpenGL version you're running with $ glxinfo | grep -i opengl
OpenGL version string: 1.3 Mesa 7.10.1
Regards.
On Thu, 7 Apr 2011, Massimo Gengarelli wrote:
I'm experiencing the same issue here, using radeon's drivers on an ATI Radeon 9200 Pro.
In my opinion it's related to the OpenGL version we're running; probably the Gnome Shell (specifically Mutter) requires some OpenGL extension that were introduced with OpenGL 1.4 that we don't have, since we're stuck with OpenGL 1.3.
You may find which OpenGL version you're running with $ glxinfo | grep -i opengl
OpenGL version string: 1.3 Mesa 7.10.1
And this is something that is not going to change any time... right?
Regards.
Then we are out of.. the game. Last ATi to be dropped around me.
-- Massimo Gengarelli gengarel@cs.unibo.it
Adam Pribyl
2011/4/7 Adam Pribyl:
OpenGL version string: 1.3 Mesa 7.10.1
And this is something that is not going to change any time... right?
You are mistaken. Current F15 including updates-testing (enabled by default) features mesa-libGL-7.11-0.5.20110401.0.fc15:
$ glxinfo | grep -i opengl OpenGL version string: 2.1 Mesa 7.11-devel
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 04/08/2011 09:58 AM, Christoph Frieben wrote:
2011/4/7 Adam Pribyl:
OpenGL version string: 1.3 Mesa 7.10.1
And this is something that is not going to change any time... right?
You are mistaken. Current F15 including updates-testing (enabled by default) features mesa-libGL-7.11-0.5.20110401.0.fc15:
$ glxinfo | grep -i opengl OpenGL version string: 2.1 Mesa 7.11-devel
providing the entire output, might give more useful information Note: I'm using mesa from git, so my info is slightly different
glxinfo | grep OpenGL EE r600_pipe.c:430 r600_get_param - r600: unknown param 45 OpenGL vendor string: X.Org OpenGL renderer string: Gallium 0.4 on AMD RV635 OpenGL version string: 2.1 Mesa 7.11-devel (git-d456178) OpenGL shading language version string: 1.20 OpenGL extensions:
Kevin
- -- Get my public GnuPG key from http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x7D0BD5D1
On Fri, Apr 08, 2011 at 17:58:43 +0200, Christoph Frieben christoph.frieben@googlemail.com wrote:
2011/4/7 Adam Pribyl:
OpenGL version string: 1.3 Mesa 7.10.1
And this is something that is not going to change any time... right?
You are mistaken. Current F15 including updates-testing (enabled by default) features mesa-libGL-7.11-0.5.20110401.0.fc15:
$ glxinfo | grep -i opengl OpenGL version string: 2.1 Mesa 7.11-devel
That depends on which card you have. I get: bash-4.2$ glxinfo | grep -i opengl OpenGL vendor string: Tungsten Graphics, Inc. OpenGL renderer string: Mesa DRI R200 (RV280 5961) 20090101 TCL DRI2 OpenGL version string: 1.3 Mesa 7.11-devel OpenGL extensions:
On Fri, 2011-04-08 at 15:20 -0500, Bruno Wolff III wrote:
On Fri, Apr 08, 2011 at 17:58:43 +0200, Christoph Frieben christoph.frieben@googlemail.com wrote:
2011/4/7 Adam Pribyl:
OpenGL version string: 1.3 Mesa 7.10.1
And this is something that is not going to change any time... right?
You are mistaken. Current F15 including updates-testing (enabled by default) features mesa-libGL-7.11-0.5.20110401.0.fc15:
$ glxinfo | grep -i opengl OpenGL version string: 2.1 Mesa 7.11-devel
That depends on which card you have. I get: bash-4.2$ glxinfo | grep -i opengl OpenGL vendor string: Tungsten Graphics, Inc. OpenGL renderer string: Mesa DRI R200 (RV280 5961) 20090101 TCL DRI2 OpenGL version string: 1.3 Mesa 7.11-devel OpenGL extensions:
Christoph was pointing out that Mesa had gone from 7.10 to 7.11, I believe.
On 4/7/11 3:54 AM, Massimo Gengarelli wrote:
I'm experiencing the same issue here, using radeon's drivers on an ATI Radeon 9200 Pro.
In my opinion it's related to the OpenGL version we're running; probably the Gnome Shell (specifically Mutter) requires some OpenGL extension that were introduced with OpenGL 1.4 that we don't have, since we're stuck with OpenGL 1.3.
I respect your opinion and all, but I'm pretty sure that's just a bug not a missing feature. To which GL 1.4 feature do you refer?
- ajax
On Fri, 2011-04-08 at 10:57 -0400, Adam Jackson wrote:
I respect your opinion and all, but I'm pretty sure that's just a bug not a missing feature. To which GL 1.4 feature do you refer?
- ajax
Could be something related to GLX_textures_non_power_of_two; the screenshot posted by Adam is just showing one of the many bugs that are lying under the shell: for example Mutter isn't capable of properly drawing drop shadows around the windows (he just draws some white textures here and there), the dash shows some ugly white textures where the small light should be (indicating that the application is opened) and the workspace switcher flashes while coming out from the right side in overlay mode.
I'm almost sure it's not a Fedora's bug because I'm running Gnome Shell successfully on my Asus EeePC 1000H (which has a i915 Intel video card, capable of running OpenGL 1.4+).
Obviously, I'd like to be wrong on this because I'm really loving the Shell and I'd like to use it on my Desktop PC too :-P
On Fri, 2011-04-08 at 17:53 +0200, Massimo Gengarelli wrote:
I'm almost sure it's not a Fedora's bug because I'm running Gnome Shell successfully on my Asus EeePC 1000H (which has a i915 Intel video card, capable of running OpenGL 1.4+).
The logic here is:
* Card A / Driver B is capable of OpenGL 1.3 * Card X / Driver Y is capable of OpenGL 1.4 * GNOME Shell runs well on Card X / Driver Y * GNOME Shell runs badly on Card A / Driver B * Therefore, GNOME Shell must run well with OpenGL 1.4 and badly with OpenGL 1.3!
The reason this is a fallacy is that you haven't established that the level of OpenGL support is the *only* difference between the configurations in question, and it certainly isn't. There are all sorts of others. As Ajax says, it's likely not to do with the exact level of OpenGL support available, it's more likely just a bug in the problematic hardware / driver combo.
On Fri, 2011-04-08 at 09:02 -0700, Adam Williamson wrote:
The reason this is a fallacy is that you haven't established that the level of OpenGL support is the *only* difference between the configurations in question, and it certainly isn't. There are all sorts of others. As Ajax says, it's likely not to do with the exact level of OpenGL support available, it's more likely just a bug in the problematic hardware / driver combo.
I strongly hope you're right, but I really think that the bugs I listed in my previous mails (especially those related to Mutter) depend on some OpenGL's features that are missing.
I'm assuming that because I've experienced the same bugs using Ubuntu's new shell (Unity). Unity has some checks for the OpenGL version hardcoded, hacking through the code and removing them made the shell start up with the same identical bugs (white lights where drop shadows should have been, ...).
Digging into the code I discovered that the way Unity draws drop shadows around the panels is by using some GL functions that require the extension GLX_texture_non_power_of_two to be active.
Mutter suffers the same bugs.
More than that, my video card (ATi Radeon 9200 PRO) will never be able to use the missing extensions, just because the hardware is way too old. I'm not complaining against GNOME developers or Radeon's drivers developers, both are doing an excellent work and it's quite normal that hardware becomes obsolete after more than 6 years.
Again, I'd really love to be wrong because I'm really missing GNOME shell here ;-)
On Fri, Apr 8, 2011 at 9:31 PM, Massimo Gengarelli gengarel@cs.unibo.it wrote:
On Fri, 2011-04-08 at 09:02 -0700, Adam Williamson wrote:
The reason this is a fallacy is that you haven't established that the level of OpenGL support is the *only* difference between the configurations in question, and it certainly isn't. There are all sorts of others. As Ajax says, it's likely not to do with the exact level of OpenGL support available, it's more likely just a bug in the problematic hardware / driver combo.
I strongly hope you're right, but I really think that the bugs I listed in my previous mails (especially those related to Mutter) depend on some OpenGL's features that are missing.
I'm assuming that because I've experienced the same bugs using Ubuntu's new shell (Unity). Unity has some checks for the OpenGL version hardcoded, hacking through the code and removing them made the shell start up with the same identical bugs (white lights where drop shadows should have been, ...).
Digging into the code I discovered that the way Unity draws drop shadows around the panels is by using some GL functions that require the extension GLX_texture_non_power_of_two to be active.
Mutter suffers the same bugs.
More than that, my video card (ATi Radeon 9200 PRO) will never be able to use the missing extensions, just because the hardware is way too old. I'm not complaining against GNOME developers or Radeon's drivers developers, both are doing an excellent work and it's quite normal that hardware becomes obsolete after more than 6 years.
Again, I'd really love to be wrong because I'm really missing GNOME shell here ;-)
pot_textures aren't are hard requirement so yes you are wrong.
On Fri, 2011-04-08 at 21:55 +0200, drago01 wrote:
pot_textures aren't are hard requirement so yes you are wrong.
well, he may not be - it may be that not having them causes it to hit a fallback path which we haven't tested well and which is buggy.
in any case, the sensible path here would seem to be to file a bug on it and make sure ATI driver and mesa devs are CCed.
On Fri, 2011-04-08 at 21:55 +0200, drago01 wrote:
pot_textures aren't are hard requirement so yes you are wrong.
Nice to know, but that doesn't change the fact that two new-generation window managers (Mutter and Compiz with Unity plugin) that are both lying on OpenGL are showing the same identical bugs on a video card that can run OpenGL 1.3 while they are running fine on a video card that can run OpenGL 1.4+.
I could be wrong in stating that the required missing extension is pot_textures, but I don't think I'm wrong when I say that this is strictly related to OpenGL.
Probably the missing extensions could be implemented via software emulation but *for now*, who's not having a combination of video card/driver which is capable of running OpenGL 1.4+ is not getting Mutter (and therefore is missing the new Gnome-Shell).