Gentoo Wiki ArchivesGentoo Wiki

AIGLX/Troubleshooting

Contents

Troubleshooting

Detect the Xorg errors and warnings concerning AIGLX:

grep -e EE -e WW /var/log/Xorg.0.log

Determine generally what Xorg thinks about AIGLX:

grep -i aiglx /var/log/Xorg.0.log

Obtain verbose information from glxinfo:

LIBGL_DEBUG=verbose glxinfo

Determine how well DRI works (that is, what frames per second it is capable of):

glxgears

inline unit growth error

inline-unit-growth error is a known problem (and a few other threads I can no longer find), so if the following happens:

[...].deps/libfbmmx_la-fbmmx.Tpo -c fbmmx.c  -fPIC -DPIC -o .libs/libfbmmx_la-fbmmx.o
cc1: error: invalid parameter `inline-unit-growth'
cc1: error: invalid parameter `large-function-growth'
make[1]: *** [libfbmmx_la-fbmmx.lo] Error 1
make[1]: Leaving directory `/tmp/portage/portage/xgl-0.0.1_p20060418/work/xgl-xo
rg/fb'
make: *** [all-recursive] Error 1

!!! ERROR: x11-base/xgl-0.0.1_p20060418 failed.
Call stack:
  ebuild.sh, line 1894:   Called dyn_compile
  ebuild.sh, line 941:   Called src_compile
  ebuild.sh, line 1609:   Called x-modular_src_compile
  x-modular.eclass, line 320:   Called x-modular_src_make

Fix it by adding -mmx to x11-base/xgl in your /etc/portage/package.use file.

Warning: Latest Xgl ebuild lacks an MMX USE flag! A fix that works for me is to ensure your CFLAGS in /etc/make.conf does not include --force-adr (http://forums.gentoo.org/viewtopic-t-470211-postdays-0-postorder-asc-highlight-xgl-start-600.html)

Digest verification failed

SHA256 is a known bug, so if the following happens:

Would you like to merge these packages? [Yes/No] y
>>> Emerging (1 of 1) media-libs/mesa-6.5.1_alpha20060418 to /
>>> checking ebuild checksums
>>> checking auxfile checksums
!!! Digest verification failed:
!!! /usr/portage/local/layman/portage-xgl/media-libs/mesa/files/makedepend-location.patch
!!! Reason: Failed on SHA256 verification
!!! Got: 3541f6528b402328886e3169169fc400db11ed91d39ddd2b5d3b1f6be77d7ff5
!!! Expected: c0ec505e5fb263d4cde901ce9afe3197346acf67372460a2177231be8425e0ab

Fix it by simply running the following command to fix the SHA256 error:

ebuild /usr/local/overlays/xgl-coffee/media-libs/mesa/mesa-6.5.1_alpha20060418.ebuild digest

X_GLXvop_BindTexImageEXT error

For AMD64 users who are having problems compiling Mesa because of the following error:

error: 'X_GLXvop_BindTexImageEXT' undeclared (first use in this function)"

Try doing the following:

cd /usr/include/GL
ln -sf /usr/lib64/opengl/global/include/glext.h glext.h
ln -sf /usr/lib64/opengl/xorg-x11/include/gl.h gl.h
ln -sf /usr/lib64/opengl/global/include/glxext.h glxext.h
ln -sf /usr/lib64/opengl/xorg-x11/include/glx.h glx.h
ln -sf /usr/lib64/opengl/xorg-x11/include/glxmd.h glxmd.h
ln -sf /usr/lib64/opengl/xorg-x11/include/glxproto.h glxproto.h
ln -sf /usr/lib64/opengl/xorg-x11/include/glxtokens.h glxtokens.h

If the error persists, do the same with /usr/local/include/GL

cmdveclinear error

mesa-6.5.1_alpha20060701 CVS version seems to be broken. To fix the error, emerge mesa with VIDEO_CARDS set to none, i.e.:

VIDEO_CARDS=none emerge -av mesa

gcompizthemer/cgwd segfault and other cgwd problems

If you ran into this:

root@localhost ~ # gcompizthemer
Segmentation fault
root@localhost ~ # cgwd
cgwd: Connection Error (Unable to determine the address of the message bus (try 'man  dbus-launch' and 'man dbus-daemon' for help))
Segmentation fault

You have to add dbus-launch before each of these commands, i.e.

dbus-launch gcompizthemer
dbus-launch cgwd

If you start cgwd via script (/usr/local/bin/compiz-decorator), don't forget to change that.

do not forget to update your dbus and libwnck!!!

GLcore

Don't put Load "GLcore" in the modules section of your xorg.conf. It gets pulled in as a dependency anyway and with Xgl there is apparently a problem with it if it gets loaded before some module (GLX?) (I haven't narrowed it down) that is requiring it.

Patch XYZ failed during unpack

This usually means that a cvs update broke a patch. The first thing to try is to comment it out in the ebuild and redo the digest. If that doesn't work there is no easy solution. You can check the forum and Hanno's blog to see if someone found a solution, if not, probably the best place to post is Hanno's blog as he seems te regularly read it and fix his overlay.

GLX_EXT_texture_from_pixmap is missing

compiz outputs this:

compiz: GLX_EXT_texture_from_pixmap is missing
compiz: Failed to manage screen: 0
compiz: No managable screens found on display localhost:2.0

On ubuntuforums.org, it is stated that this error can occur from a wrong libGL.so (only the libGL.so from mesa can do the trick). Check into where mesa has installed its libGL.so files to and set the export LD_LIBRARY_PATH, in compizrc, to this instead of /usr/lib/opengl/xorg-x11/lib/ which may have a libGL.so that is a symbolic links to nothing or to the Nvidia libGL.so (the one we don't want). Try

LD_LIBRARY_PATH=/usr/lib compiz --replace gconf &

It is reported that using the right library, compiz should work on a GeForce FX 5200 -- doesn't work for me, though (not being too much of an expert as far as libraries are concerned, i assume that setting LD_LIBRARY_PATH does roughly the same as LD_PRELOAD, though on a more general level).

If the above doesn't work then try remerging mesa and xgl after switching back to the xorg-x11 libGl.

eselect opengl set xorg-x11
emerge -va mesa xgl
eselect opengl set nvidia # (for nvidia users, or)
eselect opengl set ati # (ati users)

Comment: While this has solved for many, it hasn't for me. If I emerge -C nvidia-drivers, everything that depended on libGL.so.1 failed to load as a user, but not as root. Why? After zillions of emerges, the problem was: LD_ASSUME_KERNEL. Try and unset LD_ASSUME_KERNEL and run compiz again. :-)

Comment: To force loading the correct (xorg-x11) libGL this works for me:

LD_PRELOAD=/usr/lib/opengl/xorg-x11/lib/libGL.so compiz --replace gconf


Another possible cause is not emerging the correct version of Mesa. Be sure you have it unmasked properly and are pulling in a version greater than mesa-6.4.2. It should be mesa-6.5.1_alpha*.

The above broke my working Xgl (on amd64 and FX5200). Xgl needs to be compiled against nvidia (or ati, should it be your vendor), compiz against mesa; Hence the opengl implementation selected is responsible for explicitly defining the pltform against which the compilation will take place in any occasion. So be aware on switching implementations between 'x0rg-x11' and 'your_vendor' for useable compilations- if in doubt, re-emerge in the correct order, and against the correct implementation once more.

Be advised that in some occasions, while emerging, the system switches between implementations automatically- I haven't found a way to bypass this other than the traditionall ^c "hotkey" =) whilst emerging so any feedback is welcome on this issue- watch out for any debugging output while emerging.

The above (the ati implementation) was also tested on an amd turion 64 HP pavillion dv5063ea (I estimate any dv5000 series will work though) with a 200M ati pciXpress 128MB dedicated _PLUS_ another 128 from system memory (bug on HPs to get the acceleration on X), using an 64bit system, 2.6.15.1 kernel, proprietary ati-drivers 8.23.7 (to get the acceleration working) and, until now, goes on smoothly (I've been using XGL for about one month now, and have experienced no crashes at all -thankfully havent seen any error messages after the "GLX_EXT_texture_from_pixmap is missing")

In my experience and with my setup ( ati mobility 9600 M10 128 MB - ati driver 8.28.8 ) I found out that it's not necessary to switch between the opengl implementations to make mesa-xgl-compiz-quinnstorm-cgwd work. Assuming you're using the latest packages downloaded from the portage-xgl overlay all I had to do is to compile mesa without the nptl USE FLAG. After that compiz and everything else will correctly work even without the LD_LIBRARY_PATH variable.

If you're using beryl try to launch beryl-xgl (instead of beryl)!

Comment: recently, the >0.1.1 versions of beryl cause this issue on many users' systems (with ATI fglrx + xgl), regardless of distribution. None of the above fixes solves the problem. A solution has reportedly been found for Ubuntu at this link: Ubuntu Fix. So far (01/03/07) no fix has been posted for Gentoo, and the older, stable versions of beryl have been removed from portage.

Comment: In my case just following the steps described here resolved this error (go to Hardware link). However, in my xorg.conf I had to leave Mode 0666 as the only line in Section "DRI", but that's another story.

Extra Window Decorations

Sometimes Compiz will try to add window decorations to your gnome-panel. When this happens you will see a title bar labeled "bottom expanded edge panel" just above your bottom panel and a thick line just below your top panel. To fix this just open a terminal and run:

killall gnome-panel
gnome-panel &

After that the extra window decorations should be gone. This first happened to me when I changed the my gnome default fonts, but recently it has also occured on other occasions for unknown reasons.

No Window Decorations

Important: You need to enable the gnome use flag to build the decorator!

Compiz will be built with gnome-window-decorator only if cairo is installed and pango and gtk+ are built against it. gnome-base/control-center and libwnck are also required. Make sure the gnome and svg use flags are enabled for compiz. It has also been reported that in some cases you need to remove the compiz cvs directory to get it built (usually /usr/portage/distfiles/cvs-src/compiz).


Failed to load shadow images

If the gnome-window-decorator doesn't start and shows the error

 gnome-window-decorator, Failed to load shadow images

you may have set XLIB_SKIP_ARGB_VISUALS=1.

Sometimes this is the case with Mozilla Firefox, so try commenting out the line in /etc/env.d/10MozillaFirefox and login again.

Even if the decorator works, some apps' windows get decorated wrongly (dialogs without a close button) or aren't decorated at all (eg vlc).

Try to remove "gconf" from compiz's command line option, if you haven't activated compiz' plugins using gconf-editor. According the author(see above), if you start compiz with "gconf" option, all other options on the command line will be ignored.

SDL games or other apps like rdesktop that should be opaque are unevenly translucent

Open a console, export XLIB_SKIP_ARGB_VISUALS=1, then run the app from there or prepend XLIB_SKIP_ARGB_VISUALS=1 before the command

No Visual Format

When you start up Xgl you get an error like:

Fatal error:
no visual format 

Solution:

emerge nvidia-glx

Do not forget to load the glx module in your /etc/X11/xorg.conf file.

Also make sure you're using a color depth of 16 or 24.

Question: I'm having this error, but an Intel i915 chipset, so emerging the nvidia-glx package won't work for me, will it?

Black Windows

All windows turn black when running compiz.

compiz may output errors like this:

compiz: pixmap 0x200043 can't be bound to texture
compiz: Couldn't bind redirected window 0x40001a to texture

Affected by this problem:

  • Nvidia Geforce (mostly GF4 MX)
  • Intel i810 and i915
  • Radeon XPRESS 200M

For me, and probably many people with the binary nvidia drivers, this appears to have been caused by glitz building against mesa's GL libraries rather than the nvidia libraries. Hanno's instructions are:

emerge mesa glitz xgl compiz

But as soon as mesa is finished, the active GL libraries are switched to be mesa's, so glitz is built against them.

glitz and xgl need to build against the nvidia libraries, compiz needs to see the mesa libraries. Try:

Code: Emerge order for nvidia users
emerge mesa
eselect opengl set nvidia
emerge nvidia-glx glitz xgl 
eselect opengl set xorg-x11
emerge compiz
eselect opengl set nvidia

Everything works relatively correctly now on a Geforce2Go. A few issues remain. When Xgl is started from within an already running Xserver without the -fullscreen option, it looks great. When started with the -fullscreen option or from the console its super slow, but I believe that's because my card only has 16MB of ram on it. If I set the depth to 16 bits rather than 24 in xorg.conf, things improve. There are some refresh and stability issues, but it does seem to work.

For me this is not the problem. The problem could be that you had to reinstall the mesa (cvs) package each time you reinstall the xgl package, because xgl checkout the mesa cvs and if there are differences this could cause problem. So cvs change in mesa + xgl merge => mesa merge first.

The problem can also be caused due to the following. Compiz and Xgl have hardcoded DEFINEs for certain values. These values should really be inherited from a file called glxtokens.h. The following two patches fix this issue:

Patch Compiz with the followig:

--- include/compiz.h    2006-06-02 10:33:04.000000000 +0200
+++ include/compiz.h~   2006-06-02 10:33:15.000000000 +0200
@@ -878,11 +878,8 @@
 typedef FuncPtr (*GLXGetProcAddressProc) (const GLubyte *procName);

 #ifndef GLX_EXT_render_texture
-#define GLX_TEXTURE_TARGET_EXT              0x6001
-#define GLX_TEXTURE_2D_EXT                  0x6002
-#define GLX_TEXTURE_RECTANGLE_EXT           0x6003
+#include <GL/glxtokens.h>
 #define GLX_NO_TEXTURE_EXT                  0x6004
-#define GLX_FRONT_LEFT_EXT                  0x6005
 #endif

 typedef Bool    (*GLXBindTexImageProc)    (Display      *display,
 

Patch Xgl with the followig:

--- GL/glx/glxserver.h  2006-06-02 10:34:45.000000000 +0200
+++ GL/glx/glxserver.h~ 2006-06-02 10:39:15.000000000 +0200
@@ -63,15 +63,8 @@
 #include <GL/gl.h>
 #include <GL/glxproto.h>
 #include <GL/glxint.h>
-
-#ifndef GLX_TEXTURE_TARGET_EXT
-#define GLX_TEXTURE_TARGET_EXT              0x6001
-#define GLX_TEXTURE_2D_EXT                  0x6002
-#define GLX_TEXTURE_RECTANGLE_EXT           0x6003
+#include <GL/glxtokens.h>
 #define GLX_NO_TEXTURE_EXT                  0x6004
-#define GLX_FRONT_LEFT_EXT                  0x6005
-#define GLX_Y_INVERTED_EXT                  0x6006
-#endif
 
 /* For glxscreens.h */
 typedef struct __GLXdrawable __GLXdrawable;
 

Distorted Screen

Some distortion errors fixed with new drivers from here http://dri.freedesktop.org/snapshots/ They may run very slow though. Slower than the older Radeon drivers.

Half-screen problem

Some users get a "half-screen" problem:

  • Radeon Mobility M6 LY (IBM ThinkPad X32 and X31, Sony Vaio Z1, HP Omnibook 6100) -> screenshots [1], [2], [3], [4], [5]
  • Radeon Mobility M9+
  • Radeon Mobility R250 Lf (FireGL 9000) - The problem goes away compiling Mesa _WITHOUT_ any optimizations.

This seems to be specific to ATI cards. There is a bug report filed against the X.Org ATI driver at bugs.freedesktop.org that may be related.

One X31 Thinkpad user reports getting the same half-screen problem in the Gnome XGL of Kororra LiveCD, but NOT the KDE XGL of Kororra LiveCD. One guess is that it is either related to Gnome or the Radeon drivers that load with Gnome.

  • COMFIRMED: KDE XGL of Kororra LiveCD works with no half-screen problem, but you are unable to maximize a window to full screen.
  • WORKAROUND: set in xorg.conf to use 800x600 resolution, both KDE and GNOME works well.
  • BETTER WORKAROUND: patch and recompile compiz. In compiz source code (currently in file compiz/src/display.c) looking for

(it seem that this piece of code have been moved to the file compiz/src/window.c, however the fix still works)

static void recalcNormalHints (CompWindow *window) {
  ...
  maxSize = window->screen->maxTextureSize;

and change it to

static void recalcNormalHints (CompWindow *window)
  ...
  maxSize = window->screen->width;

Recompile compiz, now everything works fine under 1024x768. This is because maxTextureSize might be wrongly detected under resolution 1024x768 (due to unknown reason? mesa? radeon driver?)

  • SUGGESTION: I've found that this problem goes away when I manually set the AGPSize parameter to 16 (the correct size) for my Radeon M6 LY inside xorg.conf. I'm using Ubuntu Dapper Drake, with AIGLX and Compiz, on a Compaq N410c. (Maybe the correct amount of memory is not autodetected on these cards - therefore the offscreen surfaces are not the correct size either - resulting in corruption?) - Eddie Hung
    • On Sony Vaio V505, Eddie's suggestion works well
 Section "Device"
      Option "AGPSize" "16"
  • Sony Vaio Z1: Also suffered from the half-screen problem if I switched to resolutions > 1024x768. To workaround this issue I created /etc/drirc:
 <driconf>
    <device screen="0" driver="radeon">
       <application name="all">
           <option name="allow_large_textures" value="2" />
       </application>
    </device>
 </driconf>

I can now use 1400x1050 without any restrictions.

  • The /etc/drirc suggestion fixed the problem on an IBM ThinkPad R40 with ATI Radeon Mobility 7500
  • It likewise fixed half-sceen problem on IBM TP A31 with ATI Radeon Mobility 7500

Whole distortion

Some users get a completely distorted screen:

fuxed.png

This issue is apparently only present for ATi users. Furthermore, the issue is only present when using compiz, and not when using glxcompmgr. A solution is currently unknown.

Example: Compaq nx7010 with radeon mobility 9200 (9000 M9) and fglrx driver.

It is also present in that OTHER OS when coming out of fullscreen 3D applications with all ATi Mobility Radeon users...

Partial distortion

It is usually accompanied with the following error messages:

No matching visual for __GLcontextMode with visual class = 4 (32770), nplanes = 24
No matching visual for __GLcontextMode with visual class = 4 (32770), nplanes = 24
NOTE: __GLcontextMode is due to Color Depth, setting it to 24 solves the problem with these messages.

Partial distortion of the screen has also been reported. This seems to be an issue when using 16 bit color depth. A workaround is to use 24 bit color depth.

Increasing VideoRam from 16384 to

VideoRam  32768

allowed my to set my i810 to 24bit color depth.

Mouse Distortion

If your mouse leaves a distorted trail check your xorg.conf for:

Option     "HWCursor" "false"

Delete the line or outcomment it.

If you have added this line in response to xorg bug #3009 [6], the problem reported in the bug does not exist when running XGL. In fact, leaving this line in xorg.conf may cause the mouse cursor to disappear periodically under XGL.

Support for non power of two textures missing

compiz outputs this:

compiz: Support for non power of two textures missing

The driver has to support either the GL_{NV,EXT,ARB}_texture_rectangle or the GL_ARB_texture_non_power_of_two extension. The open source ATi driver, x11-drivers/xf86-video-ati, does not support these extensions. If you are using an ATi card, try using the proprietary fglrx drivers, x11-drivers/ati-drivers. This info may only relate to r300 and newer cards.

For nvidia users affected by this, the problem appears to have been fixed in a recent update to glitz:

2006-02-16 David Reveman <davidr@novell.com>

* src/glitz_texture.c: Fix so that GL_ARB_texture_rectangle and GL_ARB_texture_border_clamp are not required for texture object creation.

Updating glitz to 0.5.3 should be sufficient to remedy this problem.

To check if the driver supports the needed extension, use the following command:

glxinfo | grep 'GL_\(NV\|EXT\|ARB\)_texture_rectangle\|GL_ARB_texture_non_power_of_two'
I only get this if I run Xgl from a console, on tty8.  If I run it nested in an X server
it works fine.  Any ideas please?
I had the same problem, i believe the reason is that Xgl uses MESA opengl instead of ATI for some reason, 
in console this works for me: LD_LIBRARY_PATH=/usr/lib/opengl/ati/lib Xgl -ac -accel glx:pbuffer -accel xv:pbuffer

I915_PARAM_LAST_DISPATCH not declared in kernel DRM headers

Mesa CVS requires some DRM headers "i915_drm.h" from linux kernel RC 2.6.16. Either upgrade your kernel to this experimental version, or find a copy of that header file (I hear the coffee overlay ebuilds come with it). If you upgraded kernels, Copy it into your kernel header include dir like this:

cp /usr/src/linux/drivers/char/drm/i915_drm.h /usr/include/drm/

It is confirmed that Xorg 7.0 DRM works with i915GM on 2.6.16-rc3-nitro1 on an IBM X41.

Compiz is loaded but all my special effects are missing

Start up the gconf-editor and go to apps/compiz/general/allscreens/options -> active_plugins.

Make sure your desired plugins are in that list - if not, add them following these instructions.

Wrong keyboard layout

The keyboard layout seems to have problems if it's not us. To fix this

setxkbmap -model pc105 -layout dk -variant basic 

Change "dk" in the above line to your own keyboard layout country code if it's not "dk".

Or, to switch between layouts:

 setxkbmap -model pc105  -layout "us,ru(winkeys)" -variant winkeys -option "grp:alt_shift_toggle,grp_led:scroll"

If you need to swap the Caps Lock and Ctrl keys add to the above: -option ctrl:swapcaps

Check that you haven't simply cut and pasted the /usr/bin/compizrc script from the HOWTO XGL without changing the base layout from 'it' (Italy?) to whatever you need!. (this one caught me!)

For grabbing the cube you also have something launch in the root window. The best way to achieve that is to launch nautilus. If you run KDE, the process that runs in the root window is "kdesktop"


CAN'T VT SWITCH ?

I found after many hours of googling that the xkb database can get burned during the xorg 7.x upgrade.. to solve this:

emerge -Cva xkbdata && rm -Rf /usr/share/X11/xkb/ && emerge xkeyboard-config

~mordjah --67.185.152.175 04:07, 8 March 2006 (GMT) hopefully google will index this edit nicely ;)

KDE Application Tray Icon Problems

This can be solved with this patch for kdelibs. The KDE Panel then works as expected.

Using: CoffeeBuzz's overlay
echo "kde-base/kdelibs pertty" >> /etc/portage/package.use
echo "kde-base/kdelibs ~x86" >> /etc/portage/package.keywords
emerge kdelibs

In some cases this doesn't work, a solution is to remove all task bars, remove ktaskbarrc in your .kde :

Code:

rm -f .kde/share/config/ktaskbarrc

After that restart kde and add a new taskbar applet to your panel.

KDE kicker remains empty

If the kicker remains empty although applications were started, try the following:

  • Start KDE withough Xgl.
  • In the properties of the taskbar, ensure that buttons from all desktops will be shown in the taskbar (this option appears only if Xgl isn't running!).
  • Restart KDE, this time with Xgl enabled.

MPlayer crashes Xgl

If your MPlayer crashes Xgl (pops you back to the textmode console), try using 'mplayer -vo x11', worked for me. Note: direct rendering is enabled and working well!

Alternately, replace x11 with gl2.

If either x11 or gl2 work, then put the following line in your .mplayer/config

vo=gl2,x11,xv

BadLength (poly request too large or internal Xlib length error) with radeon/fglrx

Running Xgl on fglrx (radeon 9600) I get:

X Error of failed request: BadLength (poly request too large or internal Xlib length error)
 Major opcode of failed request: 161 (GLX)
 Minor opcode of failed request: 1 (X_GLXRender)
 Serial number of failed request: 94
 Current serial number in output stream: 95

When running

LD_PRELOAD=/usr/lib/opengl/ati/lib/libGL.so.1.2 Xgl :1

Xgl starts and seems to work, but when running opengl-apps (as glxgears or glxinfo) inside it crashes. Note: I use a fresh install of gentoo.

I got it working this way:

# eselect opengl set xorg-x11
# emerge -av glproto mesa
# eselect opengl set ati

--212.242.54.49 16:39, 5 March 2006 (GMT)

Make certain you are using mesa 6.5.0 alpha. I was still getting the mesa-6.4.2r1 because the script didn't unmask the package that I used from coffee's overlay. Davebailey 15:16, 9 March 2006 (GMT)

I got it working this way:

eselect opengl set ati
cd /usr/include/GL
rm gl.h glext.h glx.h glxext.h glxmd.h glxproto.h glxtokens.h
ln -s /usr/lib64/opengl/(ati|nvidia)/include/* ./
emerge -av --oneshot xgl
eselect opengl set ati

after noticing that the culprit was /usr/lib/xorg/modules/xgl/libxglx.so, which is naughty and links directly to /usr/lib/opengl/(xorg-x11|ati|nvidia)/lib/libGL.so.1, depending on which OpenGL profile is set when you compile XGL. This was on an AMD64 machine; substitute 64 for 32 as necessary, and fill in your desired vendor. Xgl runs without needing LD_PRELOAD after this.

BadLength (poly request too large or internal Xlib length error) with nvidia mx440 / FX 5200 / FX 5700

X Error of failed request: BadLength (poly request too large or internal Xlib length error)
Major opcode of failed request: 143 (GLX)
Minor opcode of failed request: 1 (X_GLXRender)
Serial number of failed request: 90
Current serial number in output stream: 91

LD_PRELOAD=/usr/lib/opengl/nvidia/lib/libGL.so Xgl :1 -ac -accel xv -accel glx:pbuffer &

note: I get distorted screen after running this command with GFFX5600

BadLength (poly request too large or internal Xlib length error) with i915

A potential solution is to install a STABLE version of mesa and copy the driver (i915_drm.so) from:

/usr/lib/xorg/modules/dri/

to a temporary directory. Then install the cvs version and copy the old driver back.


Another potential solution that has been tested and proven to work is to install the latest xorg-server from the xgl-coffee SVN repository. This was tested when the previous possible solution failed to be of assistance in the resolution to this error. So, if that didn't work for you, please try this.

Getting Xgl working with i915 mesa driver

A potential solution is to compile STABLE mesa (6.4.2), copy the driver from /usr/lib/xorg/modules/dri to a temporary directory, emerge the CVS mesa and copy the driver back.

This is not possible, driver from 6.4.2 will not work with CVS, there are some request structures that have changed. For me setting VideoRam to 65536 and DefaultColorDepth to 24 in xorg.conf solved all problems. Using mesa-cvs This worked for me too - I had corrupt windows and other such problems and this (unpredictably) fixed it all

mesa and ATI

If compiling mesa with VIDEO_CARDS="ati" fails just leave it blank or set it to VIDEO_CARDS="vesa". There seems to be a bug in cvs mesa. You will still get acceleration by using fglrx.

Do note that "ati" will build several drivers -- (mach64 r128 radeon r200 r300) for my Radeon 9000 (R250 lf M9) it was failing on r300 so i reduced it to just radeon and r200. You can either pass the DRI_DIRS along with the command, modify the linux-dri-x86 config file at the right moment, or modify the ebuild. (xgl-coffee)

If the problem is about unions and scratch, see this threat http://forums.gentoo.org/viewtopic-t-436358.html .....just need to get a CVS of libdrm

x11-libs/libdrm-9999.ebuild for those people, that actually want libdrm from CVS. In fact, installing libdrm is kind of mandatory to get working mesa, otherwise you are just lucky..

# $Header: $
ECVS_SERVER="dri.freedesktop.org:/cvs/dri"
ECVS_MODULE="drm"
ECVS_USER="anonymous"
S=${WORKDIR}/${ECVS_MODULE}
inherit cvs
DESCRIPTION="X.Org libdrm library"
HOMEPAGE="http://dri.freedesktop.org/"
KEYWORDS="~x86 ~amd64"
RDEPEND=""
DEPEND="${RDEPEND}"
SLOT="0"
src_compile() {
        ./autogen.sh --prefix=/usr || die "./autogen.sh failed."
        make || die "make failed"
}
src_install() {
        make DESTDIR=${D} install || die "make install failed."
}

The bug seems to be due to a broken r300 driver. I managed to compile mesa by modifying the ebuild.

cd /usr/local/portage/portage-xgl/media-libs/mesa
modify mesa-*.ebuild.
 if use video_cards_ati; then
       add_drivers mach64 r128 radeon r200 r300
 fi
ebuild mesa-*.ebuild digest

BuddhikaS 3:15, 13 March 2006 (GMT)

White screen bug on loading compiz

With nvidia-cards: If you get a white screen on loading compiz, try to start Xgl without loading compiz, switch to a virtual console and start compiz or alternatively redirect the compiz' output to a file (with compiz > compiz.log 2>&1) If you should see something like "couldn't redirect window texture...", then look at #Black_Windows

This fixed it for me:

eselect opengl set xorg-x11;
emerge compiz;
eselect opengl set nvidia;

An other effect is that windows and the background turn white after they been overlapped by an other window or after a short time. At my place it could be fixed to compile mesa using the motif-USE-flag:

Add the following line to /etc/portage/package.use

media-libs/mesa motif

and reemerge mesa using:

emerge --ask --verbose --oneshot mesa

White screen bug on loading xgl

try emerge -1 =x11-proto/glproto-1.4.7 and emerge xgl

Also: If using an ati mobility card. Disable the Sideport+UMA in the BIOS. This fixed it for me and it is no longer needed for DRI. I specifically have an ati radeon xpress 200m. Questions or comments about t his edit send to jtryonksu_@_gmail_dot_com


White screen bug on loading beryl

I encountered a white screen at the time beryl is stated. However the desktop is fully functional (e.g. Rotating "the white cube", starting programs etc.).

Using

beryl-xgl --use-copy

solved that Problem for me at Macbook Pro 1st gen., (ati x1600, fglrx 8.33, kernel 2.6.20.2, debian). Written by Sammy

Compiling glitz fails at glitz_glx_drawable.c

Emerge mesa before emerging glitz.

TV programs display corrupted images

If your tvtime, xawtv, etc look like this, try the program xdtv and run it with:

xdtv -noxv

or add following line to your .tvtime/tvtime.xml:

<option name="OverScan" value="0.0"/>

if tvtime is still very slow, also add this line to .tvtime/tvtime.xml:

<option name="InputWidth" value="360"/>

you can also launch these programs in another X like this:

xinit tvtime -display :1 -- /usr/bin/X :1 -ac

another option will be to use that great program called mplayer, for example:

mplayer -tv  driver=v4l2 dev=/dev/v4l/video0  tv://54

Shift backspace

By default, shift backspace restarts Xgl, which can lead to some very irritating restarts. To solve this enter the following after Compiz is started (where <keyboard_code> is the two letter identification for your keyboard layout):

xmodmap /usr/share/xmodmap/xmodmap.<keyboard_code>

For US keyboard users, the code is 'us'. Making the command:

xmodmap /usr/share/xmodmap/xmodmap.us

Also for US keyboard users, if this breaks certain keystroke commands (such as F12 or ctrl-alt-left), try this instead:

xmodmap /usr/share/xmodmap/xmodmap.us-101
  • Otherwise, without having the /usr/share/xmodmap directory, the folowing line should do the trick :
xmodmap -e "keycode 22 = BackSpace BackSpace" 

If this works, add this to your ~/.Xmodmap in order to make it permanent

keycode 22 = BackSpace BackSpace

no screens found

On my Radeon Mobility 9000 (R250 lf) M9 running radeon drivers (mesa/drm cvs) I had to enable large texture support (r200) with driconf. Doing this and/or running at a lower resolution may solve this problem.

A modified (or newly created) /etc/drirc could looks like this:

<driconf>
    <device screen="0" driver="r200">
        <application name="all">
            <option name="allow_large_textures" value="2" />
        </application>
    </device>
</driconf>

For an intel 915GM increasing VideoRam to 64megabytes (65536) in your xorg.conf also seems to work.

mesa compile fails: glxcmds.c: In function `glXBindTexImageEXT':

Emerge =app-emulation/emul-linux-x86-xlibs-7.0-r1, then

eselect opengl set ati

(for details please see http://forums.gentoo.org/viewtopic-p-3151471.html#3151471 and http://bugs.gentoo.org/show_bug.cgi?id=109922)


Screen distortion when launching applications from the gnome-panel

If you are experiencing difficulties with screen distortions when you are launching programs from the gnome-panel (ie. not from the menus or the desktop but from the launcher "buttons" on the panel) there is a workaround that may work for you.

The problem is due to the way Xgl (mis)handles the "zoom-border" function that is part of the gnome-panel. (Normally when you click on a launcher, a rectangular "border" will emanate quickly from the launcer to indicate that the click was registered). The workaround is to simply disable the animations for the gnome-panel.

  1. Run the gconf-editor (Applications->System Tools->Configuration Editor)
  2. Navigate into the panel options (apps->panel->global)
  3. Uncheck the "enable_animations" property
  4. Exit the configuration editor


As of gnome-panel-2.14.0, the "zoom-border" functionality is the only thing that is effected by the "enable_animations" property in apps->panel->global. This might change with future releases of gnome-panel.

Totem/xine/gxine video players crashing and/or hard-locking the system

If you are experiencing crashes/hard-locks when using Totem or Xine for video playback, it might be because they are trying to use the "xv" output video plugin. Try the following to correct this.


1. Rebuild totem and xine-lib without xv support

USE=-xv emerge -av xine-lib totem

OR add the following two lines to /etc/portage/package.use and re-emerge

echo "media-libs/xine-lib -xv" >> /etc/portage/package.use
echo "media-video/totem -xv" >> /etc/portage/package.use
emerge -av xine-lib totem


2. Change the default video output plugin for gnome, if applicable

  • Open the "Multimedia Systems Selector" tool (Desktop->Preferences->Multimedia Systems Selector)
  • Select the "Video" tab
  • Under the "Output Video Plugin" section, change the "Output" drop-down to "XWindows (No Xv)"
  • Click Test, click OK when finished
  • Click "Close"


Rebuilding without xv support may solve this problem for other video players as well.

Nvidia GeForceFX 5200 Go (Issues)

Overclocking the GPU causes screen to distort (much like it would normally, but now it doesn't clean up, and instead locks the system)

Quake4 dies with this

   ...using GL_ARB_multitexture
   ...using GL_ARB_texture_env_combine
   ...using GL_ARB_texture_cube_map
   ...using GL_ARB_texture_env_dot3
   ...using GL_ARB_texture_env_add
   X..GL_ARB_texture_non_power_of_two not found
   ...using GL_NV_blend_square
   X..GL_ARB_texture_compression not found
   signal caught: Segmentation fault

Enemy Territory (and TCE) will not fullscreen. (affect also GF 6200 TC pci-e)

Darwinia crashes system if you try to switch from fullscreen to windowed.

Everything else works beautifully. Used XGL-Coffee overlay and script.

If anyone has any idea's, i think it would be prudent to list them here, or link to a card specific wiki, and/or mail me at blistovmhz at gmail dot com


On my Radeon Mobility 9000

ATI PCI-Express x300/x700/x600/x1400/x??? random lockups

-Add these lines to the device section of your xorg.conf

   Option       "UseInternalAGPGART" "no" #this already exists, so just update it if you have "yes here", make sure to have agpgart compiled on your kernel.
   Option       "KernelModuleParm" "agplock=0"

The Ubuntu Wiki has this config section for ATI cards, which seems to work for (at least) the R200M and the Mobility X300:

Section "Device"
   Identifier  "card0"
   Driver      "fglrx"
   Option       "no_accel" "no"
   Option       "no_dri" "no"
   Option       "DynamicClocks" "on"
   Option       "mtrr" "on"
   Option       "DesktopSetup" "Single"
   Option       "ScreenOverlap" "0"
   Option       "Capabilities" "0x00000000"
   Option       "CapabilitiesEx" "0x00000000"
   Option       "VideoOverlay" "on"
   Option       "OpenGLOverlay" "off"
   Option       "CenterMode" "off"
   Option       "PseudoColorVisuals" "off"
   Option       "Stereo" "off"
   Option       "StereoSyncEnable" "1"
   Option       "FSAAEnable" "no"
   Option       "FSAAScale" "1"
   Option       "FSAADisableGamma" "no"
   Option       "FSAACustomizeMSPos" "no"
   Option       "FSAAMSPosX0" "0.000000"
   Option       "FSAAMSPosY0" "0.000000"
   Option       "FSAAMSPosX1" "0.000000"
   Option       "FSAAMSPosY1" "0.000000"
   Option       "FSAAMSPosX2" "0.000000"
   Option       "FSAAMSPosY2" "0.000000"
   Option       "FSAAMSPosX3" "0.000000"
   Option       "FSAAMSPosY3" "0.000000"
   Option       "FSAAMSPosX4" "0.000000"
   Option       "FSAAMSPosY4" "0.000000"
   Option       "FSAAMSPosX5" "0.000000"
   Option       "FSAAMSPosY5" "0.000000"
   Option       "UseFastTLS" "0"
   Option       "BlockSignalsOnLock" "on"
   Option       "UseInternalAGPGART" "no"
   Option       "ForceGenericCPU" "no"
   Option       "KernelModuleParm" "agplock=0"
   Option       "PowerState" "1"
   BusID       "PCI:1:0:0"
EndSection

If you still experience lock-ups try disabling mtrr. Check your dmesg output - if you see an error regarding mtrr, (i.e. mtrr allocation failed) you will need to do this.

For more information on MTRR errors and the fglrx driver, please see this link: [7]

ATI MOBILITY RADEON 9700 (ISSUES) AND USING KDE ENVIRONMENT

I've changed initxgl.sh a bit to make it work pretty good with my kde:

Code: initxgl.sh for KDE users
#!/bin/bash
# xgl startup script by RiPz @ www.binary-pulse.co.uk
# modified by Rafael Fernández López @ www.maestroprogramador.com
# info@maestroprogramador.com
# GNU Public Licence
# woo!
# Edit: 20060301 Added xfce-svn CF (CoffeeBuzz www.tripthelight.net/xgloverlay)

echo "###### STARTXGL SCRIPT ######"

if [ "$1" = "" ]; then
    echo "Usage: startxgl <display number> <window manager> <gfx>"
    echo "<gfx> is optional, values 'ati' or 'nvidia/other', default is nvidia/other"
    echo "Window Managers currently supported: xfce, gnome, kde."
elif [ "$2" = "" ]; then
    echo "Usage: startxgl <display number> <window manager> <gfx>"
    echo "<gfx> is optional, values 'ati' or 'nvidia/other', default is nvidia/other"
    echo "Window Managers currently supported: xfce, gnome, kde."
else
    echo "Starting X Server with XGL"
    echo "   Using Display $1"
    echo "   Using WM: $2"

    if [ "$3" = "ati" ]; then
        echo "   Using Card: ati"
        Xgl :$1 -ac -accel xv:pbuffer -accel glx:pbuffer &
    else
        echo "   Using Card: nvidia/other"
        Xgl :$1 -ac -accel xv -accel glx:pbuffer &
    fi

    sleep 2
    echo "   Starting compiz"
    #tambien tenia decoration despues de --replace
    DISPLAY=:$1 LD_LIBRARY_PATH=/usr/lib/opengl/xorg-x11/lib/ compiz --replace gconf decoration wobbly fade minimize cube rotate zoom scale move resize place switcher &
    sleep 2

    echo "   Starting Window Manager"

    if [ "$2" = "gnome" ]; then
        DISPLAY=:$1 gnome-window-decorator &
        DISPLAY=:$1 /usr/libexec/gnome-settings-daemon &
        DISPLAY=:$1 nautilus -n --sync &
        DISPLAY=:$1 gnome-panel
    elif [ "$2" = "xfce" ]; then
        DISPLAY=:$1 xfce-mcs-manager
        DISPLAY=:$1 gnome-window-decorator &
        DISPLAY=:$1 xftaskbar4 &
        DISPLAY=:$1 xfdesktop &
        DISPLAY=:$1 exec xfce4-panel
    elif [ "$2" = "xfce-svn" ]; then
        DISPLAY=:$1 xfce-mcs-manager
        DISPLAY=:$1 gnome-window-decorator &
        DISPLAY=:$1 xfdesktop &
        DISPLAY=:$1 exec xfce4-panel
    elif [ "$2" = "kde" ]; then
	DISPLAY=:$1 gnome-window-decorator &
	DISPLAY=:$1 startkde
    fi
    
    echo "   DONE!"
fi

When my system boots up (it starts X as normal, NOT XGL), I go to the terminal through Ctrl+Alt+F1. Then what I usually do is a /etc/init.d/xdm stop. My next step is to run "sh initxgl.sh 1 kde ati" as a normal user.

If you do use nvidia you just has to change ati with nvidia entry.


nvidia-settings complains about NV-CONTROL extension

According to information on the Gentoo and Ubuntu forums (google search for "Xgl nvidia"), the current 8178 Nvidia drivers won't allow you to use direct rendering.

If you have an Nvidia card you can check to see if you have direct rendering by running

   glxgears

or

   glxinfo | grep direct

or

   nvidia-settings

where running nvidia-settings should complain about NV-CONTROL extensions not implemented on your current display.

According to same forums, support for direct rendering should come soon in the next Nvidia drivers, but if you're impatient like me you can run two X servers simultaneously; one would have direct rendering enabled (and thus no fancy-pants 3D effects) and the other would be a server running the good ol' Xgl shiznazz that this excellent wiki addresses. (Using Xnest for this probably won't work.)

Some caveats 1) Start up your riced-out 3D server before you start the plain-Jane X server. If you don't, then the window manager for your regular X setup will prevent gnome-window-decorator from skinning your windows, and you won't get any cool effects. You could start up your other X server first, but just don't start up a window manager at the same time. For example,

   X :1     # plain-Jane server starts first
   startxgl # Let gnome-window-decorator start first
   blackbox -display :1 & # Now run your window-manager of choice.

2) Running nvidia-settings for display 0 will cause it to crash. To be safe, if you ever need to run nvidia-settings, always run it like so

   nvidia-settings -c :1


Running an Xgl-enabled server next to a normal X server You could run Xgl, compiz, and xfce4 on display 0 (ctrl-alt-F7) and your old configuration on display 1. The instructions that follow definitely could be improved upon, but that's why this is a wiki. ;)

You could get /usr/bin/startx to run on display 1 by editing the following lines:

from

   serverargs=""

to

   serverargs=":1 vt8"

and from

   authdisplay=${display:-:0}

to

   authdisplay=${display:-:1}

Then you could also start up compiz, etc, like so on the 0 display:

   #!/bin/bash
   Xgl :0 -fullscreen -ac -accel xv:fbo -accel glx:pbuffer &
   DISPLAY=:0 gnome-window-decorator &
   LD_LIBRARY_PATH=/usr/lib/opengl/xorg-x11/lib/ DISPLAY=:0 compiz --replace gconf &
   DISPLAY=:0 xfce4-session

and switch between them using ctrl-alt-F7 and F8.

Eric DB Wendel 71.134.181.209 17:31, 2 April 2006 (GMT)

X-Chat Won't Start

X-Chat refuses to start if you have transparency enabled. To fix this run gconf-editor, go-to apps > xchat > main_window Change the background_type key from 2 to 1.

Black screen freeze on X-startup with nvidia mobile chips

This could be caused by a bug in the never versions of the nvidia drivers (see http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=343186 )

This bug doesn't occur when you load the nvidia driver with the NVreg_Mobile option enabled. To enable this under gentoo edit the file "/etc/modprobe.d/nvidia" and add (or uncomment) the line:

options nvidia NVreg_Mobile=1

after that you should unload the module (if you have module unload support disabled in your kernel skip this and restart your system after the last step) :

rmmod nvidia

and then regenerate the modules.conf file by typing:

update-modules

Driver does not provide FireGL X11 extensions. Panel component will operate only partially

I get the error message:

"Driver does not provide FireGL X11 extensions. Panel component will operate only partially."

When I try and start the ATi Control panel from within Xgl.

I get no window decorations (nor any other effects), but I have all the options in gconf-editor. I'm using the script from here: http://www.binary-pulse.co.uk/random/startxgl Any ideas please?

error: `GLX_SWAP_METHOD_OML' undeclared while compiling Xorg or Xgl

If you get messages like...

 glxcmds.c: In function `DoGetFBConfigs':
 glxcmds.c:1105: error: `GLX_SWAP_METHOD_OML' undeclared (first use in this function)
 glxcmds.c:1105: error: (Each undeclared identifier is reported only once
 glxcmds.c:1105: error: for each function it appears in.)
 make[2]: *** [glxcmds.lo] Error 1
 make[2]: *** Waiting for unfinished jobs....
  i686-pc-linux-gnu-gcc -DHAVE_CONFIG_H -I. -I. -I../../include -I../../include -I../../include -I../../include -I../../include -I../../include -I../../GL/include -I../../hw/xfree86/os-support -DHAVE_DIX_CONFIG_H -I/var/tmp/portage/xorg-server-1.0.2-r3/work/Mesa-6.4.2/include -DXFree86Server -DIN_MODULE -DXFree86Module -DXFree86LOADER -Wall -Wpointer-arith -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations -Wnested-externs -fno-strict-aliasing -D_BSD_SOURCE -DHAS_FCHOWN -DHAS_STICKY_DIR_BIT -I../../include -I../../include -I../../Xext -I../../composite -I../../damageext -I../../xfixes -I../../Xi -I../../mi -I../../miext/shadow -I../../miext/damage -I../../render -I../../randr -I../../fb -I../../lbx -Os -march=pentium4 -pipe -MT glxcmdsswap.lo -MD -MP -MF .deps/glxcmdsswap.Tpo -c glxcmdsswap.c  -fPIC -DPIC -o .libs/glxcmdsswap.o
 make[2]: Leaving directory `/var/tmp/portage/xorg-server-1.0.2-r3/work/xorg-server-1.0.2/GL/glx'
 make[1]: *** [all-recursive] Error 1
 make[1]: Leaving directory `/var/tmp/portage/xorg-server-1.0.2-r3/work/xorg-server-1.0.2/GL'
 make: *** [all-recursive] Error 1

... when compiling xorg-server and/or xgl it may help to switch the opengl implementation from your video card vendor's to the x11/mesa implementation:

 eselect opengl set xorg-x11

After the successful installation of the problematic package you have to switch back:

 eselect opengl set ati # for fglrx (ati proprietary driver)
 eselect opengl set nvidia # for nvidia proprietary driver

-- hoffie, 84.148.213.21, (solved by some #gentoo-xgl users)

Unresolved symbol __glXActiveScreens

Add:

 Load "glx"

to your xorg.conf in Section "Module"


No GLXFBConfig for default depth

Got this error during latest installation. Does anybody know anything about it?

 mypc ~ # LD_LIBRARY_PATH=/usr/lib/opengl/xorg-x11/lib/ DISPLAY=:1 compiz gconf
 compiz: No GLXFBConfig for default depth, this isn't going to work.
 compiz: Failed to manage screen: 0
 compiz: No managable screens found on display :1

Got the same, tried to make it work but no effect. Seems to be connected with the last xorg-x11 update.

I too received this error. Any ideas on the cause/solution? --Azdruid 01:32, 20 May 2006 (UTC)

The same situation: xgl starts but compiz gives errors like above.

POSSIBLE FIX: I had this error with compiz-vanilla from 20060418, I downgraded to compiz from 20060404 and compiz starts up now.

POSSIBLE FIX: I had this error with compiz-*, I've upgraded to compiz-quinnstorm-0.0.10.14

The compiz BTS [8] says it's a bug in mesa. I upgraded to mesa cvs (20060529), but the error remains. -- I now built the whole system from scratch using xgl-coffee's rev 255, and it works like a charm.

I have tried 20060404 and vanilla 0408 ,0410 ,0416 ,and quinnstorm 0.0.9.[1-4] The problem appears in all the version after 20060408 and quinnstorm 0.0.9.0. Maybe someone imported a bug from 0408 to 0410 ? But 20060408 and quinnstorm 0.0.9.0 can only start up compiz and then give me some white windows and messages like this:

compiz: Couldn't bind redirected window 0x400032 to texture

POSSIBLE FIX: re-emerging mesa gives the following output:

* Rebuild x11-base/xorg-server without USE=nptl
* or AIGLX (compiz, etc.) will not work.
* This is because of a bug in the Mesa NPTL assembly code
* in all Mesa 7.0.x versions (Mesa 6.x is OK).

ctrl key not working

If your ctrl key isnt functioning correctly then try changing the hotkey in the water plugin in the gconf from "ctrl" to "none" for some reason it seems to interfere, even when the water plugin is switched off

If KDM is unusably slow on startup

Note: As an alternative you can switch to a different style (kwin) for the login manager (e.g. CDE works for me). Edit /usr/kde/3.5/share/config/kdm/kdmrc and set "GUIStyle" to CDE within the [X-*-Greeter] section.

Using kdm does not always work reliably, at least for me kdm was unusable. Therefore I used gdm for logging in and replaced kwin with compiz+gnome-window-decorator. Here is how to do just that:

Save the following script to /usr/bin (or anywhere else, but be consistent below):

File: /usr/bin/compizrc
#!/bin/bash
DISPLAY=:1 LD_LIBRARY_PATH=/usr/lib/opengl/xorg-x11/lib/ compiz --replace gconf > ~/compiz.log 2>&1 &
sleep 1;
DISPLAY=:1 gnome-window-decorator > ~/gnome-window.log 2>&1 &

Don't forget to make it executable! The above script will replace kwin as your KDE window manager. To instruct KDE which window manager to start you have to specify the executable in the environment variable KDEWM. In order to make this change global, edit /etc/env.d/99kde-env and add the following line:

File: added to /etc/env.d/99kde-env
KDEWM=/usr/bin/compizrc

Now you need to tell gentoo to use gdm instead of kdm. For this purpose edit /etc/rc.conf and change DISPLAYMANAGER to the following line:

File: /etc/conf.d/xdm
DISPLAYMANAGER="gdm"

That's it for the KDE part. For configuring gdm read section Gdm and gnome session. You have to leave out section 2 as you want to run kde, don't you? After having gdm setup to use Xgl instead of X, logout and execute the following as root at the console login:

Code: Update your environment

env-update && source /etc/profile && /etc/init.d/xdm restart

After that gdm should show up and you should be able to log into a KDE session. You can check ~/compiz.log and ~/gnome-window.log for errors if anything goes wrong.

A note on CFLAGS

AMD64 users should compile compiz with CFLAGS="-march=k8 -O0" as it fixes crashes with the rotate and wobbly extensions. If you experience crashes on any platform, the first thing you should try is recompiling compiz with -O0, as compiz does not seem to optimize very stably.

The newest compiz ebuilds from Hanno and CoffeeBuzz should correctly filter CFLAGS on amd64

Compiling mesa breaks at [...] r400 [...]

Make sure you have set your VIDEO_CARDS-option correctly in make.conf Example:

VIDEO_CARDS="nvidia"

Compiz does a segmentation fault while editing gconf

This has happened to me quite some times now. I have also heard about it happening on other boxes besides mine. Both running compiz with KDE. The symptoms is that the window decorations suddenly are gone, you cannot resize or move windows anymore.

It happens while using gconf-editor to edit compiz settings and it can happen while editing any part of the configuration. Changing the value back has no effect.

The only solution to this problem I have found is to run the following as root:

# gconftool-2 --recursive-unset /apps/compiz/
# GCONF_CONFIG_SOURCE="xml:merged:/etc/gconf/gconf.xml.defaults" gconftool-2 --makefile-install-rule /etc/gconf/schemas/compiz.schemas

Afterwards run compiz again and it should be back to normal only that all the settings are gone back to default. You can safely backup the abovementioned dir before you delete it, and then reinstate it, but remember to kill off any running instances of gconfd-2, since it will override any changes you do outside of gconf-editor.

KDE right click menu not working

This happened to me on KDE-3.5.4, but i believe it is the same for older versions. I was unable to get the right click on desktop KDE menu. The water plugin wouldn't work on the desktop either. After some searching the problem turned out to be that I turned off the icons on the desktop which also turned off programs on the desktop. The settings are in the KDE Control Panel Desktop/Behavior. I turned on "Show Icons on Desktop" which allowed me to see the other setting "Allow programs in desktop window". This allowed the right click KDE menu to work on the desktop and the water plugin now works on the desktop not just windows.

Flickering on compiz effects after a VT switch

This happened to me as I was running Xgl on a Radeon Mobility 7500 (ati-6.6.1) with "Page Flip" enabled. Make sure you comment out or remove the corresponding line in xorg.conf, as in the following:

Section "Device"
   Identifier  "card0"
   Driver      "ati"
   ...
   #Option "EnablePageFlip" "true" #Either comment out this line, remove it, or put "false"
   ...
EndSection

Windows do not redraw correctly

When I updated fglrx to 8.30.3, surprisingly my windows stopped refreshing correctly. Only an external activity, such as changing focus or changing desktop, the window was able to refresh its content.

After trying different approaches and solutions, like these posts, and playing with FSAAEnable and EnablePrivateBlackZ options, as follows

Section "Device"
   Driver      "fglrx"
   ...
   Option      "EnablePrivateBackZ" "no"
   Option      "FSAAEnable" "no"
   ...
EndSection

I figured out that the best solution was to downgrade to 8.29.6 verion of fglrx. I insisted with solutions until version 8.32.5, but at the end gave up, and didn't notice any performance loss.

I am at 8.37.6-r1, still with the same problem

Mesa errors using intel

I used a version from 05/2006 of the xgl-coffee (aka. portage-xgl) overlay before, without big problems. After switching to the xeffects overlay xgl became damn slow and aiglx lead to several drawing bugs.

I saw several mesa errors on the console. So resolved the problem by downgrading mesa to version =6.5.1* and anything works fine, fast and quite stable now.

Firefox crashes XGL/Beryl on refresh/new tab functions

Independently on fglrx and Firefox versions I had usually this annoying XGL crashes. After a while it seemed obvious that the problem came up when I was reading PDF documents with Acrobat Reader using Beryl zoom function instead of Acroreader's one (because of speed), and later check on the internet some additional information.

Making some tests I figured out that after a single zoom-in and zoom-out, browser crashed XGL every time I was pressing F5 or Ctrl-Tab sequences (sometimes following a link as well). Therefore the problem seemed to be Inputzoom plugin.

Fortunately the workaround exists, and is explained on this post: simply disable "Hide normal cursor" option in the inputzoom plugin.

Slow scrolling in Firefox/Konqueror on Intel

After I emerged compiz Firefox scrolling was very slow, and the commonly suggested 'disable smooth scrolling' workaround had no effect at all.

After some research it appeared that Intel drivers (in my case x11-drivers/xf86-video-i810-2.2.1) needed to be explicitly instructed to use XAA acceleration in xorg.conf (Section "Device"):

        Option      "AccelMethod" "XAA"
        Option      "XAANoOffscreenPixmaps" "true"

As I understand XAA is getting replaced by EXA, but Intel drivers don't implement a usable EXA at the moment.

No GLX Visuals Available

This means that your GLX settings have been messed up. When I got this message, I ran glxinfo and saw that there was no more glx and no more direct rendering. I restored the backups of the configuration files that I modified, than I modified X11/gdm/custom.conf by hand. Rebooted and it worked great!


Last modified: Sat, 13 Sep 2008 15:38:00 +1000 Hits: 91,773

Created by NickStallman.net, Luxury Homes Australia
Real estate agents should list their apartments, townhouses and units in Australia.