Fix for the NVidia problems in the main faq

[vegan_linuxguy]

Concerning the errors with NVidia drivers mentioned in the faq, including the common Xlib DRI error, you suggest running "rpm -e --nodeps XFree86-Mesa-libGL". This is not a wise solution to suggest because it will break RPM, which makes apt/synaptic unusuable. The solution discussed in this link:

http://www.redhat.com/archives/fedora-test-list/2003-October/msg01600.html

is much better, as it doesn't break RPM. A simple downloadable script with the commands in the link may be the best solution to the problem.

[kai_t]

What about using rpm -ihv --justdb XFree86-Mesa-libGL*.rpm after removing the package?

[Spark]

Another problem:
I couldn't compile the game "neverball" because of a missing GL lib ("cannot find -lGL"), the only way I could make it compile was by installing the XFree86-Mesa-libGL rpm. The resulting binary doesn't work anyway, it crashes immediately. Is there something else I could try?

[risc]

Originally Posted by kai_t

What about using rpm -ihv --justdb XFree86-Mesa-libGL*.rpm after removing the package?

Mike Harris is the guy who packages X for Red Hat / Fedora I'd suggest you use his fix.

BTW How long till this new information gets put in the FAQ?

[kai_t]

Originally Posted by risc

Mike Harris is the guy who packages X for Red Hat / Fedora I'd suggest you use his fix.

Sorry to contradict you, but the posting referenced in this thread quotes a workaround suggested by nosp[AT]xades.com.

Actually, Mike wrote in https://listman.redhat.com/archives/.../msg03048.html:

Mesa software GL should be completely uninstalled with --nodeps
prior to installing Nvidia's proprietary drivers. Otherwise you
will get very mixed problems with software trying to use Mesa
libGL instead of Nvidia libGL. Since we ship multiple optimized
versions of libGL now, this will confuse things even greater.

I got the "--justdb" idea from http://www.redhat.com/archives/fedor.../msg02038.html.

[Avatraxiom]

Actually, the currently recommended solution (as the FAQ states) is to use the nVidia driver RPMs -- they work much better and you should never have to uninstall Mesa.

-M

[D@ Mick]

Could anybody tell me what you have to do when you get an error conserning your kernel when you install the NVIDIA driver???he couldn't found one on the nvidia site and could'n compile one on himself. So what to do????

[mea]

1. Down-load the driver from www.nvidia.com.
2. As user root (or su) run '/sbin/init 3' (leave out the single quotes). This will initialize to run level 3 (i.e. no xserver).
3. Log in as root and run "export CC=gcc32". (CC is upper case)
4. Run 'sh NVIDIA-Linux-x86-1.0-4496-pkg2.run' (leave out the single quotes and, if different, substitute the name of the nvidia driver .run file down-loaded in step 1).
5. Follow the on-screen instructions.
6. When the driver installation completes successfully, edit the /etc/X11/XF86Config file by commenting out the line in the "Module" section that reads 'Load "dri"' and by changing the line in the "Device" section that reads 'Driver "nv"' to 'Driver "nvidia"'. (leave out the single quotes)
7. Run '/sbin/init 5' (leave out the single quotes). This will initialize to run level 5 (i.e. the xserver will be started).
8. The nvidia driver should now be operational. Just log in and enjoy!

NOTE: Anytime the kernel has been updated, the x-server will not load after rebooting. You will NOT have to edit the /etc/X11/XF86Config file again. Just repeat steps 2, 3, 4, 5, 7 and 8 to prepare the driver for the new kernel.

[Avatraxiom]

There is also an answer in the normal FAQ for this. :-)

-M