On Tue, 12 Nov 2002, Robert P. J. Day wrote:
> (i have *got* to put this on a web page somewhere ...)
I agree with Jonathan, this is very well-written. Just a couple of
corrections (for when it goes on the Web page).
>[...]
> NVIDIA_GLX-1.0-3123.i386.rpm # the binary GLX driver file
> NVIDIA_kernel-1.0-3123.src.rpm # the source kernel driver file
>
> First, make sure that these RPMs have matching version numbers;
> bad things will likely happen if they don't.
>
> And second, while you can grab the prebuilt binary of the GLX
> package, you're *strongly* encouraged to download (and compile)
> the source version of the kernel package. As I read it, you
> should try to use the prebuilt binary kernel RPM only if you're
> running the stock, standard, as-it-was-shipped-from-Red Hat
> kernel. So play it safe -- get the source RPM for the kernel
> package and build it yourself. If you do this, you can't go wrong.
It is true that you can always rebuild yourself, but there are some
NVIDIA-built drivers for update kernels, and you can use these as long as
you have the exactly corresponding update kernel.
Note that you don't seem to be able to have NVIDIA drivers for two
different kernels on the same machine, so if you are booting into a kernel
different from the one you built the driver for, then you need to start in
console mode and fix your XF86Config before starting X.
> * Step 2 -- Building and installing the RPMs
>
> Assuming that you have no earlier versions of NVIDIA RPMs
> cluttering up your machine, first, compile the kernel source
> RPM with something resembling:
>
> # rpmbuild --rebuild NVIDIA_kernel-<whatever>.src.rpm
One advantage to NVIDIA-built kernels is they have some
architecture-specific ones. To build these yourself,
# rpmbuild --rebuild --target <arch> NVIDIA_kernel-<whatever>.src.rpm
where <arch> is i686 or athlon, as appropriate.
> Then install the binary GLX RPM. Once again, you'll get some
> installation warnings, this time regarding conflicting MESA RPMs.
> As before, ignore them.
Note, if you have this installed already remove it and install the new
one, don't upgrade.
> * Step 4 -- Starting X
>
> If you're using any version of Red Hat up to 7.3 (and possibly
> either of the limbo betas, if memory serves), at this point, you
> should just have to type "startx &" to get an X session.
>
> If, however, you're using the latest "null" beta, there is a mismatch
> related to the gcc compilers that were used to compile the NVIDIA drivers
> and what Red Hat shipped with "null". No problem -- just force loading
> of the driver first with:
>
> # insmod -f NVdriver (ignore any warnings)
>
> then try "startx &" again.
>
> You can either force loading of the driver each time you bring up
> your machine this way, or I found adding the line
>
> install NVdriver insmod -f NVdriver
>
> to the file /etc/modules.conf will take care of this for you.
In runlevel 5, this doesn't seem to be an issue. The NVIDIA driver loads
without complaint (other than the kernel taint warning).
> * Step 5 -- Getting fancy
>
> If you're bored, you might want to experiment with a few more
> XF86Config settings that are documented in the aforementioned
> README.txt file at www.nvidia.com. A couple of examples (that
> would be added under the "Device" section for the NVIDIA card:
>
> Option "NoLogo" "1" # don't display logo
> Option "CursorShadow" "1" # have shadowed cursor
Hmm, I'll have to try these.
--
Matthew Saltzman
Clemson University Math Sciences
[EMAIL PROTECTED]
http://www.math.clemson.edu/~mjs
--
redhat-list mailing list
unsubscribe mailto:redhat-list-request@;redhat.com?subject=unsubscribe
https://listman.redhat.com/mailman/listinfo/redhat-list
