installation of nvidia drivers
Fabio Massimo Di Nitto
fabbione at fabbione.net
Thu Sep 2 10:31:42 CDT 2004
On Thu, 2 Sep 2004, Mark Shuttleworth wrote:
> We already have some binary-only kernel modules, so these can simply be
> built with those. ATI, Nvidia should definitely be built and be
> available for any kernel we ship. And it should be trivial to use them
> for those who want to.
That's why we need to investigate how to build them properly. The nvidia
driver for exaple builds only on top of the running kernel. This means
that we probably cannot build it on our buildd, but we must do it manually
on our warty machines.
> >Not the autodetection part, but once the drivers are installed a simple
> >dpkg-reconfigure xserver-xfree86 will show the nvidia driver as available
> >and the user can select them without any problem.
> >
> >
> Why not autodetect too?
Several reasons:
Autodetect doesn't use the drivers available on the filesystem but the
hardcoded database in discover1.
Discover1 has no knowledge of nvidia, but only nv (the free driver) and we
can choose either one or the other. There is no way to fallback here.
In case of missing autodetection, we cannot propose nvidia if it is not
going to be installed, otherwise X will not start and non-expert users
will have problems to understand why.
Once everything is installed xserver scripts are clever enough to build
the list of available drivers from what is available on the disk and
present them to the user.
Fabio
--
<user> fajita: step one
<fajita> Whatever the problem, step one is always to look in the error log.
<user> fajita: step two
<fajita> When in danger or in doubt, step two is to scream and shout.
More information about the sounder
mailing list