NVidia packages - let's not use nvidia-installer [RFC]

Question

Hi. After messing around with Xgl I (for some reason) decided that I hated the way the nvidia driver package doesn't account for all the files the nvidia-installer spits out/modifies. I wanted all the files involved to be properly managed by pacman. So I made some PKGBUILDs, looking at Fedora and Gentoo stuff for inspiration, and getting annoyed trying to work out what needs to be where based on NV's unmaintained install makefile..:shock:
The PKGBUILDs that follow separate the kernel module from the rest of the nvidia package, but this is by no means necessary - I just wanted to be able to reinstall one without the other (to do with installing gl parts without restarting X - see below for more).
(If there is to be a separation, it may be more logical to put the x.org driver module with the kernel module, with just the GL parts separate.)
pkgname=nvidia-kernel
pkgver=1.0.8178
_pkgbinary=NVIDIA-Linux-x86-1.0-8178
pkgrel=1
pkgdesc="NVidia driver kernel module"
url="http://www.nvidia.com"
depends=('bash' 'gcc' 'binutils' 'glibc' 'make' 'nvidia-glx')
source=(ftp://download.nvidia.com/XFree86/Linux-x86/1.0-8178/$_pkgbinary-pkg0.run nvidia.rc
nvidia.patch nv2.diff)
install="nvidia.install"
#provides=()
conflicts=('nvidia')
build() {
cd $startdir/src/
chmod +x $_pkgbinary-pkg0.run
./$_pkgbinary-pkg0.run --extract-only
cd $_pkgbinary-pkg0
#strip stuff that's in nvidia-glx to make download lighter
rm -rf usr/bin
rm -rf usr/include
rm -rf usr/lib
rm -rf usr/share
rm -rf usr/X11R6
mkdir -p $startdir/pkg/usr/share/nvidia
cp -p LICENSE $startdir/pkg/usr/share/nvidia
# adding patches from nvidia forum and now provided by zander
patch -Np0 -i $startdir/src/nvidia.patch || return 1
patch -Np0 -i $startdir/src/nv2.diff || return 1
#clean src
install -D -m 755 $startdir/src/nvidia.rc $startdir/pkg/etc/rc.d/nvidia
cd $startdir/src
rm *
mkdir -p $startdir/pkg/opt/nvidia
mv * $startdir/pkg/opt/nvidia
md5sums=('6c8081bfde4a806a487efc2a9a1ff016' '08f4f614066c08bd0774c7e557953fbe'
'3b5a2525633e88b9d78c4721190542e6')
(the file nv2.diff patches the kernel code for 2.6.16 kernels..)
Here's the .install
_driver=NVIDIA-Linux-x86-1.0-8178-pkg0/./nvidia-installer
post_install() {
cat << EOF
IMPORTANT
==> In order to use the software, you have to agree to NVIDIA's license located in
==> /usr/share/nvidia/LICENSE
==> If you don't, please remove this package (pacman -R nvidia-kernel)
Installation starts now ...
EOF
cd /opt/nvidia/NVIDIA-Linux-x86-1.0-8178-pkg0/usr/src/nv
make install > /dev/null 2>&1 || (echo "INSTALLATION FAILED!" ; echo "==> You have to shutdown Xserver to finish installation! You have to rerun 'pacman -S nvidia-kernel'" ; echo "==> If you're running a custom kernel, make sure the source tree is available." ; echo "==> Other common solutions can be found on the wiki: http://wiki.archlinux.org/index.php/How_to_install_NVIDIA_driver" )
modprobe nvidia
rm -r /opt/nvidia
cat << EOF
To use this driver you need the nvidia-glx package too.
If you need more information about setting up nvidia drivers have a look at:
"http://wiki.archlinux.org/index.php/How_to_install_NVIDIA_driver"
EOF
post_upgrade() {
rmmod nvidia > /dev/null 2>&1
post_install $1
pre_remove() {
cat << EOF
==> Deinstallation starts now!
EOF
rmmod nvidia
_KERNELNAME=$(uname -r)
rm /lib/modules/$_KERNELNAME/kernel/drivers/video/nvidia.ko > /dev/null 2>&1
cat << EOF
==> Don't forget to update your /etc/X11/XF86Config or /etc/X11/xorg.conf!
You may want to remove nvidia-glx as well.
EOF
op=$1
shift
$op $*
Now for the other bit..
pkgname=nvidia-glx
pkgver=1.0.8178
_pkgbinary=NVIDIA-Linux-x86-1.0-8178
pkgrel=1
pkgdesc="The NVidia X.org driver and utilities (without the kernel module)"
url="http://www.nvidia.com"
depends=('bash' 'gcc' 'binutils' 'glibc' 'make' 'nvidia-kernel')
source=(ftp://download.nvidia.com/XFree86/Linux-x86/1.0-8178/$_pkgbinary-pkg0.run)
install="nvidia-glx.install"
provides=('libgl')
conflicts=('libgl' 'nvidia')
md5sums=('6c8081bfde4a806a487efc2a9a1ff016')
build() {
cd $startdir/src/
chmod +x $_pkgbinary-pkg0.run
./$_pkgbinary-pkg0.run --extract-only
cd $_pkgbinary-pkg0
mkdir -p $startdir/pkg/usr/share/applications
mkdir -p $startdir/pkg/usr/bin
mkdir -p $startdir/pkg/usr/man/man1
mkdir -p $startdir/pkg/usr/lib/xorg/modules/{extensions,drivers}
mkdir -p $startdir/pkg/usr/share/nvidia
#Point .desktop to correct location
chmod +w ./usr/share/applications/nvidia-settings.desktop
sed 's:__UTILS_PATH__:/usr/bin:' ./usr/share/applications/nvidia-settings.desktop > ./nv.desktop
sed 's:__DOCS_PATH__:/usr/share/nvidia:' ./nv.desktop > ./usr/share/applications/nvidia-settings.desktop
chmod -w ./usr/share/applications/nvidia-settings.desktop
#Grab what we need and put it where we want
install -m 755 usr/bin/nvidia-bug-report.sh ${startdir}/pkg/usr/bin
install -m 755 usr/bin/nvidia-settings ${startdir}/pkg/usr/bin
install -m 755 usr/bin/nvidia-xconfig ${startdir}/pkg/usr/bin
install usr/lib/libGL.so.${pkgver} ${startdir}/pkg/usr/lib
install usr/lib/libGLcore.so.${pkgver} ${startdir}/pkg/usr/lib
install usr/lib/libnvidia-tls.so.${pkgver} ${startdir}/pkg/usr/lib
install usr/lib/libnvidia-cfg.so.${pkgver} ${startdir}/pkg/usr/lib
install usr/X11R6/lib/libXvMCNVIDIA.a ${startdir}/pkg/usr/lib
install usr/X11R6/lib/libXvMCNVIDIA.so.${pkgver} ${startdir}/pkg/usr/lib
install usr/X11R6/lib/modules/drivers/nvidia_drv.o ${startdir}/pkg/usr/lib/xorg/modules/drivers
install usr/X11R6/lib/modules/drivers/nvidia_drv.so ${startdir}/pkg/usr/lib/xorg/modules/drivers
install usr/X11R6/lib/modules/extensions/libglx.so.${pkgver} ${startdir}/pkg/usr/lib/xorg/modules/extensions
cp -pr usr/include ${startdir}/pkg/usr/share/nvidia
cp -pr usr/share/doc/* ${startdir}/pkg/usr/share/nvidia
cp -pr usr/share/applications/* ${startdir}/pkg/usr/share/applications
cp -pr usr/share/man/man1 ${startdir}/pkg/usr/man
cd ${startdir}/pkg/usr/lib/
ln -fs libGL.so.${pkgver} libGL.so
ln -fs libGL.so.${pkgver} libGL.so.1
ln -fs libGLcore.so.${pkgver} libGLcore.so.1
ln -fs libnvidia-cfg.so.${pkgver} libnvidia-cfg.so.1
ln -fs libnvidia-cfg.so.${pkgver} libnvidia-cfg.so
ln -fs libnvidia-tls.so.${pkgver} libnvidia-tls.so.1
ln -fs libXvMCNVIDIA.so.${pkgver} libXvMCNVIDIA-dynamic.so.1
cd ${startdir}/pkg/usr/lib/xorg/modules/extensions
ln -fs libglx.so.${pkgver} libglx.so
..and here's the .install
_driver=NVIDIA-Linux-x86-1.0-8178-pkg0/./nvidia-installer
post_install() {
#switch in nvidia GL headers
ln -sf /usr/share/nvidia/include/GL/gl.h /usr/include/GL/gl.h
ln -sf /usr/share/nvidia/include/GL/glext.h /usr/include/GL/glext.h
ln -sf /usr/share/nvidia/include/GL/glx.h /usr/include/GL/glx.h
ln -sf /usr/share/nvidia/include/GL/glxext.h /usr/include/GL/glxext.h
cat << EOF
IMPORTANT
==> In order to use the software, you have to agree to NVIDIA's license located in
==> /usr/share/nvidia/LICENSE (installed with nvidia-kernel - required)
==> If you don't, please remove this package (pacman -R nvidia-glx)
To use this driver you need the nvidia-kernel package too.
If you need more information about setting up nvidia drivers have a look at:
"http://wiki.archlinux.org/index.php/How_to_install_NVIDIA_driver"
Note: OpenGL headers have been replaced (if applicable) by symlinks to the NVidia headers.
To switch to the mesa headers, reinstall mesa.
(You may want to do this if compiling a GL app for another system.)
EOF
post_upgrade() {
post_install $1
pre_remove() {
cat << EOF
==> Don't forget to update your /etc/X11/XF86Config or /etc/X11/xorg.conf!
You may want to remove nvidia-kernel as well.
EOF
op=$1
shift
$op $*
..As you can see, the nvidia-glx .install instates the NVidia gl headers to aid compiling against them. If mesa is installed it's versions of these files will be overwritten. Making sure to install the package for the desired headers last seemed like the simplest way to select them. Gentoo has some script to select the gl libs.
PS: This is designed alongside the current mesa-6.4.2-1 package in testing.
So. Any comments? Any good? Or shall I just keep them to myself?

Answer 1

tpowa wrote:well, problem is not that it's not possible, problem is if it is allowed.
i'm not a license expert but as far as i can remember no big distro provides the packages other then nvidia-installer.
gentoo is a special case.
Yeah, I thought that might be a problem. The LICENSE seems to have a clause to do with splitting up the package...
No Separation of Components.  The SOFTWARE is licensed as a
single product.  Its component parts may not be separated for use
on more than one computer, nor otherwise used separately from the
other parts.
But this refers to *use* of the seperate parts, which I suppose is a bit vague. Anyhoo, as I say, this modified install method can still work with a combined package. It could even not bother with the kernel module compile if it's already present.
The only other part that seems relevant is...
2.1.2  Linux/FreeBSD Exception.  Notwithstanding the foregoing terms
of Section 2.1.1, SOFTWARE designed exclusively for use on the Linux or
FreeBSD operating systems, or other operating systems derived from the
source code to these operating systems, may be copied and redistributed,
provided that the binary files thereof are not modified in any way
(except for unzipping of compressed files).
But since no binary files are modified this doesn't seem to apply. Of course, IANAL..