Nvidia 32-bit libraries for a new driver

Question

I've been experiencing kernal panics when running x with Nvidia 185.18.14 so I decided to update to the new 185.18.36 driver from the nvidia website. Now however, I can't find 32 bit lib utils for this driver in the AUR. Can I find these items myself and package them or do I just need to play the waiting game until someone finishes them and puts them up. I.e. Is this something that is out there just waiting to be put up onto the AUR, or something that someone needs some serious knowhow to create for the AUR?

Answer 1

(I am not an expert on this, but this is what I noticed..... if you want the conclusion that I found about the 32-bit libs, that is more toward the bottom of this post)
I had been wondering about the official NVIDIA driver from the NVIDIA site, and it's 32-bit libs for a while..... and this thread made me wonder about it some more, so I did a bit of investigating.
The first thing I did was to uninstall my nvidia-utils-beta package that I was using with the NVIDIA driver from the NVIDIA site.
I did this because I had noticed that the Arch nvidia PKGBUILD downloads the NVIDIA-Linux-x86_64-1**.**-pkg0.run and then runs "sh NVIDIA-Linux-x86_64-1**.**-pkg0.run --extract-only" to build/install the kernel module.
Then the nvidia-utils PKGBUILD does the same thing with the "sh NVIDIA-Linux-x86_64-1**.**-pkg0.run --extract-only" command, so it extracts everything else that is important from the NVIDIA-Linux-x86_64-1**.**-pkg0 directory, and then the PKGBUILD places them in the correct directories with the correct symbolic links:
But if you use the latest version of "NVIDIA-Linux-x86_64-190.25-pkg0.run" (from here: ftp://download.nvidia.com/XFree86/Linux-x86_64/190.25/ ), instead of using the Arch 'nvidia' and 'nvidia-utils' packages, then the "NVIDIA pkg0" will puts their libraries and symbolic links into the /usr/lib64 directory.
/usr/lib64:
libGL.la
libGL.so
libGL.so.1
libGL.so.190.25
libGLcore.so.1
libGLcore.so.190.25
libOSSlib.so
libcuda.so
libcuda.so.1
libcuda.so.190.25
libnvidia-cfg.so
libnvidia-cfg.so.1
libnvidia-cfg.so.190.25
libnvidia-tls.so.1
libnvidia-tls.so.190.25
libvdpau.so
libvdpau.so.1
libvdpau.so.190.25
libvdpau_nvidia.so
libvdpau_nvidia.so.190.25
libvdpau_trace.so
libvdpau_trace.so.190.25
tls
Then it puts all libXvMCNVIDIA into the /usr/lib directory same as Arch:
libXvMCNVIDIA.a
libXvMCNVIDIA.so.190.25
libXvMCNVIDIA_dynamic.so
libXvMCNVIDIA_dynamic.so.1
These are all put in the same place as the Arch packages:
/usr/share/applications/nvidia-settings.desktop
/usr/bin/nvidia-xconfig
/usr/bin/nvidia-settings
/usr/bin/nvidia-smi
/usr/bin/nvidia-bug-report.sh
/usr/lib/xorg/modules/drivers/nvidia_drv.so
/usr/lib/xorg/modules/extensions/libglx.so.190.25
The manual pages are also put into the same location as Arch's packages.
The last difference that I noticed is that the default Arch packages place everything that is found in here "/usr/share/doc/NVIDIA_GLX-1.0 directory" in the these directories (I'm sure I missed something else):
/usr/share/licenses/nvidia/LICENSE
/usr/share/pixmaps/nvidia-settings.png
/usr/share/doc/nvidia/README.txt
Now if you were to download the "NVIDIA-Linux-x86_64-190.25-pkg2.run" that includes the NVIDIA 32-bit libraries, this is where I noticed the problem is.
See, when you run:
sh NVIDIA-Linux-x86_64-190.25-pkg2.run
..... it installs the /NVIDIA-Linux-x86_64-190.25-pkg2/usr/lib32/*.so* files into the /usr/lib directory..... which is the wrong place..... it should be in the /opt/lib32/usr/lib directory.
So you could try this PKGBUILD that I just made..... I'm not sure if it will work, or if I did everything correct so USE IT AT YOUR OWN RISK!!!!!
# $Id: PKGBUILD 490 2009-08-01 11:12:15Z ibiru $
# Contributor: Pierre Schmitz <[email protected]>
# Contributor: Mikko Seppälä <[email protected]>
# Maintainer : Biru Ionut <[email protected]>
# Remade from : Methuselah (TRY AT YOUR OWN RISK)
_pkgsourcename=nvidia-utils-beta
pkgname=lib32-$_pkgsourcename
pkgver=190.25
pkgrel=1
pkgdesc="NVIDIA drivers utilities and libraries."
arch=(x86_64)
url="http://www.nvidia.com/"
license=('custom')
groups=('lib32')
depends=('lib32-libxext')
conflicts=('lib32-libgl' 'lib32-ati-fglrx-utils')
provides=('lib32-libgl')
source=(ftp://download.nvidia.com/XFree86/Linux-x86_64/190.25/NVIDIA-Linux-x86_64-190.25-pkg2.run)
md5sums=('4d4a3606c64f05cf975dc044ec9200a8')
build() {
cd $srcdir
mkdir -p $pkgdir/opt/lib32/usr/lib
mkdir -p $pkgdir/opt/lib32/usr/lib/tls
sh NVIDIA-Linux-x86_64-190.25-pkg2.run --extract-only
cp -dp NVIDIA-Linux-x86_64-190.25-pkg2/usr/lib32/*.so* $pkgdir/opt/lib32/usr/lib
cp -dp NVIDIA-Linux-x86_64-190.25-pkg2/usr/lib32/libGL.la $pkgdir/opt/lib32/usr/lib
cp -dpr NVIDIA-Linux-x86_64-190.25-pkg2/usr/lib32/tls $pkgdir/opt/lib32/usr/lib
# fix wrnog links
cd $pkgdir/opt/lib32/usr/lib
ln -sf libGL.so.$pkgver libGL.so
ln -sf libGL.so.$pkgver libGL.so.1
ln -sf libGLcore.so.$pkgver libGLcore.so.1
ln -sf libvdpau.so.$pkgver libvdpau.so.1
ln -sf libvdpau_trace.so.$pkgver libvdpau_trace.so.1
ln -sf libvdpau_nvidia.so.$pkgver libvdpau_nvidia.so.1
ln -sf libnvidia-tls.so.$pkgver libnvidia-tls.so.1
ln -sf libcuda.so.$pkgver libcuda.so.1
cd $pkgdir/opt/lib32/usr/lib/tls
ln -sf libnvidia-tls.so.$pkgver libnvidia-tls.so.1
So to finish this up, you could just install this NVIDIA package from their server: ftp://download.nvidia.com/XFree86/Linux … 5-pkg0.run
And then TRY my lib32-nvidia-utils-beta PKGBUILD. (if you feel it is correctly written and approved by someone else with a bit more experience with these things)