Hello Slackers!
My Slackware is multilib-enabled.
I've been searching for the correct way to install the Nvidia SlackBuilds.org drivers including the 32-bit packages for multilib. When browsing through
https://docs.slackware.com/slackware:multilib there is a piece of text that says:
Quote:
If you were considering to use the convertpkg-compat32 script to convert a non-Slackware package to a -compat32 package, I must strongly advise against this. The script is written with a single purpose and that is to make 32bit versions of the official Slackware64 binaries/libraries available in a multilib setup. As such, the script will remove a lot of stuff that is present in the original 32bit package - stuff which is expected to have been installed as part of the 64bit version of the package.
In almost all cases where you have downloaded a non-Slackware 32bit package and want to make it work on Slackware64, the best way is to find the sources and build a 64bit version of the package. Alternatively, just install the original 32bit package instead of trying to “convert it” and then run it from the commandline to find out any missing 32bit libraries you may still have to extract from an official Slackware package.
|
If I understand correctly building and installing the 64-bit packages is pretty straightforward. Having all the source tarballs and the build script in the same directory I run ./nvidia-kernel.SlackBuild to get a package in /tmp directory. Installing it is as simple as doing installpkg nvidia-kernel-... and the same goes for Nvidia drivers.
But what confuses me is how should I build the 32-bit package. Should I run . /etc/profile.d/32dev.sh change ARCH variable to "i486" and then run the build script and later convert the package to compat32 and while still in the 32-bit environment install the package? Should I do it like this considering what is written in the quoted text and is it correct to install compat32 packages with the . /etc/profile.d/32.dev.sh still active?