Arch Linux User Repository

For testing users, contributors: check out https://github.com/archlinux-jerry/nvidia-340xx
Warning: force push may happen to match the aur tree

Sadly I don’t have the expertise or setup to work with patches, but I have seen a patch for nvidia-390xx mentioned on the AUR forum. Might that help? See https://bbs.archlinux.org/viewtopic.php?pid=2118130#p2118130

Doesn't work with Linux 6.5.

@sebalis thank you. Your reply tells me that I'm not the only one concerned. The command you mention does indeed get around the opengl problem: by typing: 'GDK_DEBUG=gl-glx:opengl gnome-calculator' the application launches fine. Ditto for baobab... On the other hand, Totem always returns "OpenGL support initialization failed". Well, nothing serious for the moment, but I wanted to point it out.

I think it is getting old in a sense, because it seems that the driver doesn’t fully support what a modern OpenGL is expected to provide ... If I understand correctly.

I encountered something similar with baobab, the GNOME „disk usage analyzer“. Starting it from the command line, I get:

No provider of glGenSamplers found. Requires one of: Desktop OpenGL 3.3 GL_ARB_sampler_objects OpenGL ES 3.0

Research led me to https://gitlab.gnome.org/GNOME/gtk/-/issues/5858 and I ended up prefixing the command line with “usr/bin/env GDK_DEBUG=gl-glx” in my baobab starter. Now baobab is working again but I fear that this problem might spread.

Hello. I've been having some problems for a while with applications using OpenGL: 1- playing a video on Gnome's "totem" is no longer possible: "OpenGL support initialization failed"... 2- "gnome-calculator" no longer launches... 3- Qemu with "virtio-vga-gl" parameter is no longer possible (error messages). Is nvidia-340 getting old? or is it something else?

The very simple solution for me, is updating first gcc and so on, after that restart, and then updating linux and rebuilding nvidia dkms

I tried it without rebooting after the upgrade, but then my X session would not start on the next reboot. I did not need to disable the login manager precisely for that reason :-) Just did the install again from the terminal and rebooted once again, and I’m back.

Regarding IGNORE_CC_MISMATCH=1, I still believe that prepending “env CC=gcc” to the two calls to “make” is the best recipe (although the maintainer disagreed previously). Normally I don’t get any messages about a mismatch, but I did get the error this time when dkms tried to install the old nvidia-340xx-dkms on the new kernel. That is of course correct and helpful as an alert that recompiling is needed.

It's just stupid every time gcc updates to ask for it. The kernel is build with the new gcc and your driver with the old one. You have to recompile the driver. And don't use IGNORE_CC_MISMATCH=1. This is how I do it (it also might be easier to build it in clean chroot https://wiki.archlinux.org/title/DeveloperWiki:Building_in_a_clean_chroot#Building_in_the_chroot):

Upgrade your system. Uninstall "nvidia-340xx-dkms" and "nvidia-340xx-utils". Reboot in tty (in grub add "3" to the kernel parameters) or disable the login manager. Clone and build both "https://aur.archlinux.org/nvidia-340xx-utils.git" and "https://aur.archlinux.org/nvidia-340xx.git". Install the new driver and reboot.

You may be able to rebuild it without reboot (after the system upgrade) but cannot confirm it because I haven't tried.

dkms install --no-depmod nvidia/340.108 -k 6.4.7-arch1-2 Error! Bad return status for module build on kernel: 6.4.7-arch1-2 (x86_64) Consult /var/lib/dkms/nvidia/340.108/build/make.log for more information. `dkms install --no-depmod nvidia/340.108 -k 6.4.7-arch1-2' exited 10

Workaround: sudo env IGNORE_CC_MISMATCH=1 dkms install -m nvidia -v 340.108 -k 6.4.7-arch1-2