Arch Linux User Repository

to conform with the other VA-API driver packages

@kajlao What part of this do you not understand. libva-intel-driver is also changed for this reason.

@Neurognostic libva-nvidia-driver should be renamed to nvidia-vaapi-driver, this is the correct name of the repo and the project.

@vkxx Your warning is unnecessary. This is a VCS PKGBUILD and behaves exactly as expected. Only thing wrong with it is it should be renamed to libva-nvidia-driver-git to conform with the other VA-API driver packages.

@alphabitserial The repository includes the headers from the open-gpu-kenel-modules project, so to use the direct backend the patch is not strictly necessary but can be used if you want the latest headers.

Warning: do not install this AUR because libva-nvidia-driver AUR is the same library but current version with bugs corrected! Maybe it's more optimal to remove AUR nvidia-vaapi-driver-git completely, but I don't know how to request it.

@kajlao: Please see the repository's README.

IMPORTANT: The direct backend is currently required on NVIDIA driver series 525 due to a regression (see issue #126).

and in the section on the direct backend:

This backend uses headers files from the NVIDIA open-gpu-kernel-modules project.

Sure, it compiles fine without it, but the EGL backend doesn't work on the latest version of the Nvidia driver. The PKGBUILD should be updated as per Sapphirus's suggestion to enable the user to choose between the default EGL backend and the Direct backend that works with the latest driver.

Please makes it conflict with libva-vdpau-driver

It compiles fine for me.. And why you're using open-gpu-kernel-modules? What is this, Sapphirus.. Unflag this package.

For the uninitiated, here's a patch to make the PKGBUILD work:

3c3
< pkgver=0.0.4.r8.g8806b61
---
> pkgver=0.0.7.r42.ge2d256e
24c24
<     #sed -i "s|usr/include/cuda|opt/cuda/include|" meson.build
---
>     #sed -i "s|usr/include/cuda|opt/cuda/include|" meson.build
25a26,29
>
>     git clone --progress https://github.com/NVIDIA/open-gpu-kernel-modules.git nvidia-open
>     ./extract_headers.sh nvidia-open
>

Just copy and paste the two lines below #sed -i if you don't want to use patch.

It's fixed in the latest git

Compilation fails due to recent merging of the direct branch which necessitates that NVIDIA's open-gpu-kernel-modules be included for its header files. This requires the build script be modified to include pulling down that repo in addition and using the required extract_headers.sh to copy them to the correct working directory.