Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#41078 - [nvidia-libgl] [mesa-libgl] must not conflict
Attached to Project:
Arch Linux
Opened by Evert Vorster (evorster) - Thursday, 03 July 2014, 21:38 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Saturday, 28 February 2015, 14:15 GMT
Opened by Evert Vorster (evorster) - Thursday, 03 July 2014, 21:38 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Saturday, 28 February 2015, 14:15 GMT
|
DetailsDescription:
I am running Arch on a laptop with Optimus technology. I would like to use the nVidia drivers for the nVidia GPU. The nVidia GPU is not physically connected to the monitor of the laptop, the Intel GPU is. Therefore the laptop needs have to have intel drivers installed at the same time as the nVidia drivers in order to give the nVidia GPU a buffer to render to. I can get the nVidia card to work ( mostly ) but gl offloading does not work, as nvidia-libgl is not installed, and it can't be, as that conflicts with mesa-libgl, which is a hard requirement of the intel drivers needed to drive the GPU which is actually attached to a monitor. Nice catch-22 situation there. I did, however manage to get this working with Sorcerer Linux, which was my previous distro, by installing the nvidia-libgl stuff alongside the mesa-libgl stuff, just in diffirent directories, and then specifying which to load in the xorf.conf. So it is possible, from my own experience. So, why are xorg-libgl and nvidia-libgl in conflict? Can I force Arch to install both at the same time? |
This task depends upon
Closed by Sven-Hendrik Haase (Svenstaro)
Saturday, 28 February 2015, 14:15 GMT
Reason for closing: Won't implement
Saturday, 28 February 2015, 14:15 GMT
Reason for closing: Won't implement
That's how I got it working in the beginning on Sorcerer Linux in the first place, and after a fashion here.
However, I am not interested in running bumblebee.
As stated in my bugreport, I want to run the nVidia GPU full-time, to have access to VDPAU and the stable nVidia drivers.
For this to work, both nvidia-libgl and mesa-libgl needs to be installed at the same time.
I am testing a custom nvidia-libgl package now, to see if I can make it not interfere with mesa-libgl. If I manage to get it working, it might end up as a package in AUR.
That would require mesa-libgl to become an optdepend on xf86-video-intel
My laptop, a HP ENVY dv7 is equipped with nVidia's Optimus technology.
In this specific laptop, the screen is hard-wired to the intel GPU.
The nVidia GPU can only render to a buffer on the Intel GPU.
In order to get any graphical display, I need to have xf86-video-intel installed.
I do a bit of gaming and video editing, and would really like to have the nVidia GPU run all the time, power savings be damned.
Following https://wiki.archlinux.org/index.php/Optimus I am able to run a setup like that.
Unfortunately, the driver xf86-video-intel depends on glamour-egl and intel-dri, which in turn depends on mesa-libgl
Because mesa-libgl and nvidia-libgl are in conflict, I am unable to install nvidia-libgl, which I need to unlock the full potential of my nVidia GPU.
I suppose nvidia-libgl installs files with the same name and location as mesa-libgl, in order for X to automatically find them.
I have been able, in the past, to install the libgl files from nvidia into a different directory, and then am able to access them with a custom xorg.conf, which has the ModulePath for the nvidia modules ahead of the ModulePath for the mesa modules.
If I am able to install xf86-video-intel without installing the mesa-libgl, I might be able to get this setup to work the way I want.
I had a peek at the nvidia-utils PKGBUILD. It looks like the nvidia installer wants to install the libgl files into /usr/lib/nvidia, and the package maintainer moved the location of those files to /usr/lib, which is the location that mesa-libgl is using. I suppose this is to make the X server automatically detect the nVidia GL files?
Kind regards,
Evert Vorster
It feels a bit unclean, and I know it will break with the next system update.
It would be best if there was a way to specify to pacman that I want nvidia-libgl installed instead of mesa-libgl.
Kind regards,
Evert Vorster
Please close this bug. The point is moot, the nvidia vdpau drivers are no more stable than the open source ones, and my CPU is faster than either GPU at decoding video.
One thing is for certain, I don't want another laptop with this hybrid GPU technology.