nvidia-utils and nvidia new modules

EAD · 2006-05-23 18:05:36

Hii, I want to use the new nvidia-utils 1.0.8762-2 and nvidia 1.0.8762-1 both from testing, and it refuse to work right with my Xorg :!:
Now I use the 81.xxx and it works very nice, Can some one help me here?
I have tried runing hwd -xa and even nvidia-xconfig and even to do it on my own but still it doesnt work
I get this lof file

(II) Setting vga for screen 0.
(**) NVIDIA(0): Depth 24, (--) framebuffer bpp 32
(==) NVIDIA(0): RGB weight 888
(==) NVIDIA(0): Default visual is TrueColor
(==) NVIDIA(0): Using gamma correction (1.0, 1.0, 1.0)
(**) NVIDIA(0): Enabling RENDER acceleration
(EE) NVIDIA(0): Failed to load the NVIDIA kernel module!
(EE) NVIDIA(0): *** Aborting ***
(II) UnloadModule: "nvidia"
(II) UnloadModule: "ramdac"
(II) UnloadModule: "fb"
(EE) Screen(s) found, but none have a usable configuration.

Fatal server error:
no screens found

Can some one help me out here please?

Fackamato · 2006-05-23 18:48:55

I have the exact same problem. Anyone? I'm using the latest nvidia-utils and nvidia aswell as the latest beyond kernel.

EmbraceThePenguin · 2006-05-23 20:30:06

I get this as well. Its almost as if the kernel module is not being built.

Thoughts???

Joe

Cerebral · 2006-05-23 20:42:26

Well, the kernel module _isn't_ being built anymore - it's prebuilt. The nvidia package in testing works with kernel26 - are you all using beyond? That's most likely the problem -- wait for nvidia-beyond.

EmbraceThePenguin · 2006-05-23 20:45:39

Ok, Yeah, I am using beyond.

Thanks for the explination...

Joe

EmbraceThePenguin · 2006-05-23 20:48:51

I actually have another question on this...

What if I want to build my own kernel (Or modify the beyond kernel). Is there going to be a package that allows you to build the driver for your own custom kernel?

Using pacman to install the nvidia driver is simpler than building the one from nvidia because Arch uses a different location for the libraries than what the stock nvidia driver uses and its a 10 step process to get the files in the right place.

For example, the stock nvidia driver installs to /usr/X11R6 and that is not existent with Arch's Xorg install.

Joe

Cerebral · 2006-05-23 20:59:32

Building modules for custom kernels is generally fairly simple - run abs (or check out the web-cvs entries at http://cvs.archlinux.org/cgi-bin/viewcv … ag=TESTING ) to get the nvidia PKGBUILD, then a quick modification to the _kernver variable and package name, and rebuild.

Most other kernel modules have been this way for a while; only nvidia was behind the times in that manner.

This method allows you to easily have the driver installed on multiple kernels at once, and also have all the modules properly managed by pacman.

EmbraceThePenguin · 2006-05-23 21:05:29

Cerebral wrote:

Building modules for custom kernels is generally fairly simple - run abs (or check out the web-cvs entries at http://cvs.archlinux.org/cgi-bin/viewcv … ag=TESTING ) to get the nvidia PKGBUILD, then a quick modification to the _kernver variable, and rebuild.
Most other kernel modules have been this way for a while; only nvidia was behind the times in that manner.

Well, for drivers that are built against something like a kernel, I would think that the best, simplest way would be to let the driver continue to build against the kernel.

This way, someone can make changes to their kernel and not worry about maintaining their own packages when updates come out; like modifying a PKGBUILD files.

Thoughts?

Cerebral · 2006-05-23 21:24:06

EmbraceThePenguin wrote:

Well, for drivers that are built against something like a kernel, I would think that the best, simplest way would be to let the driver continue to build against the kernel.

Heh, this was discussed a while ago when I made fglrx packages that mirrored the way the nvidia package used to work - at the time, I thought I had an awesome idea.

It turns out. however, that it's a bad solution, because then you only have one package, and so pacman only knows about one module. If you have multiple kernels, then it gets really hackish to track all the modules installed with just a single package, and they aren't directly managed by pacman that way.

Plus, there's the argument that the arch devs shouldn't need to maintain modules for kernels other than the kernels they have in their repos - they can't account for everybody's setup.

If you're knowledgeable enough to maintain your own kernel build, then certainly you're knowledgeable enough to maintain modules for it; especially when most of the work is done for you.

EmbraceThePenguin · 2006-05-23 21:58:12

Cerebral wrote:

EmbraceThePenguin wrote:
Well, for drivers that are built against something like a kernel, I would think that the best, simplest way would be to let the driver continue to build against the kernel.
Heh, this was discussed a while ago when I made fglrx packages that mirrored the way the nvidia package used to work - at the time, I thought I had an awesome idea.
It turns out. however, that it's a bad solution, because then you only have one package, and so pacman only knows about one module. If you have multiple kernels, then it gets really hackish to track all the modules installed with just a single package, and they aren't directly managed by pacman that way.
Plus, there's the argument that the arch devs shouldn't need to maintain modules for kernels other than the kernels they have in their repos - they can't account for everybody's setup.
If you're knowledgeable enough to maintain your own kernel build, then certainly you're knowledgeable enough to maintain modules for it; especially when most of the work is done for you.

Oh, I can maintain it, that is not a problem at all; however, I would rather "Keep It Simple" and only make changes to a kernel, then run pacman -S nvidia to reinstall the driver; instead of downloading CVS or abs files, moding a PKGBUILD file, makepkg, then, finaly, pacman -A or -U nvidia.

I don't mind maintaing my own packages (as I have had to more than a few times), but that is what I loved about Arch; up till recently, I haven't had to.

Hell, I mine as well go back to Gentoo or start a LFS project if I need to start maintaing my own packages...

*Sigh* - Don't mean to rant and complain, but Arch has realy gone down hill within the last year or so. Maybe its time I look else where; I am almost at the same frustration level I was at when I left Gentoo when it got too big and maintainers started the Microsoft way of imposing their way of thinking whats best for you and giving you no choice but to do things the hard way...

Anyway, thanks for the replys...

brain0 · 2006-05-23 22:03:20

You can keep it simple, just use the stock kernels. There are almost no reasons for not using them, as they are very modular and still provide all functionality you might need.

If you miss something, a bug report will most likely result in a change so that the feature can be added.

iphitus · 2006-05-24 07:48:58

brain0 wrote:

You can keep it simple, just use the stock kernels. There are almost no reasons for not using them, as they are very modular and still provide all functionality you might need.
If you miss something, a bug report will most likely result in a change so that the feature can be added.

The modules for the beyond kernel have yet to be uploaded, this is most likely an oversight by tpowa... I'll add them tonight...

James

iphitus · 2006-05-24 08:48:03

ok nvidia-beyond is in the repositories.

brain0 · 2006-05-24 11:21:06

iphitus wrote:

brain0 wrote:
You can keep it simple, just use the stock kernels. There are almost no reasons for not using them, as they are very modular and still provide all functionality you might need.
If you miss something, a bug report will most likely result in a change so that the feature can be added.
The modules for the beyond kernel have yet to be uploaded, this is most likely an oversight by tpowa... I'll add them tonight...
James

By "stock kernels" I meant -ARCH and -beyond, of course.

tpowa · 2006-05-24 11:27:15

the nvidia package was one of biggest mess up your system packages.
ok perhaps it was good for custom kernel users, but now it's a clear package that is completely tracked by pacman, you have to customisize it for yourself like any other extra module package we have.

iphitus · 2006-05-24 12:04:07

and the PKGBUILDs are much nicer and cleaner now too

nvidia-legacy-beyond is building now, it'll be in the repos in a few minutes.

edit: make that nforce-beyond too.

EAD · 2006-05-24 17:13:22

I have tried using the nvidia-utils and nvidia-beyond and still get the same error,Why is it? and how can I fix it up?
:?

Cerebral · 2006-05-24 19:45:12

xhectorx wrote:

why not use _kernver=`uname -r` that way the user won't have to change anything in the PKGBUILD?

Because the package is intended to be built for kernel 2.6.16-ARCH or 2.6.16-beyond, not necessarily for whatever kernel the dev happens to be running when they build the packages.

Besides, the user would still want to change the name of the package; otherwise you'd get pacman asking you to upgrade to the version in the arch repos every time it's updated, and if you're not careful you'd get your custom-kernel package 'upgraded' by a stock kernel package.

Mythoz · 2006-05-24 20:16:28

EmbraceThePenguin wrote:

Oh, I can maintain it, that is not a problem at all; however, I would rather "Keep It Simple" and only make changes to a kernel, then run pacman -S nvidia to reinstall the driver; instead of downloading CVS or abs files, moding a PKGBUILD file, makepkg, then, finaly, pacman -A or -U nvidia.

My first reaction to the new nvidia package was pretty much the same as yours. But putting this files under the control of pacman is really a nicer and cleaner way. I remember the update to xorg 7: I had to search and delete files which were left in the old location. Still do not know if I found all of them.

A simple solution for our problem (accommodativeness) is to write a simple shell script which do the work for us: copy the nvidia stuff from abs tree, modify _kernver to current kernel version and call makepkg && pacman -U.

Maybe the maintainer of the nvidia package is kind and puts such script into nvidia package itself?

iphitus · 2006-05-25 00:19:07

I have written a shell script you can use already, see my comment at the end of this closed bug.

http://bugs.archlinux.org/task/4691

I'll document it, enhance it, and package it up over the weekend for you. Then it should be *simpler* to update modules for custom kernels than it is now. I'll attempt to make it just a one line command like 'abs-module --update'

James

MillTek · 2006-05-25 05:32:47

Hi,
I was trying to install the new nvidia (1.0.8762-1) from testing and pacman said;
pacman -S nvidia
:: nvidia-utils conflicts with nvidia. Remove nvidia? [Y/n] n

error: package conflicts detected

Should I remove nvidia-utils?

Jim

iphitus · 2006-05-25 06:49:16

hit 'y' to let it remove nvidia and then reinstall it.

Mythoz · 2006-05-25 10:09:07

iphitus wrote:

I have written a shell script you can use already, see my comment at the end of this closed bug.
http://bugs.archlinux.org/task/4691
I'll document it, enhance it, and package it up over the weekend for you. Then it should be *simpler* to update modules for custom kernels than it is now. I'll attempt to make it just a one line command like 'abs-module --update'

Looks really nice! That is exactly the reason why I have chosen ArchLinux few years ago. It is simple, does it just right (pacman file tracking) but still give you the power if you want to go a slightly different way.

Thank you very much!

Neuro · 2006-05-26 20:47:16

Right, anybody could direct me to instructions on how to checkout the [testing] branch using ABS instead of the normal [current] branch? What should I put in /etc/abs supfiles?

If anyone could provide me with CVS commands to retrive the [testing] snapshot of nvidia, perhaps I'd be able to make up a script to automatically fetch the latest and greatest nvidia and compile a package automatically.

Arch Linux

#1 2006-05-23 18:05:36

nvidia-utils and nvidia new modules

#2 2006-05-23 18:48:55

Re: nvidia-utils and nvidia new modules

#3 2006-05-23 20:30:06

Re: nvidia-utils and nvidia new modules

#4 2006-05-23 20:42:26

Re: nvidia-utils and nvidia new modules

#5 2006-05-23 20:45:39

Re: nvidia-utils and nvidia new modules

#6 2006-05-23 20:48:51

Re: nvidia-utils and nvidia new modules

#7 2006-05-23 20:59:32

Re: nvidia-utils and nvidia new modules

#8 2006-05-23 21:05:29

Re: nvidia-utils and nvidia new modules

#9 2006-05-23 21:24:06

Re: nvidia-utils and nvidia new modules

#10 2006-05-23 21:58:12

Re: nvidia-utils and nvidia new modules

#11 2006-05-23 22:03:20

Re: nvidia-utils and nvidia new modules

#12 2006-05-24 07:48:58

Re: nvidia-utils and nvidia new modules

#13 2006-05-24 08:48:03

Re: nvidia-utils and nvidia new modules

#14 2006-05-24 11:21:06

Re: nvidia-utils and nvidia new modules

#15 2006-05-24 11:27:15

Re: nvidia-utils and nvidia new modules

#16 2006-05-24 12:04:07

Re: nvidia-utils and nvidia new modules

#17 2006-05-24 17:13:22

Re: nvidia-utils and nvidia new modules

#18 2006-05-24 19:45:12

Re: nvidia-utils and nvidia new modules

#19 2006-05-24 20:16:28

Re: nvidia-utils and nvidia new modules

#20 2006-05-25 00:19:07

Re: nvidia-utils and nvidia new modules

#21 2006-05-25 05:32:47

Re: nvidia-utils and nvidia new modules

#22 2006-05-25 06:49:16

Re: nvidia-utils and nvidia new modules

#23 2006-05-25 10:09:07

Re: nvidia-utils and nvidia new modules

#24 2006-05-26 20:47:16

Re: nvidia-utils and nvidia new modules

Board footer