The 304 driver supposedly will be kept useful until 2017, so I'd like to think there is a way for this to work.



I sent a bug report a few hours ago. It seems others are able to get the 304 driver working with the 3.16 kernel but I can't. The nVidia installer log suggested to do "make oldconfig && make prepare" in the kernel source, but I'm not sure how to do that.

I can't use the Stable distro because it doesn't have the graphics tablet drivers I need (Huion P608N). If there were rogue repos in my sources.list I don't know why, I guess I have to check it every time I do anything just to make sure some program hasn't trashed it (and it couldn't have been anything but the Debian GUI utilities). I have pruned it down to one line at the moment.

The only other thing I can think of trying is to keep installing older kernels and hope something works.
Back to top

I assume you've tried installing latest 304.125 driver with sgfxi, that should be working with 3.16, if not that, then 3.15, smxi has old kernels you can try, though I am starting to skip odd numbered ones to save server space.

I haven't gotten any 304 issue reports for newer kernels, but that doesn't mean much.

I know nvdia did a recent release, and those usually support current kernels, which would have been I think 3.16 when it was released, or thereabouts.

If you've filed an nvidia bug report, and have tried latest 304.125 sgfxi installs for your card, there's not much you can do beyond trying previous kernels, smxi has them back to I think 2.6.32 or so, though for your other device you probably need one that is more recent, just try it, it's easy.

Install old kernel, try driver install, reboot/start x, and see if it works, it should take about 5 minutes per test give or take.

I haven't followed to see if any patches are required, but usually patches are for failure to install at all, not failure to work due to issues in actual driver.
Back to top

In the smxi kernel install menu, option 3 (should be 3.15) just shows amd64 and if I select it it complains that I did not make a valid selection.

The error seems to be basically the same for each kernel, it complains about the -k option and suggests to do "make oldconfig && make prepare".
Back to top

Here is a recent thread that seems to be related to my problem:

forums.debian.net/viewtopic.php?t=119055
Back to top

That debian thread makes me want to cry, lol, the fact that after this long stupid installer failures like not checking headers etc are still present in the debian nvidia package is just sad.

There was a bug in the 64 bit kernel list, there is no 3.15, as noted, I only keep the last 2 kernels, plus even numbered ones due to space concerns.

So 3.14 is the one you want to try, I fixed the bug.

I honestly can't believe that the same tired nonsense is posted year after year about nvidia/fglrx driver installs, and that the nvdia packages don't do the same checks sgfxi has in it to determine which version to install, etc. sgfxi code is gpl, so there's truly no excuse for such pathetic handling of driver installation in debian, which of course is the only reason that sgfxi ever existed in the first place.

Anyway, try it again.

Note that the black screen has many possible causes, but you can totally ignore that debian thread in my opinion because you cannot debug nvidia issues using a repackaged driver, which is why sgfxi defaults to the real direct installed nvidia package.

Generally when you get a black screen, you want to go to console, and do this:

This almost always shows what error type it was.

Also, if sgfxi itself exits with an error on driver install, it will have logged the errors in /var/log/sgfxi/sgfxi.log

So ignoring the debian thread (look at devtalk.nvidia.com/default/board/98/ forums for the real issues and solutions, I would in general ignore debian forums for this type of information.

Once you've discovered what the actual Xorg error is, that helps, though that information is in nvidia's hands assuming you generated their bug report using their bug report generator tool, but it helps you know.

So you have to discover a few things in order to debug, and test other kernels, of course.
Back to top

Well, I did 'sgfxi -U', and the version number didn't change, so it doesn't look like sgxfi was updated.

I used synaptic to purge all the debian nvidia-related packages as well as the nvidia-related glx packages. I had to do it individually for each one, otherwise it would break all the dependencies. Are we still in the stone age?

After I did this I restarted the computer. This seemed to get the Noveau drivers working fully with GLX as a side-effect.

I then entered SMXI, and scraped everything clean except SMXI and it's config.

I then proceeded to graphics install, which surprisingly worked. The installer asked me about DKMS and I declined.

This was all on the default Debian 8 3.16 kernel. I restarted the computer and tried to do it on the Liquorix 3.16 kernel, which resulted errors which I sent a bug report for.

I won't be fully happy until someone indicates they know exactly what went wrong. But for now, I'm happier than I have been for the past year while I just gave up and booted up a livecd every day.
Back to top

If debian 3.16 works and liquorix 3.16 doesn't, you should probably talk to damentz/liquorix, either on these forums or irc: irc.oftc.net #smxi

Driver issues are not commonly varying between different spins of the same kernel version.

Since when has the nvidia driver asked about dkms? I've never seen that.

When I first started sgfxi, I literally went to the debian wiki for installing nvidia, and took all their ridiculous steps and checks and whatever, and automated them in sgfxi. That was about 6 years or so ago. My assumption had been that once these steps were clearly automated, the debian and other linux distros would start to fix their installers/packages, since the fixes were clearly available and using gpl code that could be safely copied and used. Instead, year after year, you have the same ridiculous failure to do basic card tests in the installers, poor dependencies, etc. The one area they did get working, though it took a long time, was dkms IF and only IF you used distro kernel and packaged driver. That is much better than it was before.

What errors resulted? sgfxi would have them logged.
Back to top

I've been away awhile so let me say this about that.

I have 32 & 64 sid on the same computer and use the 304 drivers for my Nvidia card.

I use towo's siduction kernels exclusively and am on 3.18 -2 at the moment. He does move along!

I don't recall ever having a problem with the drivers on any recent kernel.

I use only sgfxi since it's way faster, less than a minute, for an update.

I have no idea what this thread is about.
Back to top

I will try installing it on Liquorix again to get the logs.
Back to top

sedonix, I agree about having no idea about the debian stuff re issues with 304, but there are always possible situations where specific cards/chipsets have bugs with some kernels, even though most of the supported chipsets work. That happens I think more on the legacy branches where not as much testing happens.

I hadn't heard any 304 issues either, good to hear it works on 3.18 though.
Back to top