I've been away all weekend but I did have a chance on Friday to poke around my system and I can't find any reference to DKMS installed on my system.

Could it be there are other packages related to dkms that are breaking things? Perhaps you could give me a few to search for and I can post the results?
Back to top

the error you listed suggests dkms, but it's quite possible that sgfxi did remove the parts it knows about, but that there's a part left.

Sadly, I was hoping someone could tell me, I really don't use or follow dkms since ever since the first time I saw/heard about it, I also saw the problems that would predictably come from its use.

Now there might be something else going on, but the problem is you used a variety of online how-tos but don't remember/know which you really used, which is what's wrong with using such random non scripted methods, they are simply unpredictable, especially when mixed.

I've tried to clean up all things known and handled, but I just don't have the energy to track stuff I don't deal with or use myself, if you link to all the how-tos you can find that you used I can maybe find the culprit.
Back to top

Well, in an interesting turn of events everything is working now and of course, I haven't a clue what fixed this. I performed a system-wide update a few days ago; I carefully combed through the list and no x.org, nvida, glx or other seemingly graphic or kernel related packages were install or upgraded. The second thing that happened is I got the .45 version of your script which I then ran.

So any one of these could have fixed it -although again, nothing appears to logically been done to directly affect or fix my issue. The most interesting part of all this is I rebooted several times after all of that just in normal day-to-day activities and it was still suffering from the same symptoms. But then out of nowhere I start the system when I got to clients yesterday or the day before I was presented with gmd and the login without having done anything. I was so shocked because I hadn't done anything different that I immediately rebooted to see it again...Sure enough it's been fixed since.

My money is on your latest revision of the sgfxi script but I really don't know so I may very well put myself back in that position at any point in time...

Thanks for all your assistance, it's well appreciated.

Cheers,
Phil
Back to top

Unfortunately you have just precisely duplicated the last bug report I got on this specific issue, including the problem magically going away.

I'm going to guess it's this:

when you upgraded, a new kernel was installed, and that rebuilt your initrdimg or something like that.

But again, I don't know, your issue was/is real, and it was caused by some procedure you followed on some how to, but until I find that procedure and learn how to unravel it via sgfxi, I won't be able to say more.

Since this issue pops up every few months, it's clear that there is a procedure out there that I am unaware of that does something that requires one additional cleanup step, or steps, to fully purge the dkms generated stuff from your system.

Apparently when something, probably a kernel, gets upgraded, that step is overwritten or removed, and the system returns to normal.

sgfxi always searches for and removes all nvidia packages in your system, my guess is one leaves some cruft that makes the system try to rebuild the driver with dkms. what that is, is beyond me. If I could make myself care even slightly I'd dig up all the main debian dkms installation howtos, follow them, then analyze the system until I could recreate the issue, then find out what else isn't being done to remove dkms totally.

Even better, if someone else can do this, that would be great, because I can't bring myself to spend any time on dkms at this point, though I'll do it some day when I'm totally bored I guess if nobody else does it for me first.
Back to top

Firstoff, I can help you further by telling you I'm 100% certain it's not a kernel upgrade. As I mentioned before I keep those for rainy days ;). On the other hand, I'm pretty sure I remember an initrd rebuild during the update before last - close to when this all started.

Second, I may have jumped the proverbial gun on this one. This morning everything was great. I booted up when I got to my first client and didn't have to run sgfxi.

No I just got to my last client of the day and have lost Direct Rendering. Contrarily to previous times however, the X server started fine. But I notice Compiz wasn't running all my pre-configs so I looked into it a little further and noticed I have no GL or DRI or Compiz. It seem almost like X.org defaulted to nv rather than the nvidia driver. Although it's never done that before!?

I was confused before. Now I'm completely lost in this. I'm 110% certain I did not update or change anything in any way from the time it was working this morning until now...Yet something has obviously changed!
Back to top

Okay, I did a little more digging and here's the deal:

dri gl and the likes are fine so is the nvidia driver. I exited X and re-ran the script and it essentially did nothing because nothing had changed.

Turns out I had simply lost all my Compiz settings. The same thing happened to many other apps including Evolution and XFCE4. No idea why since I didn't update or change anything. Anyway, that has nothing to do with your script.

For all intents and purposes the problem I had has quite literally disappeared for no apparent reason.

As for crud being left behind from all those howtos I tried before miraculously stumbling upon this site and sgfxi; I can say for certain that I first tried the Debian wiki and followed "the Debian way" which did not work for me. Not saying it doesn't work but for whatever reason it didn't work for me. I then followed this one tinyplanet.ca/~lsorense/debian/debian-nvidia-dri-howto.html
which didn't work either. I'm pretty sure it's after those howtos I found your script. And I suspect based on what your telling me my problems began with me not doing any housekeeping before embarking on another method of getting nvidia working.
Back to top

Oh, that link made me laugh, that's a very badly written how-to version of the main debian wiki version, which is what I used to create the framework of the original sgfxi script.

Hint: when you have a lot of steps, and a lot of embedded knowledge, that is what programming is for, and that's why we make scripts like sgfxi, to embed this knowledge and these methods into a simple, easy to use method.

I can't believe that debian is still putting that how-to as the way to install debian drivers when these problems have been solved for years now. I'm still honestly amazed that debian hasn't scripted these complicated things, it's really not impossible, especially not if you copy the GPL sgfxi script, which does that stuff if you want.

I didn't see a dkms thing there, but I think it's part of the nvidia package, so the question is, as it was, what did the removal of the debian nvidia packages leave behind that made the system look for the old driver instead of using the sgfxi installed one?
Back to top