|
Since I never use dkms, I can't easily say what is wrong, but the two times issue is almost certainly related to dkms cruft that sgfxi doesn't know about.
I've added extensive logging to sgfxi, and a debugger tools, so if you run: sgfxi --debugger it may give me an idea, that will gather a bunch of data then upload it to my server. I recently added a bunch of dkms debugging stuff to try to figure out why it's not removing cleanly like it's supposed to. When you run apt-get purge <something> that thing is supposed to be removed, and at least for a while, that was working with dkms cr#p, but now it looks like dkms is not purging correctly, and is leaving some cruft that tells the kernel to try to build a non existent module again on reboot. Tracking stuff like this down is a real pain in the butt, particularly when the package in question is incapable of actually keeping stable methods in place. dkms has always been a hack, though it is vastly improved over the early days. D stands for dell, who developed it, which should be all you need to know. As to why bad software is written, it's very simple: bad management leads to bad decisions which leads to bad employees, since nobody good will stay. Since software is an essentially 100% human construction, when you have bad management, bad project leaders, years of bad decision making, and the resulting bad programmers left over from all the decent ones who leave, disgusted, there is nothing and nobody remaining to make anything good happen. It's not rocket science. Good people want to work for good or at least decent companies. There are no software faeries who magically overrule all these human issues and give you good software and methods despite all these shortcomings. Further, amd a year or two ago started to massively defund their linux group, as part of a larger cost cutting effort, which probably means, this is my guess, most of the work has now been outsourced as well, which can only work if you have super strict oversight, and super competent project leaders, but see what I just wrote, so instead, you get cr#p, from crappy software outsourcing firms, that's my guess based on the pure randomness of the releases now. So there is no 'good' or 'right' you can appeal to, this is business, some are well run, others are not. Since amd cards are cheap, and generally good enough, and they generally do good enough windows driver support, that means that people buy them basically as a budget nvidia, which is frandkly silly because you can get last generation nvidia cards for 30 bucks, more if you want the super duper jumbo ones. Plus amd now has the apu things, which includes the amd video chip in the cpu, which I guess is good for various high volume computing environments. But linux, no, amd doesn't care about it. That's obvious now. Back to top |
|
I ran it, and it indicated it was successfully uploaded.
When you told me that dkms wouldn't work/apply, I went into synaptic and completely removed (purged) it, before running sgfxi -B after the addition of the patch. But as the upload indicates, the debugger must have found leftovers, because there are files referencing dkms. Back to top |
|
No, there's nothing I see leftover of dkms.
I'd have to actually see the install process for dkms to see what else it's doing. If you reboot and you don't get that attempting to build kernel module fglrx that means it's gone, if you do see it, it means some file somewhere is telling the kernel to try to build missing modules on boot. dkms has always been a pain. sgfxi by the way allows for easy switch back and forth between distro dkms driver and run download driver, that's: $ for dkms package distro driver sgfxi -d # regular default driver for your card, download/run package sgfxi It's also possible that on first reboot install attempt, if you get that message to reboot, what is actually needed is 'halt', not 'reboot', because reboot can leave some data in the motherboard ram that goes away on halt, I don't know for sure, but that can be a part. Back to top |
|
|
|
All times are GMT - 8 Hours |