On Sat, Oct 31, 2009 at 2:27 PM, Nikos Chantziaras <[email protected]> wrote: > On 10/31/2009 11:07 PM, Alan McKinnon wrote: >> >> On Saturday 31 October 2009 22:03:04 Nikos Chantziaras wrote: >>>> >>>> For instance, you might be running 2.6.31-r4 and also have 2.6.31-r3 >>>> installed. To install nvidia-drivers, you must build it twice - against >>>> each kernel you want to use it with (nvidia-drivers builds and installs >>>> a >>>> kernel driver into /lib/modules/<kernel version>) >>> >>> It's a bit more obfuscated than that. Maybe nvidia-drivers work >>> different, but ati-drivers will build against /usr/src/linux but install >>> the actual modules in /lib/modules/running_kernel. If /usr/src/linux >>> doesn't point to the running kernel, the modules will be installed in >>> the wrong place. >> >> >> That is just so mind-bogglingly absurdly stupid I doubt if ATI should even >> be >> allowed near a computer.... >> >> Compiling code never depends on something running, it only depends on >> things >> being present that can be linked against. >> >> Thanks for reminding me why I insist on NVidia GPUs, I'd forgotten. > > This isn't ATI's installer. It's the ebuild that does this.
And from deep memory it seems like there were other packages that operated this way 8-10 years ago. I know in 1999 I had to be very careful about where the linux link pointed, and while it's not as necessary today to do so i'm still quite careful. I use the ATI drivers on my AMD64 machine. I think I've always found that I needed to emerge fglrx after the new kernel had been booted but never understood why. This email is helpful. It seems to me that if it is the ebuild that's doing this is needs to be fixed. If I understand correctly I could be building for 2.6.31 but installing in 2.6.29? That's not right... Thanks to all for the info. Cheers, Mark
