On Sunday 02 July 2006 15:54, Florian Kulzer wrote: > On Sun, Jul 02, 2006 at 13:41:22 +0300, David Baron wrote: > > After attempting to use Debian's version which crashed when trying to run > > any DRI application, I returned to the same version number from Nvidia. > > Works fine ... until the next bootup in which case the X simply does not > > start up. The log reveals a nearly complete sequence, no errors. The > > bootup logcheck may show: > > Jul 2 13:10:57 d_baron kdm: :0[4689]: IO Error in XOpenDisplay > > Jul 2 13:10:57 d_baron kdm[4644]: X server for display :0 terminated > > unexpectedly > > Jul 2 13:10:57 d_baron kdm[4644]: Display :0 cannot be opened > > Jul 2 13:10:57 d_baron kdm[4644]: Unable to fire up local display :0; > > disabling. > > > > I removed the Debian packages, reinstalled Nvidia's. Worked fine but next > > bootup, the same thing. In both cases Nvidia's reinstall says the > > installation was "modified" (it was not, not by me). > > > > One problem brought to mind is that the two installations use different > > directories for their various pieces. I should "only" have Nvidia's > > around. > > > > Is this a bug in the most recent version? > > What do I do about it? (could go back to previous version which gave me > > no trouble at all but that was before installing Debian's). > > Did you purge the nvidia-glx package? If you only removed it, then you > still have the /etc/init.d/nvidia-glx script which messes with the > symlink for libglx.so on every reboot. The failure to start X is > probably the result of trying to use the nvdidia driver with Xorg's > original libglx.so.
Seem to have done it! Thanks. I assume the other nvidia file /etc/init.d/nvidia-kernel is theirs since it is not necessary to have nvidia in /etc/modules (as suggested for the Debian install).
