December 24, 2003
So, What’s the Problem Again?
I have returned! But you will not yet receive the updates I wrote while away. Or the ones I wrote before I left.
For those, you’ll have to wait until I get my laptop somewhere that it will work its wireless magic. It could happen tomorrow, but I wouldn’t bet on it.
Instead, I give you most of the text of one of the emails sent to the linux users group at Mudd, to which I subscribe out of a mixture of equal parts political obligation, technical interest, and pure morbid curiosity. I’m posting it here in what is probably a violation of numerous federal laws.
[quote]I have a laptop with redhat 7.3 (kernel 2.4.18-3 with ieee1394 added) and an nvidia quadro4 550 GoGL. I installed the nvidia drivers version 1.0-3123, the regular way with kernel driver and GLX library source RPMs.
Recently, I was running the SPECViewperf 7.1.1 benchmarks and having trouble with crashes, so I upgraded to the latest driver, 1.0-4496. This was with the much fancier (ugh) NVIDIA installer binary, and had some trouble – I had to do a few installs before things got working (I made a stupid mistake involving .xinitrc that made me think X was failing, so I rolled back a few versions, but finally fixed it all and went back to 4496), and things were good. the SPECViewperf worked just fine, and compiled binaries (eg. glxgears) run no problem.
However, now when I go to build my old OpenGL projects, they fail miserably. the first problem was a preprocessor error, that there was an unclosed #if statement in /usr/include/GL/gl.h – I took a look, and sure enough, the file ended prematurely. I’m not sure exactly how it got truncated, but it was. so I moved it out of the way and copied /usr/share/doc/NVIDIA_GLX-1.0/include/GL/* into /usr/include/GL/. now, when I try to build my old projects, the preprocessor error goes away, and I get a long list of linker errors:
<Long List of Linker Errors Removed>
I get around 950 of these errors with different 5-digit numbers. there’s one other error that stands out:
/usr/lib/libGL.so: undefined reference to `__glProcessGlobalData’
which is towards the end.
Soooo… I ran ldconfig, same problem. Manually checked /usr/lib/libGL* :
<Long List of Things Modified Dec 10 Removed>
Dec 10 is when I did the upgrade to 4496. according to NVIDIA’s troubleshooting section of the README, this is all as it should be. I also double checked that the driver files are in place where they should be for the current kernel build and everything.
I’ve done a few objdumps and can’t figure out where __nvsymXXXXX is supposed to be defined, and I can’t get google to return a relevant result. this is starting to drive me crazy. Does anyone have any ideas about what could cause this problem…[/quote]
And it continues, on into the night.
My solution, and this is just off the top of my head, is to apply periodic force with a rock (version 184.108.40.206.3 pre 78) or other blunt instrument until the computer is beaten into submission