> I didn't get a chance to read the entire message yet, but I
> thought I'd
> fire this off to you as you have run across a very common
> nVidia problem.
> 
> You should not build gl-dependent programs against the
> nVidia proprietary
> drivers (or the ATI proprietary drivers for that
> matter).  There is
> absolutely no guarantee that they will work without the
> nVidia GL library;
> e.g. on Intel or other non-proprietary GL-enabled
> graphics.
> 
> That being said, here is the fix:
> http://www.nvnews.net/vbulletin/showthread.php?s=a45bca4e6a691eb1552f284ab2a7036b&threadid=22875

"Doctor, doctor, my arm hurts when I do this!"

"Don't do that anymore!"

Sure, I inadvertently tripped over the solution: build without the proprietary nvidia package installed.

The challenge is that we can't guarantee end-users downstream will comply. In the build scripts for those three packages I can insert a test for libGL.la. If found, then halt/pause the build with a warning message. If a person builds only for personal use, as I have done for the past many years, building packages with nvidia installed is harmless because that is the only person using the packages. Such people will never notice problems or build failures, as I never did.

With that said, the thread you linked is 6 years old. Thus the problem still exists and the problem never was successfully resolved as promised.

I can't provide build scripts to other users that way, or pre-built packages. As end-users can use upstream versions of these packages, just as I did, probably the best I can do is insert warning messages in the build scripts along with commentation.

Ideas anyone?