Comment 20 for bug 289925

It seems like we have something like the following:

 * race condition where native drivers are fast enough not to crash (i.e. they are filling out an internal structure before the call to glutInit)
 * use of a GL operation by FreeGLUT that requires a GL context under MESA, but does not under the proprietary drivers
 * use of a MESA operation that is buggy in software but correct in the common hardware implementations

I'll try again to reproduce the effect on my Ubuntu machine.