NVidida OpenGL and XFree86 4.0

by David Beie » Wed, 24 May 2000 04:00:00

Hi,

does anybody knows how to work my TNT2-Card with XFree86 4.0 and with
the GLX-driver from NVIDIA.

The normal X-Server is working well, but when I start an GL-aplication,
it crashes, when I use the driver nv.

When I use the driver nvidia, so linux crashes when I start the
X-Server.

I the first case I can use Telnet to shutdown this machine, in the
second one there is no network available, so I need to do hard reset, to
restart my machine.

David

Top

1. XFree86 4.0.x, DRI and Mesa/OpenGL

I am in the process of setting up machines capable of hardware
accelerated OpenGL. Let me summarize what I have accomplished so far:
The machine(s) run(s) the 2.4.0 kernel (it is based on RockLinux,
www.rocklinux.org for the curious) and also has XFree86 4.0.2.
Following the clear compile instructions at dri.sourceforge.net, I
have in parallel to the /usr/X11R6 hierarchy a /usr/X11R6-DRI
hierarchy which contains the DRI enabled X-Server and basic libraries
(plus support for my Matrox g450). The /usr/X11R6-DRI/lib hierarchy
also contains the special Mesa 3.4 libGL library for use in
conjunction with DRI. DRI works and a program relying heavily on
OpenGL (vmd) is running nicely and clearly hardware accelerated.
(Seeing the before and after was quite an experience :-)

As you see, things work and I am (should be) quite content, but the
current procedure leaves somewhat of a mess. Not only do I have two X
hierarchies (/usr/X11R6, /usr/X11R6-DRI), but I also have OpenGL
(Mesa) related libraries in several places (on this machine
in /usr/X11R6/lib, /usr/X11R6-DRI/lib and /usr/lib). Further, I have been
following instructions, but I have to admit that it is / was not always
clear to me what I am / have been doing (in this sense pointers
to RTFMs appreciated as well!).

I would like to accomplish two things:

First, I don't want two X hierarchies. Is the X11R6-DRI code from the
DRI CVS a full XFree code base which allows me to build a full XFree
installation and not
just the server? If yes, what do I need to do (probably change
some entries in xf86site.def or host.def)? If no, any hints on
how to merge the two -- is it possible?

Second, what about Mesa (3.4): As far as I can tell, the code coming
with X11R6-DRI gives me libGL and the include files. Building standard
Mesa, however, also gives me libGLU (which I definitely need
for compiling certain applications). How can I unify/merge these two?

I am afraid if the library locations are not unified/cleaned up, then
I might end up in dynamic library conflicts. For example, what works
on this Rocklinux machine (with 2.4.0 and XF4.0.2) already leads to
trouble on a RedHat 6.2 based machine. While the new X-Server is definitely
running, even self-compiled programs don't load correctly with the
enlightening message:

#:ldd /usr/local/lib/vmd-1.6/vmd_LINUX
BUG IN DYNAMIC LINKER ld.so: dl-version.c: 210: _dl_check_map_versions: Assertion `needed != ((void *)0)' failed!

(I presume the reason for this is that vmd is a C++ program, the libGLU
present is the default RH 6.2 one (= 3.2), but the hardware accelerated
libGL library is based on Mesa 3.4)

Thanks for pointers/hints,

Stefan

2. pptpd on redhat 8.0 (2.4.18-14)

3. XFree86 4.0/Voodoo3/OpenGL/Glide questions

4. Where to get the w3mir/htget mirroring utility

5. Matrox 450 (DH) OpenGL app lockups with XFree86 4.0.x

6. Q: PS/2 mouse behaves erratically -- why?

7. XFree86 4.0 And Compiling OpenGL Programs

8. Solaris 32 or 64-bit

9. KStatusBar core dump - XFree86 4.0 + FreeBSD 4.0

10. FreeBSD 4.0 moused vs XFree86 4.0

11. Can't get XFree86 4.0 working on FBSD 4.0