vmd 1.8.7 crashed by TrueColor visual in ubuntu 10.04

[said]

My laptop have a Quadro NVS 140M graphic card, and running with ubuntu 10.04. I have installed the latest nvidia drive from ubuntu ppa like:

$ aptitude show nvidia-current
Package: nvidia-current
State: installed
Automatically installed: no
Version: 256.35-0ubuntu2~xup2

vmd 1.8.7 binary is downloaded at vmd offical website. The package was chosen as
LINUX OpenGL, CUDA (Linux (32-bit, RHEL4 or later) with CUDA)

I am sure installed vmd 1.8.7 correctly since there is even no warning from make install command. However, when I run vmd command, it gives following errors:

$ vmd
Info) VMD for LINUX, version 1.8.7 (August 1, 2009)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd@ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 2 CPUs detected.
Info) Free system memory: 2520MB (83%)
Info) Detected 1 available CUDA accelerator:
Info) [0] Quadro NVS 140M 2 SM_1.1 @ 0.80 GHz, 127MB RAM, KTO, OIO
Info) Creating CUDA device pool and initializing hardware...
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
ERROR) A TrueColor visual is required, but not available.
ERROR) The X server is not capable of displaying double-buffered,
ERROR) RGB images with a Z buffer. Exiting ...
Info) VMD for LINUX, version 1.8.7 (August 1, 2009)
Info) Unable to create OpenGL window.

and vmd crashed.

I have asked vmd supporter John Stone. He recommend me post this error on nvidia forum here.

Any suggestions is appreciate.

Best regards.

Sai Duan

[AaronP]

Please see the forum sticky posts about how to report a problem, particularly the part about how to generate an nvidia-bug-report.log.gz file. Please also attach the output of xdpyinfo and glxinfo to your reply.

[said]

Originally Posted by AaronP

Please see the forum sticky posts about how to report a problem, particularly the part about how to generate an nvidia-bug-report.log.gz file. Please also attach the output of xdpyinfo and glxinfo to your reply.

Sorry, I have attached the nvidia-bug-report.log.gz and xdpyinfo.log now. The glxinfo output is quit simple and I will paste here:

$glxinfo
name of display: :0.0
Error: couldn't find RGB GLX visual or fbconfig



######################################
I just found that If I run 'sudo glxinfo', it works. And the output is attached as glxinfo.log

[AaronP]

Originally Posted by said

I just found that If I run 'sudo glxinfo', it works. And the output is attached as glxinfo.log

That's very interesting... the glxinfo output from root looks fine. Does your user have access to the /dev/nvidia* device files?

[AaronP]

Actually, I take that back. Not having access to the device files would just result in indirect rendering. I suspect that something about your environment is causing OpenGL applications to use Mesa instead of the NVIDIA OpenGL library. Mesa has a bug that causes it to fail to fall back to indirect rendering correctly, and causes an error very much like the one you're seeing.

What does "ldd /usr/bin/glxinfo" show when you run it as root vs. non-root? Do you have LD_LIBRARY_PATH set in your environment?

[said]

Originally Posted by AaronP

That's very interesting... the glxinfo output from root looks fine. Does your user have access to the /dev/nvidia* device files?

The output for the permission of nvidia* files looks like:
$ ll /dev/nvidia*
crw-rw-rw- 1 root root 195, 0 2010-07-24 12:28 nvidia0
crw-rw-rw- 1 root root 195, 255 2010-07-24 12:28 nvidiactl

I think it is fine.

Originally Posted by AaronP

Actually, I take that back. Not having access to the device files would just result in indirect rendering. I suspect that something about your environment is causing OpenGL applications to use Mesa instead of the NVIDIA OpenGL library. Mesa has a bug that causes it to fail to fall back to indirect rendering correctly, and causes an error very much like the one you're seeing.

What does "ldd /usr/bin/glxinfo" show when you run it as root vs. non-root? Do you have LD_LIBRARY_PATH set in your environment?

$ ldd /usr/bin/glxinfo
linux-gate.so.1 => (0x4001d000)
libGL.so.1 => /usr/lib/mesa/libGL.so.1 (0x40020000)
libm.so.6 => /lib/tls/i686/cmov/libm.so.6 (0x40097000)
libX11.so.6 => /usr/lib/libX11.so.6 (0x400bd000)
libpthread.so.0 => /lib/tls/i686/cmov/libpthread.so.0 (0x401da000)
libc.so.6 => /lib/tls/i686/cmov/libc.so.6 (0x401f3000)
libXext.so.6 => /usr/lib/libXext.so.6 (0x4034d000)
libXxf86vm.so.1 => /usr/lib/libXxf86vm.so.1 (0x4035e000)
libXdamage.so.1 => /usr/lib/libXdamage.so.1 (0x40364000)
libXfixes.so.3 => /usr/lib/libXfixes.so.3 (0x40368000)
libdrm.so.2 => /lib/libdrm.so.2 (0x4036e000)
libdl.so.2 => /lib/tls/i686/cmov/libdl.so.2 (0x40379000)
/lib/ld-linux.so.2 (0x40000000)
libxcb.so.1 => /usr/lib/libxcb.so.1 (0x4037e000)
librt.so.1 => /lib/tls/i686/cmov/librt.so.1 (0x40398000)
libXau.so.6 => /usr/lib/libXau.so.6 (0x403a1000)
libXdmcp.so.6 => /usr/lib/libXdmcp.so.6 (0x403a5000)
and
$ sudo ldd /usr/bin/glxinfo
linux-gate.so.1 => (0x00a25000)
libGL.so.1 => /usr/lib/nvidia-current/libGL.so.1 (0x005d9000)
libm.so.6 => /lib/tls/i686/cmov/libm.so.6 (0x00a4f000)
libX11.so.6 => /usr/lib/libX11.so.6 (0x006f2000)
libpthread.so.0 => /lib/tls/i686/cmov/libpthread.so.0 (0x00c9b000)
libc.so.6 => /lib/tls/i686/cmov/libc.so.6 (0x0024c000)
libnvidia-tls.so.256.35 => /usr/lib/nvidia-current/tls/libnvidia-tls.so.256.35 (0x00e8e000)
libnvidia-glcore.so.256.35 => /usr/lib/nvidia-current/libnvidia-glcore.so.256.35 (0x00e90000)
libXext.so.6 => /usr/lib/libXext.so.6 (0x00cd9000)
libdl.so.2 => /lib/tls/i686/cmov/libdl.so.2 (0x00110000)
/lib/ld-linux.so.2 (0x00da7000)
libxcb.so.1 => /usr/lib/libxcb.so.1 (0x00ac5000)
libXau.so.6 => /usr/lib/libXau.so.6 (0x00114000)
libXdmcp.so.6 => /usr/lib/libXdmcp.so.6 (0x00bbb000)

It seems normal users don't use libnvidia-tls.so.256.35 and libnvidia-glcore.so.256.35.

The 'echo $LD_LIBRARY_PATH' show that I used mesa at /usr/lib/mesa
If I get ride of this lib I can run vmd 1.8.7 now. But it required may resources.

$ vmd
Info) VMD for LINUX, version 1.8.7 (August 1, 2009)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd@ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 2 CPUs detected.
Info) Free system memory: 2529MB (84%)
Info) Detected 1 available CUDA accelerator:
Info) [0] Quadro NVS 140M 2 SM_1.1 @ 0.80 GHz, 127MB RAM, KTO, OIO
Info) Creating CUDA device pool and initializing hardware...
CUDA error: no CUDA-capable device is available, line 62
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
Info) OpenGL renderer: Quadro NVS 140M/PCI/SSE2
Info) Features: STENCIL MSAA(4) MDE CVA MTX NPOT PP PS GLSL(OVF)
Info) Full GLSL rendering mode is available.
Info) Textures: 2-D (8192x8192), 3-D (2048x2048x2048), Multitexture (4)


################################################## ########
I suspect the high computer resource required by vmd should be related to the high nvidia driver. So I upgraded the nvidia driver.
$ aptitude show nvidia-current
Package: nvidia-current
State: installed
Automatically installed: no
Version: 195.36.24-0ubuntu1~10.04

Everything works fine except a warning of vmd 1.8.7

$ vmd
Info) VMD for LINUX, version 1.8.7 (August 1, 2009)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd@ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 2 CPUs detected.
Info) Free system memory: 2641MB (87%)
Info) Detected 1 available CUDA accelerator:
Info) [0] Quadro NVS 140M 2 SM_1.1 @ 0.80 GHz, 127MB RAM, KTO, OIO
Info) Creating CUDA device pool and initializing hardware...
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
Info) OpenGL renderer: Quadro NVS 140M/PCI/SSE2
Info) Features: STENCIL MSAA(4) MDE CVA MTX NPOT PP PS GLSL(OVF)
Info) Full GLSL rendering mode is available.
Info) Textures: 2-D (8192x8192), 3-D (2048x2048x2048), Multitexture (4)

[AaronP]

Originally Posted by said

$ ldd /usr/bin/glxinfo
linux-gate.so.1 => (0x4001d000)
libGL.so.1 => /usr/lib/mesa/libGL.so.1 (0x40020000)

There's your problem. Like I said, Mesa is buggy, and will prevent your OpenGL applications from working correctly when used with the NVIDIA driver. Please contact your Linux distribution to find out why it's picking up libGL.so.1 from the non-standard /usr/lib/mesa directory instead of the standard /usr/lib.

[said]

Originally Posted by AaronP

There's your problem. Like I said, Mesa is buggy, and will prevent your OpenGL applications from working correctly when used with the NVIDIA driver. Please contact your Linux distribution to find out why it's picking up libGL.so.1 from the non-standard /usr/lib/mesa directory instead of the standard /usr/lib.

I think this is due to I want to compile other software which require freeglut-dev package. And this automatic installed the dependency libglu-dev.

Anyway, I removed /usr/lib/mesa from LD_LIBRARY_PATH and everything works well. Thanks for your help.