[GRASSLIST:3400] Re: NVIZ segfault

Glynn Clements glynn.clements at virgin.net
Wed May 12 13:17:29 EDT 2004


Carl Brown wrote:

> I saw it with the 5.03 debian package, now I'm still seeing it with 5.3-cvs. 
> I thought this had been fixed years ago: 
> http://www.intevation.de/rt/webrt?serial_num=783&display=History
> This suggests a video driver problem as a possible cause. 
> Video in this system is a Matrox G400, 32MB.

FWIW, I have a G400, and I have never had a segfault from NVIZ. More
generally, OpenGL tends to be less prone to causing segfaults than
most other libraries, as it doesn't use any structures (one common
cause of segfaults is accidental ABI changes resulting from unexpected
changes to structure layouts induced by changes in system types or the
compiler's alignment and/or padding rules).

Essentially, unless a backtrace indicates that the segfault actually
originated within the DRM driver, it's unlikely to be related to
OpenGL. [Note: if you are using indirect rendering, a segfault within
the GLX implementation will kill the X server rather than the client.]

Regarding NVIZ segfaults generally, my suspicion is that Tcl/Tk is the
most likely culprit, particularly if using Tcl/Tk 8.4. The use of
libpthread by Tcl/Tk or other libraries also seems to be a factor
(sometimes adding an explicit -lpthread to the command which links
NVWISH2.2 can cure segfaults).

-- 
Glynn Clements <glynn.clements at virgin.net>




More information about the grass-user mailing list