Re: the nvidia/osmesa dilema

[Daniel J Sebald]

On 05/15/2016 10:12 AM, Andreas Weber wrote:
> On 04.05.2016 12:36, Juan Pablo Carbajal wrote:
> > We still have this issue with building the documentation and running
> > tests (https://savannah.gnu.org/bugs/?44478) that casues segfaults
> > when the user is running with nvidia drivers.
>
> Hi all,
> I've been playing around with jwe's idea to create an external filter
> program which renders the plot and isn't linked with libGL.
>
> Can someone explain why this simple code works with proprietary nvidia
> drivers?
> https://github.com/Andy1978/octave-debugging/blob/master/hgload_print/render_ofig.cpp
>
>
> The Makefile create an .ofig with
> "figure visible off; sombrero; hgsave sombrero.ofig;"
>
> and render_ofig calls hgload ("sombrero.ofig") print ("sombrero.pdf")
> which uses __osmesa_print__ and gl2ps.
>
> it's also linked against libGL.so:
> $ ldd render_ofig | grep libGL
>   libGL.so.1 => /usr/lib/x86_64-linux-gnu/libGL.so.1 (0x00007f5cc908a000)
>   libGLU.so.1 => /usr/lib/x86_64-linux-gnu/libGLU.so.1 (0x00007f5cc8e1c000
>
> $ readlink -e /usr/lib/x86_64-linux-gnu/libGL.so.1
> /usr/lib/x86_64-linux-gnu/nvidia/legacy-304xx/libGL.so.304.131
>
> why doesn't this segfault?

What order are the OSMesa and GL libraries being linked into this program?

Dan