How to handle OpenGL requirement on Linux

[chapmanb]

Hi all;
I'm working on porting over R/bioconductor libraries to bioconda, and ran into an issue with those that require r-rgl. It links against libGLU, and gives errors:

** preparing package for lazy loading
Error : .onLoad failed in loadNamespace() for 'rgl', details:
  call: dyn.load(file, DLLpath = DLLpath, ...)
  error: unable to load shared object '/anaconda/envs/_build/lib/R/library/rgl/libs/rgl.so':
  libGLU.so.1: cannot open shared object file: No such file or directory

I could fix by installing an OpenGL implementation like Mesa: yum install mesa-libGLU-devel on the build box. Does everyone think this is the best plan to handle this, or know of a better/cleaner way to provide the dependency in conda without bringing in a huge X porting effort? Thanks for any help or ideas.

[johanneskoester]

To me, it sounds reasonable to assume a working X installation for such packages.

[chapmanb]

Johannes -- thanks much. Any objections from anyone if I add mesa-libGLU-devel to the CentOS build-bot Dockerfile?

[chapmanb]

Thanks again for the feedback. I added mesa in and updated the bioconda/bioconda-builder image on Docker Hub so it's now available for recipes. Thanks again.