Rebuilding Conda binaries

The binaries were built on a Scientific Linux 6.5 (Carbon). The default GCC (4.4.7) is too old for ROOT6, so we used the Developer Toolset (v2) from CERN. It provides everything you need to rebuild from the recipes: gcc/binutils/git/etc.. If you really want to rebuild or update the Anaconda binaries (typically not needed as we try keep them updated), you can:

  • use a VM: pickup a VirtualBox image from here, or create your own based on the ISO images from here. Check this guide for instructions on creating a VM.

  • Much easier: Docker image ready with the Developer Toolset (v2), cmake and Anaconda installed. To use it, install Docker and do:

     docker pull nlesc/slc6-devtoolset-anaconda
     docker run --rm -ti -e  MYUID=$UID  nlesc/slc6-devtoolset-anaconda /bin/bash
     git clone

In general, to be able to build conda packages, you need to have the conda-build package installed (already available if you are using the Docker image):

conda install conda-build

From the top directory of this cloned github repository of recipes, building a package is as easy as:

conda build ./directory-of-package-recipe

For packages which depend on Python (e.g., ROOT), you need to build a separate binary for each Python version you want to provide. Simply pass the CONDA_PY variable to the same command:

CONDA_PY={python-version} conda build ./directory-of-package-recipe

Remember, you need to build separately each such binary.

For packages which depend on ROOT (e.g., root-numpy), you need to, in addition, pass the ROOT version against which they should be build. For example:

CONDA_PY=3.4 ROOT_VERSION=5.34 conda build ./root-numpy