Building EMAN2/SPARX/SPHIRE from Source

Standard instructions for establishing a build and runtime environment for EMAN/SPARX/SPHIRE on Linux, macOS and WSL2 (Windows):

• While the instructions look long, in reality it should only take 10-15 minutes to complete the entire process (with a decent network connection).

• We use Conda for a working environment and most of the dependencies. Anaconda distribution has become ubiquitous for Python and R-based scientific computing and education over the last decade. While it may be possible to build the system without using Conda, we do not recommend doing this, and cannot provide support for a non-conda approach. If you follow the instructions below, you should have a painless source build very quickly.

• We make use of the environment system in conda, the package manager, to isolate EMAN2 dependencies from other conda packages you may have installed. If you need to install other packages you wish to use in concert with EMAN2, you will need to install them within the EMAN2 environment. Be warned that sometimes installing another package may trigger a version change in one of EMAN2's dependencies, which may or may not be a problem, depending on which dependency it is. We strongly suggest getting the base EMAN2 installed and working first, before trying to install any additional packages within the same conda environment.

• Note that even with a source build it may be difficult to get this working on systems with very old operating system installs. We normally try to support OS versions as much as 5-7 years old. Please report any problems.

• GPU Support (Linux Only!): For features which support the GPU, please complete the standard source install instructions below, then follow the GPU instructions from the binary installation page.

Initial Setup (one time)

1. Remove/disable other Anaconda/Miniconda/EMAN installations. If you do not have Anaconda/Miniconda (or another copy of EMAN2/SPARX/SPHIRE) already installed in your account, skip to step 2. If you DO have Anaconda or Miniconda installed in your account, you must ensure that any existing install you have is not active in your shell (not in PATH, no LD_LIBRARY_PATH, PYTHONPATH or PYTHONHOME set). While it may also be possible to set up an environment for EMAN2 in your existing Anaconda install, you will need to know what you're doing, and adapt these instructions to your situation.

      1 echo $PATH
      2 # make sure no Anaconda/Miniconda/EMAN2 entries
      3 echo $LD_LIBRARY_PATH
      4 echo $PYTHONPATH
      5 # ideally, both return nothing. If it set to something it is possible that it may interfere with Anaconda
      6 # strongly suggest at least during the install, ''unset'' both of these. After installation you can test
      7 # to see if they cause any issues
      8 
   

2. Download and install Miniconda: Linux or macOS

      1 bash <Miniconda-installer>
      2 
      3 # There are a variety of options you can use, but the default command above is sufficient in most cases
      4 bash <Miniconda-installer> --help
   

   and follow the prompts. When you see Do you wish the installer to initialize Miniconda3 by running conda init?, say no, then move on to the next step.

3. Initialize conda for shell interaction. These instructions will depend on what shell you use. The default on most systems is bash. If you use a different shell ( tsch, zsh, ... ), you will need to take this into account:

      1 # for bash-like shells, such as bash and zsh
      2 source <miniconda-path>/etc/profile.d/conda.sh
      3 
      4 # for csh-like shells, such as csh and tcsh:
      5 source <miniconda-path>/etc/profile.d/conda.csh
      6 
      7 # <shell-name> is bash, fish, powershell, tcsh, xonsh or zsh
      8 conda init <shell-name>
   

   As it says after you run this command, you will need to close and reopen your shell/terminal for it to take effect. This command modified your shell initialization so the conda activate command can be used properly. For more information on conda-init and activation, see Environment Activation.

   macOS Users (bash only)

   On macOS, this modifies ~/.bash_profile. If you have a ~/.profile startup file, creation of ~/.bash_profile will prevent ~/.profile from being read. A simple solution is to source .profile within .bash_profile.

   For differences between login- and non-login shells and order of reading the startup files on macOS, see, https://www.anintegratedworld.com/basics-of-osx-bashrc-v-profile-v-bash_profile/.

4. Configure and update conda.

      1 # We suggest the following, meaning you will need to use ''conda activate'' after logging in before using EMAN2
      2 conda config --set auto_activate_base False
      3 
      4 # Automatic conda updates may cause things to break, so we suggest making all package upgrades explicitly
      5 conda config --set auto_update_conda False
      6 
      7 # Install mamba (Not as necessary as it once was. Conda has improved. Mamba installation may be difficult now)
      8 conda install mamba -c conda-forge
      9 
     10 # Update base environment to get the latest conda
     11 mamba update --all -n base
   

5. Create a new environment with EMAN2 dependencies. eman2 below is the name of the environment. You may make this whatever you like, as long as you remember to use the same name when doing conda activate. Note that this name will appear as part of your prompt when activated, so you may want to keep it short.

      1 mamba create -n eman2 eman-dev --only-deps -c cryoem -c conda-forge
   

   To get the non-GUI variant which will not have pyqt and pyopengl, and can be installed on machines that do not provide GUI support and do not have any OpenGL packages installed, run one of the commands below:

      1 mamba create -n eman2 'eman-dev=*=nogui_*' --only-deps -c cryoem -c conda-forge
   

   List of dependencies: If you wish to see the list of conda dependencies that eman-deps is built from, look here. Some of the dependency versions are listed in the appropriate file under .ci_support.

   macOS Big Sur Users

   The new Big Sur macOS release (fall 2020) broke some things in Python, which are only now (Dec 2020) getting fixed. These fixes have not yet made their way into Anaconda. If you experience an error similar to "Unable to load OpenGL library", install a patched pyopengl package from cryoem channel (as of 1/13/2021).

   conda install pyopengl -c cryoem

   OR

   This quick fix seems to provide a temporary solution until Anaconda gets the necessary updates:

   You need to go into your miniconda3 or anaconda3 folder and find this file (the path may be slightly different on your install, but you should be able to find it):

   ~/anaconda3/envs/eman2/lib/python3.7/site-packages/OpenGL/platform/ctypesloader.py

   Around line 80 you should find a line like:

   fullName = util.find_library( name )

   change this line to:

   fullName = '/System/Library/Frameworks/' + name + '.framework/' + name

   (important to use 8 spaces before fullName, not a <tab>)

6. Ensure you have OpenGL:

   • EMAN2 uses OpenGL (via PyQt) for all of its graphics. OpenGL installation depends on OS variant and, for example, whether you are using proprietary NVidia drivers under Linux. You will need to have OpenGL set up on your machine as a whole before continuing. On Mac, you should already have this with XCode. On Linux with an NVidia driver you will likely also need to install the Mesa header files. If you aren't sure how to set up OpenGL, Google can probably help.

7. Get EMAN code from GitHub:cryoem/eman2.

      1 cd <path-where-you-want-eman2-source>   # eg - $HOME/src
      2 git clone https://github.com/cryoem/eman2.git # this will create an eman2 folder containing the current source code from the master branch
      3 
   

8. Create a build directory (out-of-source builds are recommended).

      1 mkdir <build-directory> # eg- $HOME/src/eman2-build
      2 
   

Build and Install: Daily development, Update code, etc.

When you start a new shell, these are the steps you will need to take before running EMAN programs or compiling the system:

1. Activate your environment. If you used a different name above, use it here too.

      1 conda activate eman2
   

2. Update and Checkout source code. Periodically you should update your source using standard git techniques.

      1 cd <source-directory>    # <path-where-you-want-eman2-source>/eman2
      2 git fetch --all --prune  # get remote branches and cleanup non-existing remote branches from local repo
      3 git checkout master      # to make sure you are building from the current master build rather than a branch someone is working on
      4 git pull                 # This is what actually updates your local copy of the code
      5 
   

3. CMake

      1 cd <build-directory>
      2 # on Mac:
      3 cmake <source-directory>   # - eg $HOME/src/eman2, optionally add -DCMAKE_VERBOSE_MAKEFILE
      4 
      5 # on Linux:
      6 cmake <source-directory> -DENABLE_OPTIMIZE_MACHINE=ON  # - eg $HOME/src/eman2, optionally add -DCMAKE_VERBOSE_MAKEFILE
      7 
   

   • cmake-gui

     • If you use cmake-gui, since conda is not in PATH anymore, cmake will fail to find the environment directory. In that case set CONDA_PREFIX to your conda environment directory manually.

     • Make sure to delete any cmake variables that cmake already found, variables like *_LIBRARY or similar, *_INCLUDE_PATH or similar, CONDA_EXECUTABLE, CMAKE_INSTALL_PREFIX and any variables that are expected to contain conda environment related values.
     • Configure and generate.

   • ccmake

     • If you use ccmake, you may get an error related to OpenGL. If this happens try quitting ccmake and running it again.

4. Make

      1 make -j 8
      2 make install
   

   Note that the make -j 8 above will compile using 8 threads. On some machines omitting the 8 will compile faster and cause no problems, and if you have more than 8 threads on your machine, you can increase the number.

Test install

• e2version.py
• e2speedtest.py
• e2display.py (not for clusters without GUI environments configured)
• note: "make test" is not recommended for individual installs, this tool is primarily aimed at developers making changes in the core of the system. It is not uncommon for one or more tests to fail, and there is no need to report such failures. In most cases this is a problem with the test, not a problem with the system.

• Other Environments. To switch to another conda environment (stop working with EMAN2), first deactivate your current environment.

     1 conda deactivate
  

Troubleshooting

Solutions to potential known problems will be listed here. If these don't help, please report your problem (http://groups.google.com/group/eman2):

• If using a newer C compiler on Linux, you may encounter an error like:

  /home/steve/miniconda3/envs/eman2/lib/libstdc++.so.6: version `GLIBCXX_3.4.29' not found

  Simply removing the offending libstdc++.so file will normally resolve this issue.
• Compilers on CentOS 6 and 7, do not support gcc's dual ABI feature which is needed to compile EMAN2's modern C++ code. On our build machine, devtools-7 and devtools-11 have been tested and confirmed to fail to compile EMAN2 code.

Reporting Problems

Please, provide the output of the following commands when reporting a problem. Record the session via script command. This, also, records the commands.

1.    1 script filename.txt
      2 
      3 conda activate eman2
      4 
      5 conda info -a
      6 conda list
      7 conda list --explicit
      8 
      9 cd <source-dir>
     10 
     11 git status
     12 git log -1
     13 
     14 cd <build-dir>
     15 
     16 rm CMakeCache.txt
     17 cmake <source-dir>
     18 cmake . -LA
     19 
     20 make clean
     21 make -j
     22 make install
     23 make test-verbose
     24 
     25 conda deactivate
     26 
     27 exit # or Ctrl+D
     28 
   

2. Send filename.txt.

Linux Clusters

The approach above will install EMAN with a precompiled version of OpenMPI, which may or may not work with the batch queuing system on your cluster. If it does not work, the symptom will be that MPI parallel jobs will use only a single node, no matter how many you have allocated in your job. Currently, we do not have alternative OpenMPI installation instructions.