Build against Python environment

This is the most common case for a development installation on a Linux environment (including Ubuntu from WSL).

VirtualEnvWrapper is a set of convenient tools for working with virtual environments in Python. Install it with

pip install --user virtualenvwrapper
export WORKON_HOME=$HOME/.envs  # Put this in your ~/.bash_profile

Using VirtualEnvWrapper, you can set up a new virtual environment for compile DLite against Python 3.9, with

mkvirtualenv -p /usr/bin/python3.9 dlite39

The -p option to VirtualEnvWrapper allows you to select Python interpreter you want to build against.

Install Python dependencies to the new virtual environment

workon dlite39
pip install -U pip
pip install -r requirements.txt
pip install -r requirements_dev.txt

Create a new build folder and build DLite against Python 3.9

workon dlite39
mkdir build-dlite39
cd build-dlite39
cmake -DCMAKE_INSTALL_PREFIX=$VIRTUAL_ENV -DPython3_FIND_VIRTUALENV=ONLY -DPYTHON_VERSION=3.9 ..
cmake --build .
cmake --install .

If the -DPYTHON_VERSION=3.9 option to CMake is not sufficient to get CMake to compile against the correct libraries, you can make a copy of cmake/build-python3.7.cmake and configure CMake with

cmake -C ../cmake/build-python3.7.cmake ..

By default, virtualenv does not set LD_LIBRARY_PATH. This will result in errors when running, for example, dlite-codegen. To address this, DLite patches the virtualenv activate script to set LD_LIBRARY_PATH when running cmake --install . (see patch activate for details).

To update LD_LIBRARY_PATH in your current shell, please do

deactivate
workon dlite39

before running the tests

ctest