Django Autotest on Hudson/Jenkins with VirtualEnv
September 29, 2011
I wanted to run an autotest (./manage.py test) on my Django project, after first installing dependencies into a virtualenv. Here's the setup I used.
Jenkins Environment Variables
In the build, set the environment variables so that the virtualenv is added to the PATH, and the workspace is added to the PYTHONPATH:
PATH=${WORKSPACE}/.env/bin:$PATH
PYTHONPATH=$PYTHONPATH:${WORKSPACE}
Jenkins Build - Execute Shell
In the Jenkins build, we need to execute a shell script:
if [ -d ".env" ]; then echo "**> virtualenv exists" else echo "**> creating virtualenv" virtualenv --no-site-packages .env fi source .env/bin/activate pip install -r requirements.txt cd project_folder ./manage.py test --settings=settingsautotest
Note that I'm using a specific settings file for the autotest.
Also note that we're loading dependencies into the virtualenv using a requirements.txt file. This file for me contains:
psycopg2==2.2.1 slimmer gunicorn
All the other dependencies I had are included in the project folder and version controlled along with my project source.
We create the virtualenv using the --no-site-packages flag in order to limit our dependency on the system installed packages.
Problems
I had a problem installing psycopg2 under the virtual env - it failed with the following error:
Error: pg_config executable not found.
This was solved using by installing a few extra packages:
# apt-get install libpq-dev python-dev
Once this was sorted I got the following error:
psycopg2.ProgrammingError: autocommit cannot be used inside a transaction
This was solved by limiting the psycopg2 version to 2.2.1 in the requirements.txt file, rather than the default 2.4.2.
