Developer Guide

Hacking

Running a local copy of the client

Running the client in developer mode from your local tree is a little different than running letsencrypt-auto. To get set up, do these things once:

git clone https://github.com/letsencrypt/letsencrypt
cd letsencrypt
./bootstrap/install-deps.sh
./bootstrap/dev/venv.sh

Then in each shell where you’re working on the client, do:

source ./venv/bin/activate

After that, your shell will be using the virtual environment, and you run the client by typing:

letsencrypt

Activating a shell in this way makes it easier to run unit tests with tox and integration tests, as described below. To reverse this, you can type deactivate. More information can be found in the virtualenv docs.

Find issues to work on

You can find the open issues in the github issue tracker. Comparatively easy ones are marked Good Volunteer Task. If you’re starting work on something, post a comment to let others know and seek feedback on your plan where appropriate.

Once you’ve got a working branch, you can open a pull request. All changes in your pull request must have thorough unit test coverage, pass our integration tests, and be compliant with the coding style.

Testing

The following tools are there to help you:

  • tox starts a full set of tests. Please make sure you run it before submitting a new pull request.

  • tox -e cover checks the test coverage only. Calling the ./tox.cover.sh script directly (or even ./tox.cover.sh $pkg1 $pkg2 ... for any subpackages) might be a bit quicker, though.

  • tox -e lint checks the style of the whole project, while pylint --rcfile=.pylintrc path will check a single file or specific directory only.

  • For debugging, we recommend pip install ipdb and putting import ipdb; ipdb.set_trace() statement inside the source code. Alternatively, you can use Python’s standard library pdb, but you won’t get TAB completion…

Integration testing with the boulder CA

Generally it is sufficient to open a pull request and let Github and Travis run integration tests for you.

Mac OS X users: Run ./tests/mac-bootstrap.sh instead of boulder-start.sh to install dependencies, configure the environment, and start boulder.

Otherwise, install Go 1.5, libtool-ltdl, mariadb-server and rabbitmq-server and then start Boulder, an ACME CA server:

./tests/boulder-start.sh

The script will download, compile and run the executable; please be patient - it will take some time… Once its ready, you will see Server running, listening on 127.0.0.1:4000.... Add an /etc/hosts entry pointing le.wtf to 127.0.0.1. You may now run (in a separate terminal):

./tests/boulder-integration.sh && echo OK || echo FAIL

If you would like to test letsencrypt_nginx plugin (highly encouraged) make sure to install prerequisites as listed in letsencrypt-nginx/tests/boulder-integration.sh and rerun the integration tests suite.

Code components and layout

acme

contains all protocol specific code

letsencrypt

all client code

Plugin-architecture

Let’s Encrypt has a plugin architecture to facilitate support for different webservers, other TLS servers, and operating systems. The interfaces available for plugins to implement are defined in interfaces.py.

The most common kind of plugin is a “Configurator”, which is likely to implement the ~letsencrypt.interfaces.IAuthenticator and ~letsencrypt.interfaces.IInstaller interfaces (though some Configurators may implement just one of those).

There are also ~letsencrypt.interfaces.IDisplay plugins, which implement bindings to alternative UI libraries.

Authenticators

Authenticators are plugins designed to prove that this client deserves a certificate for some domain name by solving challenges received from the ACME server. From the protocol, there are essentially two different types of challenges. Challenges that must be solved by individual plugins in order to satisfy domain validation (subclasses of ~.DVChallenge, i.e. ~.challenges.TLSSNI01, ~.challenges.HTTP01, ~.challenges.DNS) and continuity specific challenges (subclasses of ~.ContinuityChallenge, i.e. ~.challenges.RecoveryToken, ~.challenges.RecoveryContact, ~.challenges.ProofOfPossession). Continuity challenges are always handled by the ~.ContinuityAuthenticator, while plugins are expected to handle ~.DVChallenge types. Right now, we have two authenticator plugins, the ~.ApacheConfigurator and the ~.StandaloneAuthenticator. The Standalone and Apache authenticators only solve the ~.challenges.TLSSNI01 challenge currently. (You can set which challenges your authenticator can handle through the get_chall_pref().

(FYI: We also have a partial implementation for a ~.DNSAuthenticator in a separate branch).

Installer

Installers plugins exist to actually setup the certificate in a server, possibly tweak the security configuration to make it more correct and secure (Fix some mixed content problems, turn on HSTS, redirect to HTTPS, etc). Installer plugins tell the main client about their abilities to do the latter via the supported_enhancements() call. We currently have two Installers in the tree, the ~.ApacheConfigurator. and the ~.NginxConfigurator. External projects have made some progress toward support for IIS, Icecast and Plesk.

Installers and Authenticators will oftentimes be the same class/object (because for instance both tasks can be performed by a webserver like nginx) though this is not always the case (the standalone plugin is an authenticator that listens on port 443, but it cannot install certs; a postfix plugin would be an installer but not an authenticator).

Installers and Authenticators are kept separate because it should be possible to use the ~.StandaloneAuthenticator (it sets up its own Python server to perform challenges) with a program that cannot solve challenges itself (Such as MTA installers).

Installer Development

There are a few existing classes that may be beneficial while developing a new ~letsencrypt.interfaces.IInstaller. Installers aimed to reconfigure UNIX servers may use Augeas for configuration parsing and can inherit from ~.AugeasConfigurator class to handle much of the interface. Installers that are unable to use Augeas may still find the ~.Reverter class helpful in handling configuration checkpoints and rollback.

Display

We currently offer a pythondialog and “text” mode for displays. Display plugins implement the ~letsencrypt.interfaces.IDisplay interface.

Writing your own plugin

Let’s Encrypt client supports dynamic discovery of plugins through the setuptools entry points. This way you can, for example, create a custom implementation of ~letsencrypt.interfaces.IAuthenticator or the ~letsencrypt.interfaces.IInstaller without having to merge it with the core upstream source code. An example is provided in examples/plugins/ directory.

Warning

Please be aware though that as this client is still in a developer-preview stage, the API may undergo a few changes. If you believe the plugin will be beneficial to the community, please consider submitting a pull request to the repo and we will update it with any necessary API changes.

Coding style

Please:

  1. Be consistent with the rest of the code.

  2. Read PEP 8 - Style Guide for Python Code.

  3. Follow the Google Python Style Guide, with the exception that we use Sphinx-style documentation:

    def foo(arg):
        """Short description.
    
        :param int arg: Some number.
    
        :returns: Argument
        :rtype: int
    
        """
        return arg
    
  4. Remember to use pylint.

Submitting a pull request

Steps:

  1. Write your code!

  2. Make sure your environment is set up properly and that you’re in your virtualenv. You can do this by running ./bootstrap/dev/venv.sh. (this is a very important step)

  3. Run ./pep8.travis.sh to do a cursory check of your code style. Fix any errors.

  4. Run tox -e lint to check for pylint errors. Fix any errors.

  5. Run tox to run the entire test suite including coverage. Fix any errors.

  6. If your code touches communication with an ACME server/Boulder, you should run the integration tests, see integration. See Known Issues for some common failures that have nothing to do with your code.

  7. Submit the PR.

  8. Did your tests pass on Travis? If they didn’t, it might not be your fault! See Known Issues. If it’s not a known issue, fix any errors.

Updating the documentation

In order to generate the Sphinx documentation, run the following commands:

make -C docs clean html

This should generate documentation in the docs/_build/html directory.

Other methods for running the client

Vagrant

If you are a Vagrant user, Let’s Encrypt comes with a Vagrantfile that automates setting up a development environment in an Ubuntu 14.04 LTS VM. To set it up, simply run vagrant up. The repository is synced to /vagrant, so you can get started with:

vagrant ssh
cd /vagrant
sudo ./venv/bin/letsencrypt

Support for other Linux distributions coming soon.

Note

Unfortunately, Python distutils and, by extension, setup.py and tox, use hard linking quite extensively. Hard linking is not supported by the default sync filesystem in Vagrant. As a result, all actions with these commands are significantly slower in Vagrant. One potential fix is to use NFS (related issue).

Docker

OSX users will probably find it easiest to set up a Docker container for development. Let’s Encrypt comes with a Dockerfile (Dockerfile-dev) for doing so. To use Docker on OSX, install and setup docker-machine using the instructions at https://docs.docker.com/installation/mac/.

To build the development Docker image:

docker build -t letsencrypt -f Dockerfile-dev .

Now run tests inside the Docker image:

docker run -it letsencrypt bash
cd src
tox -e py27

Notes on OS dependencies

OS level dependencies are managed by scripts in bootstrap. Some notes are provided here mainly for the developers reference.

In general:

  • sudo is required as a suggested way of running privileged process

  • Augeas is required for the Python bindings

  • virtualenv and pip are used for managing other python library dependencies

Ubuntu

sudo ./bootstrap/ubuntu.sh

Debian

sudo ./bootstrap/debian.sh

For squeeze you will need to:

  • Use virtualenv --no-site-packages -p python instead of -p python2.

Mac OSX

./bootstrap/mac.sh

Fedora

sudo ./bootstrap/fedora.sh

Centos 7

sudo ./bootstrap/centos.sh

FreeBSD

sudo ./bootstrap/freebsd.sh

Bootstrap script for FreeBSD uses pkg for package installation, i.e. it does not use ports.

FreeBSD by default uses tcsh. In order to activate virtualenv (see below), you will need a compatible shell, e.g. pkg install bash && bash.