Creating a pre-receive hook script

In this article

• Writing a pre-receive hook script
• Setting permissions and pushing a pre-receive hook to GitHub Enterprise Server
• Testing pre-receive scripts locally
• Further reading

You can see examples of pre-receive hooks for GitHub Enterprise Server in the github/platform-samples repository.

Writing a pre-receive hook script

A pre-receive hook script executes in a pre-receive hook environment on the GitHub Enterprise Server appliance. When you create a pre-receive hook script, consider the available input, output, exit-status and environment variables.

Input (stdin)

After a push occurs and before any refs are updated on the remote repository, the git-receive-pack process invokes the pre-receive hook script with the standard input of one line per ref to be updated:

<old-value> SP <new-value> SP <ref-name> LF

This string represents these arguments:

For more information on git-receive-pack see "git-receive-pack" in the Git documentation. For more information about refs see "Git References" in Pro Git.

Output (stdout)

The script output (stdout) is passed back to the client, so any echo statements are visible to the user on the command line or in the user interface.

Exit-status

The exit-status of a pre-receive script determines if the push will be accepted.

Environment variables

Outside of the values that are provided to stdin, there are additional variables that are available to a pre-receive hook script running on GitHub Enterprise Server.

Setting permissions and pushing a pre-receive hook to GitHub Enterprise Server

A pre-receive hook script is contained in a repository on the GitHub Enterprise Server appliance. A site administrator must take into consideration the repository permissions and ensure that only the appropriate users have access.

We recommend consolidating hooks to a single repository. If the consolidated hook repository is public, the README.md can be used to explain policy enforcements. Also, contributions can be accepted via pull requests. However, pre-receive hooks can only be added from the default branch. For a testing workflow, forks of the repository with configuration should be used.

1. For Mac users, ensure the scripts have execute permissions:

   $ sudo chmod +x SCRIPT_FILE.sh

   For Windows users, ensure the scripts have execute permissions:

   git update-index --chmod=+x SCRIPT_FILE.sh

2. Commit and push to your designated pre-receive hooks repository on the GitHub Enterprise Server instance.

   $ git commit -m "YOUR COMMIT MESSAGE"
   $ git push

3. Create the pre-receive hook on the GitHub Enterprise Server instance.

Testing pre-receive scripts locally

You can test a pre-receive hook script locally before you create or update it on your GitHub Enterprise Server appliance. One method is to create a local Docker environment to act as a remote repository that can execute the pre-receive hook.

1. Ensure Docker is installed locally.

2. Create a file called Dockerfile.dev containing:

   FROM gliderlabs/alpine:3.3
   RUN \
     apk add --no-cache git openssh bash && \
     ssh-keygen -A && \
     sed -i "s/#AuthorizedKeysFile/AuthorizedKeysFile/g" /etc/ssh/sshd_config && \
     adduser git -D -G root -h /home/git -s /bin/bash && \
     passwd -d git && \
     su git -c "mkdir /home/git/.ssh && \
     ssh-keygen -t rsa -b 4096 -f /home/git/.ssh/id_rsa -P '' && \
     mv /home/git/.ssh/id_rsa.pub /home/git/.ssh/authorized_keys && \
     mkdir /home/git/test.git && \
     git --bare init /home/git/test.git"
   
   VOLUME ["/home/git/.ssh", "/home/git/test.git/hooks"]
   WORKDIR /home/git
   
   CMD ["/usr/sbin/sshd", "-D"]
   

3. Create a test pre-receive script called always_reject.sh. This example script will reject all pushes, which is useful for locking a repository:

   #!/usr/bin/env bash
   
   echo "error: rejecting all pushes"
   exit 1
   

4. Ensure the always_reject.sh scripts has execute permissions:

   $ chmod +x always_reject.sh

5. From the directory containing Dockerfile.dev, build an image:

   $ docker build -f Dockerfile.dev -t pre-receive.dev .
   > Sending build context to Docker daemon 3.584 kB
   > Step 1 : FROM gliderlabs/alpine:3.3
   >  ---> 8944964f99f4
   > Step 2 : RUN apk add --no-cache git openssh bash && ssh-keygen -A && sed -i "s/#AuthorizedKeysFile/AuthorizedKeysFile/g"  /etc/ssh/sshd_config && adduser git -D -G root -h /home/git -s /bin/bash && passwd -d git && su git -c "mkdir /home/git/.ssh && ssh-keygen -t rsa -b 4096 -f /home/git/.ssh/id_rsa -P ' && mv /home/git/.ssh/id_rsa.pub /home/git/.ssh/authorized_keys && mkdir /home/git/test.git && git --bare init /home/git/test.git"
   >  ---> Running in e9d79ab3b92c
   > fetch http://alpine.gliderlabs.com/alpine/v3.3/main/x86_64/APKINDEX.tar.gz
   > fetch http://alpine.gliderlabs.com/alpine/v3.3/community/x86_64/APKINDEX.tar.gz
   ....truncated output....
   > OK: 34 MiB in 26 packages
   > ssh-keygen: generating new host keys: RSA DSA ECDSA ED25519
   > Password for git changed by root
   > Generating public/private rsa key pair.
   > Your identification has been saved in /home/git/.ssh/id_rsa.
   > Your public key has been saved in /home/git/.ssh/id_rsa.pub.
   ....truncated output....
   > Initialized empty Git repository in /home/git/test.git/
   > Successfully built dd8610c24f82

6. Run a data container that contains a generated SSH key:

   $ docker run --name data pre-receive.dev /bin/true

7. Copy the test pre-receive hook always_reject.sh into the data container:

   $ docker cp always_reject.sh data:/home/git/test.git/hooks/pre-receive

8. Run an application container that runs sshd and executes the hook. Take note of the container id that is returned:

   $ docker run -d -p 52311:22 --volumes-from data pre-receive.dev
   > 7f888bc700b8d23405dbcaf039e6c71d486793cad7d8ae4dd184f4a47000bc58

9. Copy the generated SSH key from the data container to the local machine:

   $ docker cp data:/home/git/.ssh/id_rsa .

10. Modify the remote of a test repository and push to the test.git repo within the Docker container. This example uses git@github.com:octocat/Hello-World.git but you can use any repo you want. This example assumes your local machine (127.0.0.1) is binding port 52311, but you can use a different IP address if docker is running on a remote machine.

    $ git clone git@github.com:octocat/Hello-World.git
    $ cd Hello-World
    $ git remote add test git@127.0.0.1:test.git
    $ GIT_SSH_COMMAND="ssh -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -p 52311 -i ../id_rsa" git push -u test master
    > Warning: Permanently added '[192.168.99.100]:52311' (ECDSA) to the list of known hosts.
    > Counting objects: 7, done.
    > Delta compression using up to 4 threads.
    > Compressing objects: 100% (3/3), done.
    > Writing objects: 100% (7/7), 700 bytes | 0 bytes/s, done.
    > Total 7 (delta 0), reused 7 (delta 0)
    > remote: error: rejecting all pushes
    > To git@192.168.99.100:test.git
    >  ! [remote rejected] master -> master (pre-receive hook declined)
    > error: failed to push some refs to 'git@192.168.99.100:test.git'

    Notice that the push was rejected after executing the pre-receive hook and echoing the output from the script.

Further reading

• "Customizing Git - An Example Git-Enforced Policy" from the Pro Git website