Last Updated: Jan 23, 2020
[NOTE: The method described below should still work, but it’s much easier to just use a Heroku buildpack (see this post for details)]
I manage my open source and contract development projects using Redmine.
Since I’m “frugal”, I tend to try to push the free hosting envelope as far as possible. As a result, I have my Redmine deployment on Heroku, my files and attachments on Dropbox and my source on GitHub.
I also like to link to changesets in my projects, which is easy enough to do when you host the source and the Redmine server on the same machine.
Not so easy with Heroku+GitHub …. until now!
By the end of this tutorial, we will have:
- Setup a build system using Vagrant that matches the Heroku hosting environment
- Compiled a statically linked Subversion client
- Added the svn client to our Redmine repository and pushed it to Heroku
- Configured a project in Redmine to fetch changesets from GitHub using Subversion
In 2012, GitHub announced SVN support, which primarily opened the service up to developers who hadn’t given up the centralized development model.
As an added bonus, it gave us an alternative view into the commit logs for our projects without the need for a local copy.
In order to proceed, I’m making the following assumptions:
- You are hosting an instance of Redmine on Heroku
- You are comfortable using Git
- You’ve used a Debian-based Linux distribution in the past
- You’ve used the GNU Build System before
- You have Ruby and RubyGems configured (if not, RVM is a good place to start)
- You are not crazy enough to try doing this from Windows or OSX ;)
- You have a sense of humour and realize the winky above indicates this guide was written for Linux, but could easily be adapted for any OS
Configuring the Build Environment
Heroku (as of November, 2012 at least) deploys applications to an Ubuntu 10.04 x86_64 environment, and we’re going to be statically linking for that environment, so we need to setup a build system that reflects this requirement.
Since Rubygems should already be installed, it can be used to quickly setup Vagrant:
gem install vagrant
Now, let’s fetch a pre-build lucid system (thanks Vagrant!) and initialize it in the current directory:
vagrant box add lucid64 http://files.vagrantup.com/lucid64.box vagrant init
This will create a
Vagrantfile in the current director, which contains configuration information for our build system. In order to tell Vagrant to use the lucid64 instace we’ve downloaded, the
Vagrantfile needs to be edited and the config.vm.box section updated.
1 2 3 4 5 6 7 8
Now that everything is ready to go, let’s start the instance.
Once vagrant has started the virtual machine, we can access it via ssh using:
For the purposes of this tutorial, I’m going to be using the 1.6 branch of Subversion, which still had the subversion-deps packaged separately.
NOTE With the 1.7 branch, there is a script that automates the process, and I’ll likely update this tutorial at some point to use this.
First, we need to get the source:
wget http://subversion.tigris.org/downloads/subversion-deps-1.6.19.tar.bz2 wget http://subversion.tigris.org/downloads/subversion-1.6.19.tar.bz2 tar xvf subversion-deps-1.6.19.tar.bz2 tar xvf subversion-1.6.19.tar.bz2 cd subversion-1.6.19
Second, we’ll install any additional components required to compile successfully:
sudo apt-get install build-essential libxml2-dev
Finally, we’ll configure the build to exclude as much as possible and to produce a static binary:
./configure --with-ssl \ --without-gssapi \ --without-swig \ --without-neon \ --enable-all-static make
Once the build completes, the only file we’re interested in is the
svn client, so we’ll copy that to the
/vagrant directory of our build machine, but first we’ll strip it.
strip subversion/svn/svn cp subversion/svn/svn /vagrant
NOTE stripping the binary reduces the overall size from ~ 12 MB to ~ 4 MB, which is important since we have limited space on Heroku. vagrant@lucid64:~/subversion-1.6.19$ ls -l subversion/svn/svn -rwxr-xr-x 1 vagrant vagrant 12060462 2012-11-21 14:37 subversion/svn/svn*
vagrant@lucid64:~/subversion-1.6.19$ strip subversion/svn/svn vagrant@lucid64:~/subversion-1.6.19$ ls -l subversion/svn/svn -rwxr-xr-x 1 vagrant vagrant 4257472 2012-11-21 14:56 subversion/svn/svn*
/vagrant directory is shared between the host and the guest machine by default, so this gives us access to the Subversion client if we open a new console and access the project folder we started Vagrant in.
Since we’re now done, we can end the ssh session and stop the virtual machine
exit vagrant halt
Adding Subversion to Redmine
Go to the root of your local Redmine repository, and create a
bin folder. Now copy the
svn binary to this folder, add the result to your repository and push to Heroku.
mkdir bin cp /path/to/static/svn bin git commit -a -m "Adding a Statically Linked Subversion to Redmine" git push heroku master
If all went well, when you access the Repositories tab under Administration -> Settings, Subversion should be listed and the version we just uploaded displayed.
Accessing Changesets from GitHub
The final step is to configure an existing project to access a project on GitHub.
First, add a new repository and point it to the GitHub url you would normally use to access the project on the web.
NOTE Leave the Login and Password fields blank unless this is a private repository
Once configured, clicking on the Repository tab should fetch the changesets and show the source tree
NOTE This can potentially time out on a larger project and require clicking on the Repository tab multiple times until all changesets have been fetched and parsed.
NOTE Viewing changeset diffs doesn’t work
I’ve used my personal installation as an example at http://alexbevi-pm.herokuapp.com/projects/redmine-dropbox-attachments.
This method isn’t on-par with a self-hosted solution, but it’s good enough if you don’t have a VPS in your budget ;)