Subversion (SVN) is an alternative to using Concurrent Version System (CVS) for collaborative development, though it has other uses if you develop on more than one machine and wish to keep all your work in a central location. This guide will show you how to setup Subversion with Webaccess via the Apache2 mod_dav svn module. If you already use apache 1.3 you can continue to use that, just change the port that apache2 listens on in its httpd.conf file.
Subversion Book is an excellent resource for information outside the scope of this guide.
- Local root access on the box or be able to su to root.
- A SSH client such as puTTy or SecureCRT (if you are setting it up remotely).
- A plain text editor, I prefer nano
We have to start out by building apache2. Because we are building it with the intention of using Subversion with it, we must build it with Berkeley DB (sleepycat) support, as the Subversion filesystem is actually built as a sleepycat database. It is this that allows for the versioning of files.
# cd /usr/ports/www/apache2 # make install WITH_BERKELEYDB=db42
We make sure that the apache2 aprutil library is known, this file sometimes seems to get “lost”
# ldconfig -m /usr/local/lib/apache2/ # echo "/usr/local/lib/apache2" >> /etc/ld.so.conf # echo "/usr/local/lib/apache2" >> /etc/ld-elf.so.conf
Instruct the Operating System to run apache2 at startup
# echo 'apache2_enable="YES"' >> /etc/rc.conf
Now to build Subversion
# cd /usr/ports/devel/subversion # make install -DWITH_MOD_DAV_SVN
OK, subversion and apache with berkeley db support are now compiled and installed. You have a choice of either creating a single huge repository for all of your projects, or individual repositories for each. I will detail each now.
Create Subversion Home folder
# mkdir -p /usr/home/svn
Create a generic repository format
# mkdir -p /usr/home/svn/default/trunk # mkdir /usr/home/svn/default/branches # mkdir /usr/home/svn/default/tags
Copy some files will need for the web interface
# cp /usr/ports/devel/subversion/work/subversion-1.0.6/tools/xslt/* /usr/local/www/data-dist/
Setup Blanket Access: If you wish to enable htaccess style password login to the subversion repository then use this system. This is the basic access control system, which can be extended to enable per directory access control as well.
# mkdir /usr/home/svn/access # touch /usr/home/svn/access/users
Create the users using the htpasswd utility
# htpasswd -mb /usr/home/svn/access/users username password
Configuring Apache2 for Blanket Access control
# nano -w /usr/local/etc/apache2/httpd.conf
Add the following to the httpd.conf file, this can be at the bottom or wherever, it will create a http://www.myservername.com/svn/ URL which you can then access
## SVN WebDAV Repository Setup
DAV svn SVNParentPath /usr/home/svn/repos SVNIndexXSLT "http://www.myservername.com/svnindex.xsl" # anonymous first Satisfy Any Require valid-user # authenticating them valid ones AuthType Basic AuthName "Subversion Repositories" AuthUserFile /usr/home/svn/access/users
This can also be placed inside a virtual host directive
Extending the Blanket Access control to enable Per-Directory Access control
# touch /usr/home/svn/access/control
If you are going to use a single large repository with all your projects in and you wish to allow and deny some users access to certain parts of the repository, then you can setup the control file like this. In this example I will use users with the names of admin, manager (a project manager), commiter, client. The access rules are inheritted, but I will demonstrate how to override an inheritted value. If someone is not mentioned at all within the tree, then they are denied access.
[/] admin = rw manager = r [/bigproject] manager = rw commiter = r [/bigproject/trunk] commiter = rw client = r [/bigproject/branches] client = r [/bigproject/trunk/manager_notes] client = commiter =
In this file from the very start the admin has full rights to the whole repository, and the manager can read and see all of the projects, but he has full access to the whole of the bigproject folder and the commiter can read the whole of the bigproject folder, all the development is to be kept in trunk so the commiter has full access to that, and the clients might want to be able to get the latest builds of a project, so they can read it. They might also want access to the stable branches, so they are allowed read access to that. Finally, the project manager has some project notes which he doesn’t want anyone else to gain access to, so the client and commiter are set to empty permissions which denys them access to that folder. You can also specify general access rights for all by using the ‘*’ and assigning access rights to that.
Now the same again, but using per directory access control on a multiple repository system
[bigproject:/] admin = rw manager = rw commiter = r [bigproject:/trunk] commiter = rw client = r [bigproject:/branches] client = r [bigproject:/trunk/manager_notes] client = commiter =
Edit the apache2 config file to use per-directory access control. Find the Blanket Access control Location area and add the below lines just underneath the SVNIndexXSLT
# If we are using the Per-Directory Access Control then we leave this uncommented # Access Control AuthzSVNAccessFile /usr/home/svn/access/control
It should look like this
## SVN WebDAV Repository Setup
DAV svn SVNParentPath /usr/home/svn/repos SVNIndexXSLT "http://www.myservername.com/svnindex.xsl" # If we are using the Per-Directory Access Control then we leave this uncommented # Access Control AuthzSVNAccessFile /usr/home/svn/access/control # anonymous first Satisfy Any Require valid-user # authenticating them valid ones AuthType Basic AuthName "Subversion Repositories" AuthUserFile /usr/home/svn/access/users
Building a Single General Repository
Create our single main repository# svnadmin create /usr/home/svn/repos # svn import /usr/home/svn/default file:///usr/home/svn/repos -m "initial import"
Should you later want to divide this repository into project folders, you will need to checkout the whole repository and use the svn move, svn copy and so on commands, which can be found in the Subversion Book
Building a Multiple Project Repository
Make a new repository# mkdir /usr/home/svn/repos # svnadmin create /usr/home/svn/repos/projectname # svn import /usr/home/svn/default file:///usr/home/svn/repos/projectname -m "Initial Import"
Make sure that apache can read the svn repositories# chown -R www:www /usr/home/svn
and make the access control and userlist readable only by apache. All the contents are pretty much encrypted, but you don't want other shell users peeking at them, though they could get through via a php script, but this is the best way to go about it.# chmod 600 /usr/home/svn/access/control # chmod 600 /usr/home/svn/access/users
There, you now have a fully working Subversion Repository. To checkout the contents of the trunk of a project in your repository via the command line tool# svn checkout http://www.mydomain.com/svn/projectname/trunk projectname
this will make a folder called "projectname" in the current folder you are in when you run the command.
Once you have made changes to a project you can commit those changes by changing into the project folder then# svn commit -m "Notes regarding the changes"
And finally to update your copy of the code from the repository, change into the project folder then# svn update
These commands are all for the SVN Command line tool, which is installed as part of the devel/subversion port. If you wish to use the command line tool and not create the files needed for running a repository with apache2, then use# cd /usr/ports/devel/subversion # make install clean
and skip the remaining steps.