Archive for November, 2007

I was planning on simply republishing my previous svn article, but realized that it sucked compared to what I know now.

Prerequisites

I’ll presume you have the following things.
– a Linux machine
– subversion already installed

Terminology to Know

There are a few terms that get mangled if you’re coming from other types of source control. This is just to clear things up.
– Repository: the central storage place on the subversion server where your data is kept.
– Checked-out copy: Unlike VSS, saying something is checked out does not imply that the file is locked. Also referred to as a local copy; but bear in mind that it doesn’t contain *all* of the data of the actual Repository.
– commit: save the changes you made locally to the repository.

Mental Hurdles

All of the files stored in a subversion repository are stored in a meta-filesystem. Much like an ISO image sitting on your desktop is not simply a folder full of files, a subversion repository is not directly accessible when you open it. Instead, you’ll see the guts of the repository- DB files, hooks, locks, etc. Don’t go digging through there to manually change your files- it’ll break things.

Another important one is the meta-filesystem. The “inside” of your repository is just a big filesystem. Much like /bin and /home exist on a linux machine by convention, there are certain base-level directories you should create for convention’s sake. The first thing you should do with a new repository is create three new base level directories: /tags, /trunk and /branches. Your main development will take place in /trunk. Don’t worry about the other two at the moment.

When I refer to the root of the repository, I’ll often refer to it as / or root. This *IS NOT* your server’s root directory or the physical location of the repository- it’s the part of the meta-filesystem where /trunk, /tags and /branches reside.

Step 1: Creating the Repository

Creating a repository is fairly simple. Anyone can create a repository where ever they have write access. All they must do is run

$ svnadmin create ~/project1/


This should create an empty repository. This will demonstrate what I’m talking about when I say that your repository is a meta-filesystem:

morgajel@unicron ~ $ svn ls file:////home/morgajel/project1
morgajel@unicron ~ $ ls /home/morgajel/project1
conf dav db format hooks locks README.txt
morgajel@unicron ~ $ svn mkdir file:////home/morgajel/project1/trunk -m 'creating trunk'

Committed revision 1.
morgajel@unicron ~ $ svn ls file:////home/morgajel/project1
trunk/
morgajel@unicron ~ $ ls /home/morgajel/project1
conf dav db format hooks locks README.txt


Here you can see the repository was empty, then we created a directory called trunk (using a commit message to describe the change we made), then showing that the directory was in fact created. Do the same thing for /tags and /branches.

We now have a working repository!

Checking out the Repository

Creating a repository is fine, but using it would be much more… useful. Next you should check out a copy of your project. Under normal conditions you’ll only be checking out the /trunk, but you mileage may vary in different situations. Since this is on our local machine, we can use the file:// protocol. Other protocols exist, like http:// (webdav), svn:// (svnserve), and svn+ssh:// (svn over ssh), but you don’t have to worry about them right now.


morgajel@unicron ~ $ svn co file:////home/morgajel/project1/trunk my_project1
Checked out revision 3.


This should check out the empty /trunk directory to a local folder called my_project1. The only thing in this directory is a hidden .svn directory which holds the guts of your local copy repository info. It’s similar in function to the CVS directory in CVS. Unfortunately a nearly-empty directory isn’t much use, so let’s add some content.

Adding Content

So let’s add some content.

morgajel@unicron ~ $ cd my_project1/
morgajel@unicron ~/my_project1 $ mkdir -p lib bin share/docs
morgajel@unicron ~/my_project1 $ touch configure Makefile share/docs/README lib/foo.pm bin/widget.pl
morgajel@unicron ~/my_project1 $ svn status
? configure
? share
? lib
? bin
? Makefile


Here I created a bunch of directories and created some empty files. When I ran svn status, svn told me that there were 5 things it wasn’t versioning. Let’s add them.

morgajel@unicron ~/my_project1 $ svn add configure share lib bin Makefile
A configure
A share
A share/docs
A share/docs/README
A lib
A lib/foo.pm
A bin
A bin/widget.pl
A Makefile
morgajel@unicron ~/my_project1 $ svn status
A configure
A share
A share/docs
A share/docs/README
A lib
A lib/foo.pm
A bin
A bin/widget.pl
A Makefile

As you can see, it recursed down into share, bin and lib and added all the goodies inside of each directory. You can also see svn status shows these as well. Keep in mind they’re just slated to be added to the repository- they’re not added yet. Let’s go ahead and commit them.


morgajel@unicron ~/my_project1 $ svn commit -m "a bunch of empty files and directories"
Adding Makefile
Adding bin
Adding bin/widget.pl
Adding configure
Adding lib
Adding lib/foo.pm
Adding share
Adding share/docs
Adding share/docs/README
Transmitting file data .....
Committed revision 4.


Modifying Data

So suppose you’d like to modify these files, you you decide to move the README to the root of your local copy (~/my_project1/):


morgajel@unicron ~/my_project1 $ svn mv share/docs/README README
A README
D share/docs/README
morgajel@unicron ~/my_project1 $ svn stat
D share/docs/README
M bin/widget.pl
A + README


Notice that I used svn mv to move files rather than regular old mv- That’s to make sure svn is aware of the move and keeps the file history associated with the new file. You can also see bin/widget.pl now include some new info as well, and displays an M[odified] next to it. The + next to README shows that it copied the history over from it’s previous position. So what happens if we move a file without svn mv?

morgajel@unicron ~/my_project1 $ mv configure config
morgajel@unicron ~/my_project1 $ svn status
? config
! configure
D share/docs/README
M bin/widget.pl
A + README
morgajel@unicron ~/my_project1 $ mv config configure

You can see that svn panics(!) that configure has gone missing, and sees this new file called config that it’s currently not revisioning. It doesn’t know that they’re the same file.

Commit Messages

You’ve seen me use the -m flag a couple of times now- I’m using it to keep things flowing. If you don’t use it, you’re prompted in your favorite $EDITOR to create a commit statement, which includes the list of modified files. Using the -m flag is useful if you’re scripting commits (I use this when dumping and committing a nightly config file from our load-balancer).

Most of the time however, you’ll use your Editor. Make sure to keep your commit messages sweet and to the point- other’s will see them.


morgajel@unicron ~/my_project1 $ svn commit
[vim shows up, I enter the following text]
Small changes to demonstrate movements
- moved the README
- added shebang to widget.pl
[save and exit vim]
"svn-commit.tmp" 8L, 190C written
Adding README
Sending bin/widget.pl
Deleting share/docs/README
Transmitting file data .
Committed revision 5.
morgajel@unicron ~/my_project1 $ svn update
At revision 5.
morgajel@unicron ~/my_project1 $ svn log
------------------------------------------------------------------------
r5 | morgajel | 2007-11-20 15:55:21 -0500 (Tue, 20 Nov 2007) | 4 lines

creating trunk
------------------------------------------------------------------------


You’ll notice that revisions 2 and 3 aren’t listed- if you’ll remember correctly, they were used to commit the /tags and /branches directories. svn log only reports changes that affect the current target (in this case, ~/my_project1 which is a local copy of /trunk).

There’s a couple more tips and tricks I could go on about- if there’s any interest in this post maybe I’ll write some more about more advanced topics.