Working with SVN

From Inkscape Wiki
Revision as of 07:38, 6 September 2006 by Luci (talk | contribs) (+ by)
Jump to navigation Jump to search

(This began a copy of WorkingWithCVS; bear with us as it is updated for SVN)

Subversion (SVN) Basics

This node discusses the basics of using SVN.

For information on more advanced usage, see:


The Repository

SVN stores source code in a shared repository (in our case on Sourceforge's server). The repository contains all past and present versions of the code, and is shared by everyone.

Your Working Copy

To work with the source code, SVN requires you to check out a working copy. This copy is private, so you can make and test any changes you like without disturbing anyone else.

If you have write access to the repository, when you are finished making your changes, you may commit your changes to the shared repository, making them available to everyone.

Alternately, you may generate a file containing the changes you made (a diff), and send it to a developer with write access to be incorporated.

You can check out as many working copies as you want; they take up only your own disk space, and they are completely independent of each other.

If you no longer need a working copy, you may simply delete it.

Getting Started

Setting Up

How Do I Check Out a Working Copy?

To check out a copy of Inkscape from the SVN repository, you may use the following command:

svn checkout inkscape

Here are what the options mean:

"checkout" ("co" means the same thing) specifies the action to take (check out a working copy).

"" is a repository URL; in this case, it refers to the "main" development branch for Inkscape.

"inkscape" is the name of the directory to check out into

Now What?

You should have a complete working copy in the inkscape directory. You can cd to it and try compiling.

Note: If it is a fresh checkout, you will need to run the ./ shell script at the top level to create files needed to compile

When you're inside the working copy, you won't normally need to specify the repository URL, because SVN remembers where the working copy came from.

Note that a Subversion command only applies to the current directory and (possibly) any subdirectories. Normally (particularly for updates, diffs, and checkins), you will want to run the command from the top level of the project.

Bringing Your Working Copy Up-To-Date

Your working copy will not automatically include changes others have made to the repository since you checked it out.

To update your working copy, use this command:

svn update

Dealing with Conflicts

If you've made changes to your working copy, what happens if you update after someone has commited changes to the same files?

Normally, SVN can work this out on its own. Sometimes you have to help it along, though. If SVN says there were conflicts, look for which files have a C next to them in its progress output. Some of those may have unresolved conflicts. You can also search for '=======' in the files themselves to find any unresolved conflicts.

<<<<<<< filename
The changes in your version will be here
The changes by the other person will be here
>>>>>>> some version here

Sometimes you will keep one set of changes and discard the other, and sometimes you will combine the two. It will require a judgement call on your part either way. Talk to the person who made the other set of changes or ask on the mailing list if it's unclear what's going on. Remember that SVN is no substitute for communication.

Also, always make sure to build and test after an update to make sure that the combined changes work as intended!

Generating a Diff

(Always update before generating a diff!)

From the top-level directory, run the command:

svn diff -x -u3 > filename

-x -u3 specifies which format to use (a unified diff with three lines of context). This is the recommended diff format.

This will create a file describing the changes you have made, though if you have created new files as part of your changes, you will need to include those separately too when emailing them to the list or another developer for inclusion.

Automatic Commit Diff

[ fixme: how to do this in SVN? ]

To always generate a diff when you commit to CVS, set the environment variable "CVSEDITOR" to the following script:

# get rid of redhat 9 locale ugliness
unset LC_CTYPE
# Turn on ro access (to stop multiple locks) for local roots
grep @ CVS/Root >/dev/null 2>&1 || export CVSREADONLYFS=1
# Tell the user what's going on
echo Building diff...
cvs diff -up 2>/dev/null | perl -ne 'print "CVS: $_" unless /^\?/' >> "$1"
exec ${EDITOR:-vi} "$1"

If the above script is named "~/bin/cvsdiffvi", then assuming you run the bash shell, you can add the following your .bashrc file:

export CVSEDITOR=~/bin/cvsdiffvi

Now, every time you commit, the entire diff will be visible to you as you write your commit comments.

Applying a Diff

A diff file can be applied to the codebase using the 'patch' command.

If the diff was made as inkscape/src/thingy.cpp and you're in the "inkscape" directory, you'll want to use the command:

patch -p1 < thingy.diff

Depending on how many path elements are part of the diff, you'll need to strip them with -p1, -p2, etc.

Committing your Changes

(Always update before committing!)

If you have write access to the repository, you can commit your changes like this:

Before you commit, please make sure that you describe the changes you have made in the ChangeLog file at the top-level directory.

svn commit -m "Your description of your changes goes here"

Tip: even if you are committing directly, you still might want to generate a diff. It's a good reference to look at when you're filling out the ChangeLog, since it will remind you of all the changes you've made.

Adding Files to the Repository

To add new files as part of your commit, you will need to run:

svn add files and directories to add go here

Before you commit. This also goes for new directories.

The actual addition or removal will not take effect until you commit.

Removing Files from the Repository

svn remove files to remove go here