subversion

svn commands within a working copy

Add a file/directory
svn add file
svn ci . -m ""
Delete a file
svn delete file
svn ci . -m ""

Use a wild card
svn add *

To kill local directories changes
svn revert
To restore them, rename the directory, update to get the old directory back and then fix.
For fine grain changes only revert individual files
svn revert ./._.DS_Store
svn status
svn revert */Makefile - to revert multiple files using Bash expansion

Check for differences(local). e.g.
svn status
M       helix.cpp
M       helix.h
svn status | grep -v ?
svn diff filename
svn diff -r 133 filename compare current againt revison 133's version
svn status --show-updates
svn status -u shows the latest revision number

svn commands

Moving files
svn cp file file2
svn del file
svn ci . -m ""

Compare working copy to Repository
svn status --show-updates produced
    *     68   j013.js
Status against revision:   69
svn log | head does not get the updated log!
Comparing directories
svn status -v -N non recursive look
svn list -v global
svn log -r 139 -v gave me a list of the files that were committed in that revision
svn log -r 135:139 -vgave me the recent history of file changes given my current revision number 139.

svn log file
svn blame file
svn update does not kill unmodified modified files*

An update or status has the file's state: U for updated, A for added, ? if it has no idea, M for modified, G for merged on update and C for conflict.
In the event of a conflict, one possibility is to throw away the local changes.
svn revert filename
To edit the merge by hand edit the filename and then
svn resolved filename
I used diff tools to see the differences, as context/meaning can be the issue.

List the contents of the repository.
svn ls file:///home/svnaccount/repository
Look at files log
svn log file:///home/svnaccount/repository/trunk/polytrimon/polytrimon.h


Repository Backup
mkdir /media/KINGSTON/repository2008-09-13
svnadmin hotcopy /home/svnaccount/repository /media/KINGSTON/repository2008-09-13/

To restore copy the backup repository to where the repository lives.
svn co file:///home/svnaccount/repository/trunk/base checks out base from the trunk.

subversion server with ssh

Access repository from 10.1.1.7.
svn list svn+ssh://zero@10.1.1.7/home/svnaccount/repository

Create account on server with automatic ssh login so svn+ssh is automatic.

The use of svn+ssh can be minimized by checking out the repository and then using local commands (normal simple svn syntax). e.g.
svn co svn+ssh://harry2@10.1.1.7/home/svnaccount/repository/trunk/polytrimon
Edit polytrimon.h file, check the changes back in.
svn ci . -m ""


To check out my source code tree from home
svn co svn+ssh://zero@10.1.1.7/home/svnaccount/repository/trunk/base base

Creating a repository

Created a svnaccount and svngroup.
At home directory   /home/svnaccount/
$ svnadmin create repository
$ svn import polytrimon /home/svnaccount/repository/trunk/polytrimon
$ mkdir temp
$ cd temp
$ svn co file:///home/svnaccount/repository
$ cd repository
$ mkdir branches
$ mkdir tags
$ svn add branches
$ svn add tags
$ svn ci . -m ""

Changed the svnaccount to have primary group svngroup.
Give permission for other users in the group to read and write.
# chmod -R 770 /home/svnaccount/repository
Tested by having another user who belongs to the svngroup check out the repository, edit a file and check the file back in.

So each client user needs a local account that includes the svngroup. So you need to log in to the server before you can access svn and hence the repository is (hopefully)secure.

Symlinks

Wouldn't it be nice if subversion followed your symlink (if you told it to do so through an option switch) ... Well it doesn't - without some considerable work (externals crap) - How to make svn follow symbolic links?, External Definitions Chapter 3.Advanced Topics

In situations like this I turn to the OS. Since SVN likes all the source in one tree store all source in one tree. Then use a script to copy the code into the other locations.

Here is my bash script export.sh . I first created the /var/www/html directory with writable permissions.
#/bin/sh
# Inside base/p1/comsci/php/html
rsync -avz ../html/ /var/www/html/
rm /var/www/html/export.sh

When changes are made to the source the script needs to be executed.

For projects that share multiple source code - and do not have ownership here is a technique. Have two scripts - one to copy to the target location, another to copy from the target location. Use a common naming convention for the scripts like import.sh and export.sh .

Now to keep the current state use the import.sh to take a snapshot. So from the projects perspective it has a complete picture/copy of source of all the software - under version control!

When it comes time to deploy you are in a much better situation. Firstly you have all your code versioned. If conflicts occur deal with them. Use diff tools.

Note that the issues subversion faces are inherent in how it manages the structure. Each directory has a .svn folder so subversion is not separate from the source. If this directory was engineered to handle multiple repositories which I believe it is not then it would be easy to add the follow symbolic link flag. e.g. Hypothetical
$ svn add -follow directory/file
Alternatively the technique I described should do the job.

Update

The problem for subversion is its lack of support for multiple repositories. Branches hang around, the developer wants to take a copy of the repository - experiment then possibly merge the changes(history etc) back in, possibly with different repositories.

This is on top of the previous problem of repositories not working with each other.

Why Java picked Mercurial for source control

My Setup

• Checking out the code from ~/Desktop
  svn co svn+ssh://zero@10.1.1.7/home/svnaccount/repository/trunk/base

• repository on Linux box

Issues

• Copying a directory under subversion control copied the .svn folder. When I tried to add the new folder it was already under source control! Strange, but when I later checked in, the changes went to the original directory - no uniqueness checks here.
• Case conflict - even though I am on Unix systems. Resolved by copying and renaming a file. error message : "svn: Can't copy" ...
• Case conflict with svn cp, svn del. Needed to rename graphicsImmediateDeferred.cpp to graphicsimmediatedeferred.cpp, needed to check after deleting to copy, hence can not be working remotely without access to server and do this!

Rollback

Better way to revert to a previous SVN revisio of a file? roll back a file.
e.g.
move to svn directory
$ svn blame <file>
svn merge -r 469:465 <file>
check in

PAST

Move back in time, checked out rollback directory to temporary place, removed files for rollback and copied the correct files.

While rollback is often accomplished with merge, this technique is more primitive where you piece together the files - you are in control. For example if you stuff up a merge then you now have another problem. [This technique suggested to me by Andrew Hazon when working with TN backup software - where in place restoration can fail, defensive strategy to live another day]

Here is an example where I rolled back the changes

svn log menucontents.js found revision number
cd ~/Desktop/temp
svn co -r263 svn+ssh://zero@10.1.1.7/home/svnaccount/repository/trunk/base/p1/comsci/javascript/j021
rm -Rf j021/.svn
move to true j021 directory, remove files, including files added to repository
svn del somefile
cp ~/Desktop/temp/j021/* .
svnci - my helper check in function

While this works the issue of lost history and duplicate data in repository is not good. What is the proper way of implementing rollback?