JavaNLP Quick Guide for Using CVS

Setting up your computer to talk to CVS

NLP machines

Add the following line to your .cshrc, .tcshrc, or .login (wherever you tend to put this stuff--look at where similar commands go if you're not sure):

setenv CVSROOT /u/nlp/cvsroot

Now log out and log back in again, or type source filename where filename is the file you just edited.

Remote UNIX machines

Add the following lines to your .cshrc, .tcshrc, or .login (wherever you tend to put this stuff--look at where similar commands go if you're not sure):

setenv CVSROOT username@jamie.stanford.edu:/u/nlp/cvsroot
setenv CVS_RSH ssh

Now log out and log back in again, or type source filename where filename is the file you just edited.

If you don't want to type your password in every time you execute a cvs command, you should generate a private/public key pair and put the public key on the nlp machines. Note that the Solaris machines want an RSA1 public key, whereas the Linux machines want an RSA2 public key. You can generate both of them by doing as follows:

On your remote machine, type the following commands:

mkdir -p ~/.ssh
chmod 700 ~/.ssh
cd ~/.ssh
ssh-keygen -C "username@nlp" -f identity -t rsa1
ssh-keygen -C "username@nlp" -f id_rsa -t rsa

After each ssh-keygen command, you'll be prompted for a passphrase; enter one of your choosing or just hit return. It's convenient if you use the same passphrase for both the rsa1 and rsa2 keys, because you can activate both keys at once for a single session with ssh-agent; see below for details. (Using a blank passphrase is officially not recommended and in a moment we'll describe how to set things up so that you only need to enter the passphrase once per session, but nevertheless some people find it much easier to just go with an empty passphrase, and the result isn't that insecure, particularly on a personal computer which you physically control.) This will generate your private key (called identity) and your public key (called identity.pub). Now you have to put the public key on the nlp machines, so type the following commands :

scp identity.pub id_rsa.pub jamie.stanford.edu:
ssh -l username jamie.stanford.edu
mkdir -p ~/.ssh
touch ~/.ssh/authorized_keys
cat identity.pub id_rsa.pub >> ~/.ssh/authorized_keys
chmod 700 ~/.ssh ~/.ssh/*
rm identity.pub id_rsa.pub
logout

Now try connecting to nlp again by typing ssh -l username jamie.stanford.edu. If it worked, you shouldn't have to type in a password.

If you entered a non-empty passphrase and don't want to type it in every time you use CVS, you can use ssh-agent and ssh-add to type it only once per session. Invoke ssh-agent [command] to start your session (an X session or an interactive shell, for example). Then invoke ssh-add, and you will be prompted for a passphrase for every key in your ~/.ssh/identity file. After this, everything that happens in the session under ssh-agent will have access to those keys. So for example:

ssh-agent /usr/local/bin/tcsh
ssh-add

Presto! You can now use CVS (and ssh in general) for this session to your heart's content, without the bother of typing in a passphrase or password at every command.

Windows machines

You should be able to get things to work with any Windows client that speaks standard CVS (providing it can use ssh as the transport protocol). The commonest way to use CVS on windows is to download WinCVS, which is a CVS GUI, and to have it use CygWin SSH with a private/public keypair so you don't have to type in your password. Below you will find step by step directions for WinCVS.

Using CVS (UNIX)

Checking out JavaNLP files for the first time on a machine

Once you've set up your environment variables, you need to get an initial copy of the javanlp files on your computer. To do this, go to the directory you want the files to live and type cvs checkout javanlp. You should see a lot of lines being printed for the files you're getting, and you should end up with a javanlp folder.  You only have to do this the first time you get the files, from now on you'll just update and commit (see below).

Updating your files to get the latest copy of everything

Once you've checked out the files, all you need to do to refresh them to the latest version is to type cvs update -Pd.  This will update all the files in your current directory and below. NOTE: The -Pd option is very important--it makes sure that new directories get added and old directories get removed when you update. You will see a U in front of files that other people have modified since your last update, an M in front of files that you've modified but not yet committed, and a ? by files that you've created locally but never checked in. It's important to update regularly so that you're always using the latest official version of files, and so that you don't end up modifying a file that someone else has also modified.

Committing your changes so other people receive them

When you've made some changes and tested that they work correctly, commit the changes to the repository by typing cvs commit. CVS will go through the current directory on down and commit any files you've modified. It will prompt you to enter a little log message describing the change. Once you've committed, when other people do an update, they'll receive your changes. You should commit fairly regularly, especially when fixing minor issues, but you should never commit until your code compiles and is stable and usable.

Adding new files to the repository

If you create a new file, you have to explicitly add it to the repository by typing cvs add filename. When you do an update, you should see an A in front of the file (before the add it would show a ?). When you next commit, the file will be added permanently.

Adding new packages to the repository

You should put in your .cshrc or equivalent:

setenv CVSUMASK 002

Adding new libraries to the repository

If you have old files you want to remove, it's not enough to just rm them, because cvs will replace them next time you update. To permanently remove files, first rm them, then type cvs remove filename(s). Next time you commit, the files will be deleted from the repository. Note that if you want to remove all the files in a directory, you can also just delete them all and then type cvs remove dir. You don't ever rm the dir, CVS does it for you.

Tracking changes to a file

To determine who was responsible for changes in a particular file (e.g. in case of broken builds), type cvs annotate file. CVS will list the user and revision number for each line in the file.

Merging different versions of files

Creating and maintaining branches

For more information

CVS man page

CVS manual online

Supplemental Notes on Getting Started with JavaNLP

After getting set up with CVS, you may have to follow these additional steps before you can effectively develop for JavaNLP: