• Introduction
• Accessing CVS
• Importing a new CMT package into CVS

Introduction

This page is intended to contain a few important tips on how we use CVS. Especially in collaboration with CMT.

A few general tips on CVS commands may be found here. Though the packages described on that page are now mostly obsolete.

Our main CVS repository is presently at CERN on the CERN CVS server. Our LVL1/l1calo area within that can be viewed via the ViewCVS interface.

Accessing CVS

For full details see the HOWTO on the CERN CVS server web pages. Briefly, there are three access methods: Kerberos, SSH and anonymous pserver of which only the first two seem to work. For the first two you need an account on the LXPLUS system at CERN.

Access via Kerberos

• Ensure that your local machine has the necessary Kerberos IV infrastructure set up. See the CERN web page.
• export CVSROOT=:kserver:atdaq-sw.cern.ch:/local/reps/atlastdaq
• Authenticate to Kerberos. What works best may depend on your site NB the point at which these commands may not work for you is when you do a cvs checkout). Try one of the following three:
  • kinit yourusername@CERN.CH
  • klog -tmp yourusername@cern.ch
  • klog.krb -tmp yourusername@cern.ch
• Note will need to renew your Kerberos token every day.

Access via SSH

• export CVSROOT=:ext:atdaq-sw.cern.ch:/local/reps/atlastdaq
• export CVS_RSH=ssh # Add this to .bashrc
• Set up ssh keys on your local host and copy the .pub file to the right place at CERN. Their website suggests using keys without passphrases. If you already have keys with passphrases you can still keep them and use a separate key file just for CVS. This is what I show below.
• On your local host do:
    ssh-keygen -t rsa1
  NB to get a separate CVS only key file give another name, eg identitycvs, at the filename prompt then give no passphrase. If you use the default "identity" as the filename replace "identitycvs" with "identity" everywhere below.
• Then copy the public identify file to LXPLUS:
    scp ~/.ssh/identitycvs.pub yourusername@lxplus.cern.ch:
• Then create (or modify) ~/.ssh/config containing:
    Host atdaq-sw.cern.ch atdaq-sw isscvs.cern.ch isscvs
    Protocol 1
    ForwardX11 no
    IdentityFile ~/.ssh/identitycvs
  If you already have a "Host *" section in your .ssh/config file, the specific section for isscvs should go above it. NB the L1CaloRelease data/shell directory contains an example config file you can move to your ~/.ssh directory (but you can only get it after making a checkout of course!).
• On LXPLUS use set_ssh which moves some of your .ssh/* files into ~/public and creates links to them:
    /afs/cern.ch/project/cvs/dist/bin/set_ssh
  
• Finally authorise your new public key.
    cat ~/identitycvs.pub >> ~/.ssh/authorized_keys

Importing a new CMT package into CVS

Assume you have created a new CMT package called myPackage with the correct CMT package/version directories in your l1calo working area. You can then use the cvs import command to store it in CVS.

You can do this by hand as described below. But it is easier to use our importpkg.pl script. This script make some simple cleanups of files generated by CMT and creates some suitable .cvsignore files. But after that it imports everything in all subdirectories below the current directory, so you should first remove any files that you dont want to appear in CVS. You may like to make a backup of your package first.

  cd workarea/LVL1/l1calo/myPackage
  export CVSROOT=.... # ask for correct setup
  importpkg.pl myPackage

You can then rename the pre-CVS directory out of the way and check out the newly imported package, eg using our getpkg script.

  cd workarea/LVL1/l1calo
  mv myPackage myPackage.preCVS
  getpkg -h myPackage

Using cvs import by hand

Before running cvs import, it is best to ensure that your package directories are completely clean and only contain files that you really want stored. You should also remove, or tell CVS to ignore, all CMT binary directories such as i686-slc3-gcc323-dbg and the probably empty Visual directory. Most of the files in your cmt directory, apart from requirements and Makefile, are automatically regenerated by CMT and should also not be stored in CVS.

To ensure that these files are ignored by CVS later (or by the import command) it is best to add a suitable .cvsignore file to both your top level package directory and your cmt directory. Standard .cvsignore files tailored for each of those two directories can be copied from our other packages. Or you can look at the section below.

With a clean set of directories, you can import your new package into CVS as follows:

  cd workarea/LVL1/l1calo/myPackage
  export CVSROOT=.... # ask for correct setup
  cvs import -m "First CVS version" l1calo/myPackage start myPackage-00-00-00

This adds the package to directory l1calo/myPackage in the CVS repository and tags it as version myPackage-00-00-00. Remember this will import all files in all subdirectories, except those you have asked CVS to ignore.

The CERN CVS server doesnt report the initial tag properly to our latest script. To work around this bug, you need to tag the package again after its first import.

  cd workarea/LVL1/l1calo/myPackage
  cvs tag myPackage-00-00-01

You can then rename the pre-CVS directory out of the way and check out the newly imported package, eg using our getpkg script.

  cd workarea/LVL1/l1calo
  mv myPackage myPackage.preCVS
  getpkg -h myPackage

.cvsignore files

A file with the name .cvsignore tells CVS to ignore certain files for all commands (update, commit, import, etc) operating in that directory. For working with CMT, a suitable .cvsignore in your top package directory and another one in your cmt directory will make like a little easier.

For the top level package directory, you want to ignore all the likely CMT binary directories, plus a few other types of file. A typical set might be as follows:

i386_*
i686-*
linux-*
powerpc-*
ppc-*
sparc-*
sungnu
sunpro
Visual
*.gz
*.tar
*.tgz

For your cmt directory, you want to ignore all the files which CMT automically generates or updates. Although CMT creates a simple Makefile when the cmt create command is executed, you may want to edit this file, so it probably should be stored in CVS. However many *.sh, *.csh and all *.make should be ignored. A suitable .cvsignore file may be:

checkuses
cleanup.*sh
setup.*sh
uses.log
version.cmt
*.make
*.makesav