Revision Control Systems for Linux
Geoffrey D. Bennett
geoffrey@linuxsa.org.au

- What is revision control?

- How revision control systems such as RCS and CVS can be used to
  assist with system administration and project management

What is revision control?

- Related to configuration management

- Managing and tracking different versions of files

- Useful whenever you update a file

  - Multiple people working on the same file
    - Shared directory (conflicts)
    - Move files around with Email/SneakerNet (multiple versions)

- Concept of a repository

- Going back to previous revisions

- Essential for software development

  - Blaming people
  - Branches/merges

- Find out more: http://linas.org/linux/cmvc.html

Revision control systems for Linux:
- RCS
  - Useful for a single person, small projects
  - Easy to set up
  - Not client-server

- CVS
  - Based on RCS
  - Client-server-ish
  - Useful for larger projects
  - Slightly more work to set up, but far easier to use
  - Unreserved checkout model
  - The most popular open-source version control system

- Aegis
  - Developed in Australia
  - http://www.canb.auug.org.au/~millerp/aegis/aegis.html
  - Big & very powerful, but easy to use -- "the essential process can
    be learned in less than a day"

- Subversion, BitKeeper, Perforce, Cervisia
  - Other open-source and proprietary version control systems that may
    be worth looking at

- Katie
  - Yet another version control system, under development, to be
    described later
  - http://www.netcraft.com.au/geoffrey/katie/

RCS

- Setting up:
  - mkdir RCS

- Checking files in:
  - ci -u myfile
    (-u means that we want an "unlocked" (read-only) copy of the file)

- Checking files out:
  - co -l myfile
    (-l means that we want a "locked" (read-write) copy of the file)

- Repeat.

- Useful commands:

  - See the log of changes:
    rlog myfile

  - Checkout a particular version:
    co -u -r1.2 myfile

  - See the differences between two versions:
    rcsdiff myfile
    rcsdiff -r1.2 myfile
    rcsdiff -r1.2 -r1.3 myfile

CVS

- Setting up
  - Choose a location for your repository
  - Set your CVSROOT environment variable:
    - Plain directory:
      export CVSROOT=/path/to/cvsroot
    - Over ssh:
      export CVSROOT=:ext:username@hostname:/path/to/cvsroot
      export CVS_RSH=ssh
    - Anonymous:
      export CVSROOT=:pserver:anonymous@hostname:/path/to/cvsroot
  - cvs init
  - mkdir /path/to/cvsroot/myproject

- Using:
  cvs checkout myproject

  cd myproject
  vi myfile
  cvs add myfile
  cvs commit myfile
  vi myfile
  cvs commit myfile
  repeat...

- Useful commands:
  - Retrieve any new files from the repository and see what you've
    changed:
    cvs update
    ? = CVS doesn't know about this file
    M = You've made changes to this file
    A = You've added, but not commited the file

  - See the log of changes:
    cvs log myfile

  - See the differences between versions:
    cvs diff myfile
    cvs diff -r1.2 myfile
    cvs diff -r1.2 -r1.3 myfile

  - Get a particular version:
    - cvs update -r1.2 myfile
      and to clear:
      cvs update -A myfile

  - Tag a tree:
    cvs tag mytag

  - Checkout based on a tag:
    cvs checkout -r mytag myproject

  - $Id$
    volatile static const char *rcsid = "$Id$";
    ident

Katie
- Based on ClearCase
- The repository can be mounted as a filesystem
- Based on NFS plus extensions
- Mostly written in Perl, plus some bits of C
- All version of all files can always be seen through the filesystem
  (no need for 'katie diff')
- Directories are versioned
- Config specs
- Demonstration
- Future planned cool features:
  - A make utility that can automatically figure out dependencies
  - Sharing of built files between views