Installation

Prerequisites

ariba depends on libraries that may not be installed on your system:

• ​Boost (version >=1.39)
• ​Log4cxx (version >= 0.10.0)
• ​GMP (no specific version)

Furthermore, you need default development tools - that are most likely already installed on your system - such as gcc/g++, autoconf, automake, aclocal, libtool, liblt-dev ...

Quick Install (If all dependencies are satisfied)

ariba currently build on Linux systems. Our reference platform is Ubuntu 8.04 with the g++ compiler version 4.1 (confirmed to work on Ubuntu releases 10.04, 10.10) Download the latest ariba package from the ​download site:

Extract the archive and change into the project directory:

> tar xfz ./ariba-x.x.x.tar.gz
> cd ariba-x.x.x

Now, configure, compile, and install ariba (if not all libraries needed by Ariba are available on your system, read the section 'Prequisites'). If no configure script is available (e.g. when you checked out an svn version), run the ./bootstrap script first.:

> ./configure
> make
> make install

In case you don't want to install Ariba into your system but to a local place, do:

> mkdir build
> ./configure --prefix=$PWD/build
> make
> make install

Local install (Download library dependencies and install ariba in a local subdirectory)

If the required libraries are not available on your system or don't want to install them to the system you can use the bootstrap_libs script that comes with the ariba package. It will download all dependencies, compile and install them locally, and compile and install ariba:

> tar xfz ./ariba-x.x.x.tar.gz
> cd ariba-x.x.x
> ./bootstrap_libs

This script will automatically download all necessary libraries and compile them inside a new ./libs subdirectory and install them into ./build. In the end the ariba library itself is compiled to ./build. You can safely delete the ./libs subdirectory after the process.

Here is the manual way to go: If you install Ariba locally and have the required libraries also installed locally, you can use a config.site script to make it easier. The config.site file must reside in a folder called share. If your install path is /home/foo/local and you do a ./configure --prefix=/home/foo/local, ariba headers will be installed in /home/foo/local/include, and the ariba library in /home/foo/local/lib. To use a config.site script, create a folder /home/foo/local/share and create a file config.site. Such a file has paths towards required header files and libraries.

with_boost=/home/foo/Libraries/include
test -z "$CPPFLAGS" && CPPFLAGS='-I/home/foo/Libraries/include'
test -z "$LDFLAGS" && LDFLAGS='-L/home/foo/Libraries/lib'

If you now do a ./configure --prefix=/home/foo/local, the config.site will be found and the paths therein used for finding libraries. If you e.g. have multiple libraries distributed in their own include folders, you can also have multiple includes:

test -z "$CPPFLAGS" && CPPFLAGS='-I/home/foo/Libraries/include -I/home/foo/otherlibrary/include'

Running the PingPong Sample

The PingPong binary pingpong is installed in build/bin. It has one parameter, a configuration file. You can find sample configuration files in the etc/pingpongconfig folder. If no configuration file is given, the node will randomly select its NodeID but will not find other nodes. This is because bootstrap modules are selected in the configuration file.

> ./pingpong ../../etc/pingpong/settings_node1.cnf

If this will fail to find the libariba you may have to set the LD_LIBRARY_PATH correctly in your current terminal, or better add it to your .bashrc

> export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/user/ariba/build/lib

When running the pingpong application it will output a large number of log messages and the initiator will wait for other nodes to join. You can start them using the configuration files settings_node1.cnf and settings_node2.cnf. You may need to adjust the configurations files: currently both node1 and node2 try to join the initiator on the local machine. This will only work if you start all instances on a local machine.

Once the PingPong sample is running and the nodes have connected, each node will send out ping messages to every node he knows in the overlay structure every 5 seconds. You can now e.g. test mobility of Ariba and change the IP address of a node, or swith from LAN connection to WLAN. The links established by the PingPong sample through Ariba are mobility invariant and automatically repaired.

Selecting a compiler

As the g++-4.3 compiler is very restrictive when compiling C++ and you will have some trouble with Boost and Log4cxx, we suggest to use e.g. g++-4.1. You then have to compile the libraries and Ariba with this compiler. You can tell Log4cxx and Ariba to use a different compiler using:

./configure --prefix=... CXX=g++-4.1

This will not work in Boost as the configure script is just a wrapper around the Boost Build.System bjam. You can edit the jamfile in the Boost root directory:

using gcc : 4.1 ;

and then build using bjam as described in http://www.boost.org/doc/libs/1_38_0/more/getting_started/unix-variants.html.

Cross-Comiling for Maemo

Ariba runs on Nokia Maemo 4 (tested) and probably Maemo 5. We have tested Ariba on an N810 device. Cross-Compiling is done using ​Scratchbox. Use the ​preassembled Scratchbox version provided by Nokia which will install and configure the complete Scratchbox system automatically.

The Ariba configure will test for Maemo systems. Internally there are a number of special cases where handling on Maemo is different from normal Linux. If you require special handling, do the following in your code:

#include "ariba/config.h"
...
#ifdef HAVE_MAEMO
 // special Maemo handling
#endif

Overview of special configure options

There are several options to configure that are specific to Ariba:

• --enable-debug=yes - for building a debug build
• --enable-profiling=yes - for profiling with gprof
• --enable-logcolors=yes - for colorful logging output
• --enable-doxygen=yes - for generating doxygen documentation (do a make html-local in ariba/docu/doxygen)