Introduction à la ligne de commande

CommandLineIntro: InstallingSoftware

Installing Software

Installing software on GNU/Linux is a broad subject because each version of GNU/Linux has its own way of doing things. Most are variations on apt-get (Advanced Packaging Tool), used by Debian, Ubuntu, gNewSense, and related distributions) or yum (Yellowdog Update Manager), used by Fedora, BLAG, and related distributions.  The basic syntax is

$ sudo apt-get install packagename
$ sudo yum install packagename

Several apt-get and yum functions have the same name and act in the same way, but by no means all. When you want to go beyond the simple cases described here, be sure to check the documentation for whichever you are using.

These examples use sudo to remind you that installing software and editing configuration files require superuser privileges. You can either use sudo with each command, or switch to being superuser with the su command. (Remember to exit your superuser session before resuming normal user work.)

There are numerous options to each command. To uninstall a package, use this command.

$ sudo apt-get remove packagename 
$ sudo yum remove packagename

To read repository index files, and update the local package database.

$ sudo apt-get update
$ sudo yum update

To install all available newer versions of packages.

$ sudo apt-get upgrade

To fix broken dependencies, if any.

$ sudo apt-get --fix-broken

The yum command does not have this option. There are other ways to deal with broken RPM package dependencies, but they require more help than we can give you here.

Users can configure multiple package repositories to download from by editing /etc/apt/sources.list as superuser. Be careful. Back up the current file before making any changes.

All types of GNU/Linux allow the user to install software using the source code. For software in Debian-style packages, you can use

 $ apt-get source packagename

The yum utility does not handle source installs.

Compiling from source is especially important for software that is not available in packages, typically because it is too new. You probably don't want to tackle this process unless you know a little bit about how to use GNU/Linux commands and a little about the GNU/Linux file system, but whenever you decide to try out something brand new and possibly unfinished, this is the most common method. If you don't know about commands and file systems, you can easily get lost doing a source code installation. It is better to read up on them first, get comfortable with them, and then return here.

Installing from source works on any GNU/Linux system that has the compiler and related tools and libraries, so it's a good process to know, and it more or less follows this route once you have a source package:

  1. Unpack the archive and cd to its base directory.
  2. Run the configure script ./configure
  3. Compile the software make
  4. Install the software make install

To carry out the second and third steps, you must have compiler tools on your system. Some GNU/Linux systems come with these tools automatically, but others do not. Any system you are likely to use with this book, though, allows you to download the tools you need for free; search for the packages containing gcc and binutils.


Before we start, a word on dependencies. GNU/Linux developers often don't write an application from scratch; they rely heavily on work that has been done previously by other programmers. This is a smart practice, of course, because it saves time, and to aid this process many kind-hearted individuals have made libraries of code that other programmers can easily access and use within their own programs. These libraries are stored in fixed places in the GNU/Linux file system, usually in the directories whose names begin with /lib, /usr/lib, and /usr/share/lib.

If you install an application that requires certain libraries, it's easy as long as you have those libraries already installed on your system. However, if you don't have the required libraries, you need to find them and install them. If the programmers are thoughtful, they will have included information about dependencies in either the README or INSTALL files that you will find in the source directory of the application. Some extremely nice programmers give you both the name and the URL where you can get the necessary bits.

However, if you are installing software on a distribution other than the one it was developed on, you are likely to find libraries packaged quite differently than on the developer's system. In this case you may have to use the trial and error method: try compiling the source, and when you get an error message telling you of a missing dependency, try to install it. If you can't install it using the name given, you may need to ask someone more experienced how to find the appropriate package, or go looking for documentation of your distribution's packaging policies.

Usually, lazy GNU/Linux users don't bother to read these files so they just go through the standard process and find that the configure stage will give an error telling them what libraries are missing. These lazy types (this author included) then find the required bits and pieces online and install them.

However, if you are new to GNU/Linux, I suggest that you read the README and INSTALL before starting any installation process. It will save you time and heartache.

Just remember that although a dependency list might be long, you simply get all the necessary packages and install them one by one, following the same process described in the previous section, until finally you have everything you need for the program of your dreams to install and run.

Next, let's look at the installation process a little more in depth.

Unpack the archive

Most software sources come in the form of a compressed "tape archive" files that usually have a suffix like ".tar" or ".tgz". The GNU tar command can automatically uncompress files ending with a .gz or .tgz suffix (which means the distributor used GZIP compression) but if other forms of compression were used (such as BZIP2 or LZMA) you have to use the appropriate uncompression program to retrieve the .tar file (colloquially known as a tar ball). To unpack the archive, use the tar command:

$ tar zxvf packagename.tar.gz

Where "packagename" in the example above is the actual name of your package that you wish to install. The tar command followed by the parameters zxvf uncompresses a tar.gz file and creates a new directory with all the extracted sources. The 'z' specifies BZIP compression; if the file suffix is ".tgz2", specifiy BZIP2 compression by using a 'j'. Don't worry-- if it fails to extract you just get an error message. You can remove the tar.gz file after it successfully unpacks.

Now you must change your working directory to this new directory using the cd command. Usually the new directory name is the name of the compressed source package minus the compression suffix. For example, if my package really was called newsoftpack-1.0-alpha.tar.gz, then after running the tar zxvf command on it I would be left with a new directory called newsoftpack-1.0-alpha and would type cd newsoftpack-1.0-alpha to enter this new directory. If you are not sure of the name of the newly created directory, type ls.

Run the configure script

Once inside the new directory, we want to start the actual installation process. To do this, most of the time you will need to type the following:

$ ./configure

Properly packaged source distributions usually contain a script that checks for needed utilities and libraries and prepares the source tree for building and installation. In this case, we will assume that it is configure, since it is a very popular choice for such a script. Sometimes, the command you need to use is different. In those cases, look for information in the README or INSTALL file.

In the command shown, by putting a dot and a slash before the name of the script (./configure) you are telling GNU/Linux to execute (run) a script called configure from the current directory (denoted by "./"). The script then does its stuff, checking what kind of a computer you have, what you already have installed, what kind of GNU/Linux you are running, and so on.

One option to configure is particularly common: the --prefix option, which tells configure you want the software installed in a non-default location. On most systems, the default is fine, and it may be where other software expects to find the software or library you're installing. But sometimes you can't install the software into a shared location or you want it somewhere under your own home directory because you know you're the only person using it. To change the directory where the software will ultimately be installed, specify it with --prefix:

$ ./configure --prefix ~/bin/myprogs

The most common problem that will occur at this stage is that the configure script will halt and tell you that some software library that the new software depends on is missing. If you do experience this error, check the README and INSTALL files in case they tell you how to repair the problem, then use a search engine if necessary to find out what software the error message is talking about and where to get it. Then start the installation process again with this new package. This means that an installation sometimes can take days while you search and download all the packages you need. This is one of the great advantages of package management systems such as yum and apt-get: when developers create packages for these systems, they automate the installation of dependencies.

In some cases, dependencies are optional. The configure script actually supports a lot of options. You can see what options your software package supports by running:

$ ./configure --help

Compile the software

Assuming the configure process finished successfully, the next command to type in the installation process is:

$ make

If you have several processors or processor cores, you can use multiple jobs to speed up processing by adding a -j option:

$ make -j3

These commands actually make ("compile") the software for you. You will then end up with a whole lot of compiled files which in total makes up your software. The make process can take a while, depending on the speed of your machine and the size of the package sources you are installing. Running other processor-intensive applications will also slow down the process.

In the second command shown, -j3 tells make to try to run 3 compilation processes simultaneously, which will allow you to utilize processor resources better if you have a dual-core or bigger machine. The number after -j is arbitrary, but a good rule of thumb is the number of processor cores plus one.

As with configure, you may encounter errors during compilation. In such a case, if you can't fix the problem yourself, contact the developer of the software and politely ask for help, explaining your problem very clearly. The Web page explains how to write polite and helpful problem reports. But first, see if there are log files from the configure and make steps. These may give you more information than was presented on the screen, even if you managed to see what was on the screen as it all flashed by. You can also repeat these steps, adding " &> logfile" to the command in order to capture all output in logfile (use a filename that does not already exist). Before repeting the make step you should probably do "make clean"  in order to remove objects made by previous attempts.

Install the software

After make has finished without errors, type the following:

$ sudo make install

This will install the newly created files from your software in the correct locations in your system. This is usually under /usr/local/, though this can be overridden with a configure option, as we have seen. Because software is usually installed in a shared directory that only the root user can write to, you need to start the command with sudo to have permission to add your software. You don't need the sudo if you told configure to install into a directory under your own home directory.

So now you just need to type the name of the application in your terminal window and it should run. If it fails to start, a common remedy is to type ldconfig and then try again. ldconfig updates the system so that your operating system knows that there are new library files present.

Il y a une erreur de communication avec le serveur Booktype. Nous ne savons pas actuellement où est le problème.

Vous devriez rafraîchir la page.