3.5. Installing software packages from source code.
Main feature of Free Software is the availability of source codes,
we will learn how convert them to executable files and install
them so that you could run them just by typing their names.
We will discuss how to get them using CVS and how to
apply modifications called patches.
Section quote:
- "Talk is cheap. Show me the code."
-- Linus Torvalds
Section contents:
3.5.1. Before you start.
Souce code files we are going to "talk" about are not .src.rpm
packages but source archives like source tarballs (compressed TAR files)
that ends with .tar.gz (or equivalentely .tgz)
or TAR files commpressed with bzip2 having .tar.bz2 suffix.
Put source archive in your home, then uncompress from your favorite
GUI file manager (eg. "right click -> extract") or with
command line tools:
bash$ tar -xvzf filename.tar.gz
bash$ tar -xvjf filename.tar.bz2
bash$ tar -xvzf filename.tar.gz
bash$ cd filename
bash$ ls
INSTALL blah.c foobar.c make configure
bash$ less INSTALL
 Warning
Installing software from source code may confuse your package manager,
specially about dependency, to take over this problem you
could use your distribution specific source code packages
which could be compiled into a binary package then installed usually,
those source packages like
'.src.rpm' files in RPM distributions (eg. Red Hat)
or in Debian the compination of '.orig.tar.gz' , '.diff.gz'
and '.dsc' files.
Another solution is to change installation prefix from '/' to
'/usr/local'.
3.5.2. Compiling.
Steps after extracting files differ from package to another,
but usually we have things in common, inside the extracted folder type:
bash$ ./configure
bash$ make
bash$ su
password:
bash# make install
bash# exit
bash$
If 'make' had failed or interrupted and you want to retry make sure
you have a clean start, just run 'make clean'
(or in some cases 'make mrproper')
before you rerun 'make'.
If this source package contain programs other than the main program
on which it depends, type 'make dep' before 'make',
if this source package provide many programs and you don't
want to wait to compile them all, just type 'make' followed by target name
of the program you want to make. If every thing is done you could remove
intermediate files issue 'make clean'.
You could override default options in the 'Makefile', for example
you could set the C compiler by specifing CC variable or pass options
to it, an unusual use for it could be "make CC='gcc -s'"
which strip debuging information (to have smaller files), another
unusual use (yet useful) to enable optimization with something like this:
make CC='gcc -s -O2 -fexpensive-optimizations -fomit-frame-pointer -mpentium'
here we specify 2nd level of optimization with '-O2'
(you could ask for deeper optimization but more than two are not reliable)
other types of optimization could be used besides those used in the exmaple
like '-f optimize-register-move','-f regmove'
or '-f rerun-loop-opt', you could specify options
that makes float point calculations faster (but less accurate and harder for debuging)
with '-f fast-math', you could specify the target architecture
and CPU with '-march=TARGET' and '-mcpu=TARGET',
Mr. Isam Byasidi told me that it's better to pass compiler arguments through
'XXCFLAGS' variable. If you want to install it in a place differs from
the default you could use something like "make PREFIX='/usr/X11R6' install"
instead of 'make install',
this will case the program to be installed in '/usr/local'
(this trick could be used to have both 'wine' and the CVS version of
'winex' installed, by have each in different tree)
3.5.3. Patching.
Some source packges need a few changes to compile or to provide new features,
instead of downloading a large grand new file having many common things
with what you had downloaded before, developers saves the differences
with 'diff' tool in a file called patch,
you could download this small file and generate the new files by applying it,
the patche could be applied with 'patch -Np1 -i /path/to/file.patch'
or 'patch -b -p0 < /path/to/file.patch' according it was generaated,
this should be done before './configure', this an exmaple:
once less-378 have a problem calculating the length of line having
Arabic letters, Arabeyes.org has a patch to fix this problem,
after you download the original file 'less-378.tar.gz'
and the patch file 'less_composing.patch.tgz'
save them at your home:
bash$ tar -xvzf less-378.tar.gz
bash$ cd less-378
bash$ tar -xvzf ../less_composing.patch.tgz
bash$ patch -b -p0 < less_composing.patch
bash$ ./configure
bash$ su
bash# make && make install
3.5.4. Getting source code through CVS.
Concurrent Versions System (CVS) is heavely used by the community
of GNU/Linux, as you know free software is developed not by a few developers
but thousands of them spread all over the world, they never met in a physical
environment, they meet in the cyber space (as media call it),
CVS tools help the project administrator to manage each small change by
different developers, different versions and take over conflicts of
overwriting each others files, very fast in a small size.
CVS could be used to take a snapshot of the source package,
which mean to download the most recent fresh version as it gets written,
to use CVS you should first declare CVSROOT environment variable like this:
'export CVSROOT=":USER:PASSWORD@HOST/PROJ/DIR"'
where 'USER' and 'PASSWORD' are you account information
as supplyed by the project manager (a anonymous or guest account could be found)
'HOST' is the address or the name of the CVS server, the rest is the project
name and directory, now we could start using it, first login to the server
with 'cvs login' to downlod the project type
'cvs checkout PROJ' this will create a new directory
(use 'cd' to enter it) containing the source files
(with an extra CVS directory, you could delete it if you are not planing
to takepart in developing it) you could compile it usually.
The following example demonstrate how CVS could be used to get and
compile glchess
bash$ export CVS_RSH='ssh'
bash$ export CVSROOT=\
":pserver:anonymous@cvs.glchess.sourceforge.net:/cvsroot/glchess"
bash$ cvs login
bash$ cvs -z3 checkout glchess
bash$ cd glchess
bash$ ./configure
bash$ make
bash$ su
bash# make install
You could get involved in developing it using CVS, after you get an account,
download snapshot, and edit any file then send it back using 'cvs commit'
option '-m' used to describe your modification and '-kb' option is used with binary files,
for example:
bash$ cvs commit -m "More Comments Added to file1.c" file1.c
|
Best viewed with free web browsers
You may get more high quality software
from here for free
Generously Hosted by www.JadMadi.net
|
