You might think that it’s utterly pointless to compile programs from their original, human-readable source code, given how many awesome binary package managers exist in the Linux world. And fair enough: in most cases it’s better to grab something with apt-get or yum (or whatever your distribution uses) than to take the extra steps required to build things by hand. If a program is in your distro’s repositories, is up-to-date and has all the features you need, then great – enjoy it.
But it doesn’t always work like that. Most distros aren’t rolling-releases (Arch-style) so you only get new versions of packages in the stable branches once or twice a year. You might see that FooApp, one of your favourite programs, has just reached version 2.0, but only version 1.5 is available in your distro’s package repositories. So what do you do? If you’re in luck, you might find a third-party repository for your current distro release with the latest version of FooApp, but otherwise you need to compile the new release from its source code.
And that’s not the only reason to do it: you can often enable hidden, experimental or unfinished features by building from source. On top of this, you can apply patches from other developers that add new features or fix bugs that the software’s maintainers haven’t yet sorted out. And when it comes to security, it’s good to know that the binary executables on your machine have been generated from the original developer’s source code, and haven’t been tampered with by a malicious distro developer. (OK, this isn’t a big worry in Linux, but it’s another potential benefit.)
We’ve had several requests to run a tutorial on compiling software, and explain some of the black magic behind it. We often talk about compiling programs in our FOSSpicks section, as it’s the only way to get them – so if you’ve had trouble in the past, hopefully you’ll be fully adept after reading the next few pages. Let’s get started…
Grabbing the source
Although there are various build systems in use in the Free Software world, we’ll start with the most common one, generated by a package called GNU Autotools. Compiling software makes heavy use of the command line – if you’re fairly new to Linux, check out the “command line essentials” box before you get started here, so that you don’t get lost quickly.
Here we’re going to build Alpine, a (very good) text mode email client that works as an ideal example for this tutorial. We’ll be using Debian 7 here, but the process will be similar or identical across other distributions as well. These are the steps we’re going to take, and you’ll use all or most of them with other programs you compile:
- Download the source code and extract it
- Check out the documentation
- Apply patches
- Configure to your liking
- Compile the code
- Install the binary executable files
Alpine is a continuation of the Pine email client of yesteryear. If you search for its official site you’ll see that the “latest” version is 2.00, but that’s ancient – some developers have continued hacking away on it elsewhere, so go to http://patches.freeiz.com/alpine/ to get version 2.20. (If a newer version has arrived, adjust the version numbers in the following commands accordingly.)
The source code is contained in alpine-2.20.tar.xz, so open a terminal and grab it like so:
This is a compressed archive that needs to be extracted. You can use the same command for archives that end in .tar.gz and .tar.bz2:
tar xfv alpine-2.20.tar.xz
(If the file ends in .zip, try “unzip filename.zip”.) As the archive is extracted, you’ll see a bunch of files whizz by on the screen – enter “ls” when the process is done and you’ll see a new directory. Enter “cd alpine-2.20″ to switch into it. Now enter “ls” again to have a nosey around and see what’s inside.
If you see a green file called “configure”, that’s great – you can almost certainly start building the software straight away. But nonetheless, it’s wise to check out the program’s own documentation first. Many applications include INSTALL and README files along with the source code; these are plain text files that you can read with “less”. Sometimes the INSTALL file will contain “generic installation instructions”, with hundreds of lines of boring, non-app-specific information, so it’s a good idea to ignore it. If the INSTALL is short, to-the-point and written specifically for the program in hand, skim through it to see what you need to do.
Similarly, check out the README as well. Alpine only has a README file, but it’s fairly decent, explaining the commands required to build the source code, and listing the binary executable files that will be produced. It’s lacking something important, though: a list of dependencies. Very few programs can be compiled with just a standalone compiler, and most will need other packages or libraries installed. We’ll come to this in a moment.
If you’re new to Linux and the command line, here are some super quick tips. Open a command line via Terminal, Konsole or XTerm in your desktop menu. The most useful commands: “ls” (list files, with directories shown in blue), “cd” (change directory, eg cd foo/bar/, or “cd ..” to go down a directory), “rm” (remove a file; use rm -r for directories), “mv” (move/rename, eg “mv foo.txt bar.txt”), and pwd (show current directory).
Use “less file.txt” to view a text file, and “q” to quit the viewer. Each command has a manual page (eg “man ls”) showing options – so you can learn that “ls -la” shows a detailed list of files in the current directory. Use the up and down arrow keys to cycle back through previous commands, and use tab to complete file or directory names: eg if you have longfilename.txt, entering “rm long” and then hitting tab should complete the filename. It’s a great timesaver.
So, we’ve done steps 1 and 2 – downloading the source and reading the documentation. In most cases you’d want to go straight on the compilation step, but occasionally you may prefer to add a patch or two beforehand. Put simply, a patch (aka a “diff”) is a text file that updates lines in the source code to add a new feature or fix a bug. Patches are usually very small in comparison to the original code, so they’re an efficient way to store and distribute changes.
If you go back to http://patches.freeiz.com/alpine/, you’ll see a list of “Most popular patches” near the top. Click the “Enhanced Fancy Thread Interface” link to go to http://patches.freeiz.com/alpine/info/fancy.html. Along the top you’ll see links to patches for various Alpine versions – so because we have Alpine 2.20, click the link with that number to download fancy.patch.gz.
Now move that file into your alpine-2.20/ directory. You might be curious to have a peek inside the patch to see how it works, but as it’s compressed you’ll need to enter:
Lines starting with “***” show which source code files are to be modified, and the “+” and “!” characters at the start of a line show lines that should be added or changed respectively. So if you see something like this:
char *debug_str = NULL;
char *sort = NULL;
+ char *threadsort = NULL;
This means that the new “threadsort” line in the patch should be added after the first two (which already exist in the original code). But why doesn’t the patch simply use line numbers? Well, it’s possible to do it that way, but then the patch becomes useless if you make even the tiniest change to the file yourself. If you add a line, all of the line numbers in the patch become out of sync, so you need to make a new one. By having a few lines from the original code in the patch, you have some context, so the patch can normally still be applied even if the code has been changed by a different patch.
To apply a patch, you need the (surprise!) “patch” command. You can test the effects of the patch without actually changing the code using the “–dry-run” option like so:
zcat fancy.patch.gz | patch -p1 --dry-run
Here, “zcat” extracts the patch into plain text, and then it’s piped with the “|” character into the patch tool. We use “-p1″ here because we’re already inside the source code directory; if you’re outside it or the patch doesn’t work, try removing it. Once you execute this command, you’ll see lines like:
patching file alpine/setup.c
If it all works, re-run the command omitting the “–dry-run” option, and the changes will be made permanent. Congratulations – you’ve just spruced up Alpine with a new feature!
An example of a typical patch: changed lines of code begin with a ! symbol, whereas new lines have + at the start.
Configuring and compiling
We’re almost ready to compile the source code, but there’s still one more important step: configuration. As mentioned earlier, many programs have features and experimental options that are not compiled into the executables by default, but can be enabled by advanced users. (Why don’t the developers simply include the features, but have a command line switch to enable them? Well, certain features can impact the stability of the overall code, so they’re not compiled in by default until they’re deemed as reliable.)
Enter the following command:
./configure --help | less
This runs the “configure” script in the current directory and spits out its help text to the “less” viewer (that’s a pipe symbol before “less”). Scroll down and you’ll see that there’s a huge list of options you can change: the installation prefix (where it should be installed, eg /usr/local/), where the manual pages should go, and so forth. These are very generic and apply to almost every program you build from source using this process, so scroll down to the “Optional Features” section – this is where the fun begins.
In this section you’ll find features specific to Alpine. The “Optional Packages” section beneath it also has things you can enable or modify using the relevant command line flags. For instance, if you have a command line spell checking program that you love, and want to use it inside Alpine, you’ll see that there’s a “–with-interactive-spellcheck” option. You would use it like so:
Providing “progname” is in your usual paths (eg /bin, /usr/bin, /usr/local/bin) then this should work. Interestingly, many of the options in Alpine’s configure script let you disable things rather than enable them. This is because Alpine is highly portable and runs on many different operating systems, so if you’re compiling it for a fairly obscure Unix flavour you may need to disable some features.
Now, there may be nothing that particularly takes your fancy, so you can run the configure script on its own like so:
But configure also does something else that’s important: it makes sure that your system has everything needed to compile the program. For instance, if you don’t have GCC installed, you’ll see an error message saying that you don’t have a compiler. On Debian and Ubuntu-based systems, a quick way to get the basic packages required for compiling software is to install the “build-essential” package. On Debian (you’ll be prompted for your root password):
su root -c "apt-get install build-essential"
And on Ubuntu (you’ll be prompted for your normal user password):
sudo apt-get install build-essential
Now run ./configure again and see if any other error messages come up. This is where it can get a bit messy, thanks to the complicated world of dependencies – that is, external libraries that the program depends on. For instance, on Debian we got an error of “Terminfo/termcap not found”. But 20 seconds of Google searching for that error message provided a solution: we need to install “libncurses5-dev”.
A similar error popped up for PAM, which we resolved by installing “libpam-dev”. Ultimately there’s no magic way to solve dependency issues, but you can usually get by with the README/INSTALL files, Google and “apt-cache search” to find packages with names relating to the error messages (you usually need ones that end in -dev to compile programs from the source).
Once the configure script has run without any hitches, enter the most important command of all:
This does the compilation job, and depending on the size of the program, it could take minutes (a small command line tool) to hours or days (LibreOffice). So grab a cuppa and check back in periodically. Once the compilation is complete, you’ll need to copy the files into your filesystem – this requires root (admin) privileges. So on Debian:
su root -c "make install"
And on Ubuntu:
sudo make install
And that’s it! Start the program by typing its name – eg “alpine”. And enjoy the warm fuzzy feeling of running a program that was compiled on your own machine, with your own patches and options, because you’re no longer a slave to the distro vendors. You can now grab and install programs before someone packages them up. Happy times indeed.
And here it is: our freshly baked, self-compiled Alpine mail client. It’s not much to look at, but take it from us, it’s a very good mailer indeed.
While many programs still use the GNU Autotools approach of ./configure, make and make install (especially those programs that are part of the GNU project), an alternative is becoming increasingly popular: CMake. This does a similar job, but it requires different commands, so we’ll go through them here. As with Autotools-based programs, however, it’s well worth checking out the README and INSTALL files (if they exist) before you do anything though.
Extract the program’s source code and “cd” into the resulting directory. Then enter the following commands:
cmake .. && make
The “&&” here is important, and you might not’ve come across it before if you don’t spend much time at the command line. Basically, it means: only run the following command (make) if the previous command (cmake ..) was successful. So only try to compile the code if the configuration step (cmake ..) went without any problems.
After the software has been compiled, you’ll need to run the “make install” step as root, as described in the main text (using “su root -c” in Debian and “sudo” in Ubuntu).