Coders Edge
  
Modern Application 
Development
questions and answers on compiling
Open Source Code
in a Linux environment, but it will work for any Open Source

Other source code

Just for giggles (and with a lot of free time) I've decided to build some other packages of programs that I use quite often. I'm starting with amarok, various screan savers, ardour, libre-office (which is a fork of open office). When I don't have a dependancy usually that means that I do, actually, have the package installed on my system for use by the program that is really installed as a compiled binary rpm package. So what I need to do then is to install the devel rpm to get the headers against which the package I'm compileing can build. The headers describe the interface of the dependancy. The developer of the package we are compiling knows these headers and uses them to access the features of the other package. The code is not, though, part of the package. It is used by the package.

In the real world of software many things are dependant on other things. This is how it is. And things are aliased and wrapped and shimmed and other wise obscured. But what it all comes down to is that software has to be built in such a way that packages don't conflict in the way that they access memory. And if they do things will not act as expected. header files describe the interface of a package. The versions don't need to match as long as the headers have not changed. Many developers, for this reason, create a public shim which they do not change. Meanwhile they do all of the doing behind the scenes, without needing to publish their inner interfaces. This  technique is called shimming. However a shim also does a lot to define the interface to the point that someone else can take that shim, which are essentially just a series of headers that proxy the real interface (behind the scenes in someone else's library). The shim and then be used as a design document without any knowledge of what the source files that define these headers might be and . . . well the other module can be then replaced with a new and redesigned module. That assumes that there is a well-enough understand as to what that other module was doing. Naturally there may be all kinds of things that hte module does that can not be replicated by the knowledge just held in headers. So even if you get your code to build by creating the module from a header it will not behave like the old modules did without intimate knowledge as to the inner workings of those modules.

libreoffice, Ardour, amarok

Alright, I downloaded the libreoffice source rpm, and it was a very large collection of various packages, all of which will need to be untarred, etc, and patched, and then installed. Looks like a lot of work.

I was also wondering, after I do my compile, if I have a bunch of machines with the same OS and processor, then how do I easily create a binary rpm for this? Is the spec file provided that alows me to easily create this for use on other installs? I know that spec files are provided as part of the source rpm, so, one never knows. It is not hard to create rpm's and to destribute them. It is even fairly straightforward to set up a repository and to have systems that automatically update upon finding a newer version of a package. I suspect that hte spec files probably give you a good head start on doing this.

persistence pays off! I have had to add a bunch of devel packages to complete the various dependancies of the packages that I have downloaded and am compiling.

usually I am able to websearch the error string, or a fragment of it, and that will lead me to somewhere that tells me what I am missing. Dependanices may be met as far as cmake is concerned, but if the devel package is not install there will be no headers to build against and the build will fail.

Ah, there is no telling what you learn when you do these tasks. For example I websearched on libaio, because that was a devel package that I did not have, and got a concise explanation about what libaio is and discussions on the opinions of the author of this information:

http://www.morphisms.net/~wkj/software/libaio/index.html

That is how bazzare and obscure the web is. Some guy's left over pages from years ago. Ah, the perils of the Internet. All of your webdroppings are here forever.

success with amarok

In any case I have a compiled version  of Amarok 2.4.1 running now. It isn't much different from the one that was already installed. But is a local build and thus I can customize the build.

What is the point of doing these compiles? Is there anything we can learn from doing them? Can we make the code faster, better, stronger?

Obviously: Of course! And all of these things are controlled by the configuration. And the configuration is created by the qmake  --project or the cmake. After having compiled a version of the thing I can make another version of the same thing by creating a different build directory with different configuration options. Then one could run the various versions as desired.

In the world of development for mega-corporations there is even a deeper level of doing this configuration thing. The system is seen as a whole. The idea that you would build twenty or fifty or more various versions of code testing all of the ins and outs of configuration options seems farfetched. The way things get done at this level is that the specific thing is built within a virtualized operating system. This is a whole system, compiler, everything, all inside of an ISO image saved as a file and copied from disk to disk just like any other file. And these can be cloaned and spawned as copies, and each then modified in some way to create arrays of these almost identical running images.

  A custom of 
   excellent
 software is your
  solution for
   Custom 
   Expert 
   Software



  WP April 4, 2011

My content presented for the mobile user:
Message Mobile
Right Column Mobile
Left Column Mobile
Poem Shards Mobile
Coder's Edge Mobile