Showing posts with label DGV. Show all posts
Showing posts with label DGV. Show all posts

Saturday, June 19, 2010

Packaging Qt Applications for Ubuntu/Debian

I have recently attempted to get my Discrete Geometry 3D Viewer capable of building Debian packages, so that one may install DGV without needing to compile it or worry about dependencies for other Ubuntu Distros (like the question I got from a potential user).

I believed there would be a substantial discussion on how to do this for Qt applications, but I could only find the Maemo Guide useful. It gets even more difficult if you want to do multiple binaries from a single source. Hence, I have documented my findings of this topic in this blog.

Useful Links that I used:
Complete Ubuntu Packaging Guide
Qt App Maemo Guide
GPG Guide
Pbuilder Howto
Using Local Packages

Firstly, my situation is the following. I have three side-by-side dynamic libraries, dgv-base, dgv-contrib and dgv-vtk. The libraries are divided based on dependencies of Qt, None, and dgv-base & Qt & VTK respectively by design. Then there's the actual DGV application which depend on these libraries.

First the multiple libraries. Begin by renaming the source directory, with the name of the package and the version (with a dash as the separator, this is important). For example
libdgv-0.15

Assuming that your source is in the state you wish to distribute it, create the tarballs so that you have two tarballs
libdgv-0.15.tar.gz
libdgv_0.15.orig.tar.gz
Note the underscore between the library name and the version. This is very important.

Change to the directory of the source and execute
dh_make -e your.maintainer@address -c GPL
where the "-c GPL" assumes you are using GPLv3 license and "your.maintainer@address" is your email address. This will ask you a series of questions, where you should select library if you're doing a library or a single binary if you're doing a binary. I will assume a library for the aforementioned reasons.

This step will create a directory called "debian" with all the Debian package configuration files. Remove the example files, they are not needed for what we are doing as far as I know.
rm *.ex *.EX
Please read these pages of the packaging guide for the remaining files. You need to edit the control and rules files as I have for DGV (also see Maemo guide for a simple single binary example). Things to watch out for is also given in this link (at the end). Fill the changelog and copyright files as described by the package guide, making sure to use the name of the overall package where applicable and to match the author email to your GPG name. If you do not have one, you need to create one using the GPG guide to sign your packages.

For multiple binaries, there is one last step. You need to create the ".install" and ".dir" file for each package. The former has the list of files to be installed, but in a wild card format (e.g. "usr/lib/lib*.so"). The latter is where the files are to be installed (e.g. "usr/lib").Note that there is no "/" in front as per normal Linux directories from root. This is because they will be relative paths and when the real package is built, it will build it from the root. See the debian files I created for DGV here.

Now to build your package, we will use Personal Builder (pbuilder), a way to build your package using only the minimal (initial) Ubuntu base/setup plus your dependencies. This is the best part that makes constructing packages most useful. If your dependencies are correct, then the user just installs the package and Synaptic or apt-get just installs everything you need to get it going. Once a pbuilder environment is setup (which can be made to suit for building on different architectures and Ubuntu distros), the package is built with your configuration and placed into
/var/cache/pbuilder/result
Only downside is that all the base packages will be downloaded from the Ubuntu repository, which could take a while. If you can't use the Internet for whatever reason, you can still build it the packages using
debuild
when within the source directory. You might also want to do this initially to ensure that its all working correctly.

First create the pbuilder environment by
sudo pbuilder create --distribution $(lsb_release -cs) \
        --othermirror "deb http://archive.ubuntu.com/ubuntu $(lsb_release -cs) main restricted universe multiverse"
Then build the descriptor file
debuild -S
If you did a "debuild" already, then this step is unnecessary. Once the descriptor file is present, then build using pbuilder
sudo pbuilder build *.dsc
To build as a different distro or architecture, see this. Hopefully, all works fine and you get a few packages. There might be a few warning from Lintian. Ensure that the descriptions lines are not too long or google the lintian warning. If your warning says "empty-binary-package" then your files are installed in the wrong places. For Qt projections, I install the files of the library into "debian/tmp", then move it out into the relevant package directories by using "dh_movefiles -p$@ -Xcontrib -Xvtk usr/lib" command in the rules (see the wiki). The "-Xitem" tell the movefiles app to ignore the files with the string "item" from the file names.

Finally, on a Live CD or clean install, check you packages by installing them. Pbuilder didn't pickup the fact that I had incorrectly named libvtk5.2 as libvtk5 for one of my packages, so I recommend it.

I then build the dgv application and its all done. Hope this helps. I will post up more on how to use local packages later when I have it working.

Cheers Shakes - L3mming

Thursday, November 19, 2009

DGV 0.1 Stable Released

I have finally released a stable version of my DGV software. You can grab it on Sourceforge.
You can view some tutorials at the Google Code page.

What's DGV for? In my honours, I just wanted an easy, no fuss way to visualise my data (which was a 2D complex scalar field). To my dismay, I couldn't find one and ended up using Excel (I know.... really bad). So I wrote DGV as tool that allowed this. Most times I just wanted a tool view my data quickly. This is DGV's main goal. It has also been supplemented by my current Ph.D project on Image processing.

It has so far clocked over 2000 downloads and has far exceeded all my expectations. As such, it will be supported indefinitely into the future. I will also be adding my FTL Library to it soon, which should propel the software into the forefront of state-of-the-art Convolutions. The next major release will be after I submit my Ph.D Thesis (hopefully around March next year).

Note for Linux uses, the Linux build was built using the latest Ubuntu OS - Koala, so it requires the latest C++ standard libraries. In other words, it should run in any of the latest distros. This may be changed if it is found to be a major problem.

Cheers
Shakes - L3mming

Sunday, August 23, 2009

Discrete Geometry 3D Viewer

I thought I'd write list of features of my Discrete Geometry Viewer (DGV) program. I wrote this as a request from the New Projects section of the Linux Journal. The goal of DGV is to make visualisation of data as simple as possible (not something VTK is renowned for) while being intuitive and responsive.

In DGV, the image is placed into a 3D scene using OpenGL (via the VTK library), so it uses your graphics card to render the image. U can pan and zoom, adjust the gamma and rotate it. U can save what you're viewing or just export the data/image. U can also view the raw data in table form (may not be as useful to non-scientific users). From the table u can do surface and line plots. U can also add, subtract etc. images using the Operate feature. Finally, one can see the histogram of the image, which tells u statistical info on the image (output into console also).

The main feature of DGV is Filtering and Convolutions via Fast Fourier Transforms. This allows one to see the frequencies present within the image and edit them via the table. One can do crude Low-pass (smoothing), High-pass (edge detection) or band-pass filtering. DGV also features surface and line animations.

Future advancements will include:
- Pixel values within viewer
- Saving animations
- Python Shell rather than simple console output. See my project called QPythonShell which allows one to embed a Python shell into Qt applications.
- More file formats
- More transforms, like the Number Theoretic Transforms (via my new Number Theoretic Transform C library).

Links:
Homepage, Sourceforge, Google Code and Qt Apps.org.

Cheers
Shakes - L3mming