Georg Grabler (STiAT) - blog

Life is like an endlessly recursive fractal of perverse pain and suffering.

Category Archives: Qt

KDE ~/trunk in ArchLinux

In this tutorial, I describe how to build kde trunk on ArchLinux, using several tools, and providing you with configuration. This is a tutorial for KDE developers (or those who want to become a KDE developer), not for end users.

Basic Design

I decided I want the KDE trunk to be built in it’s own “sandbox”, meaning in a completely closed environment. This is done using a chroot, with basically a 2nd ArchLinux installation on your machine.
I do this for one simple reason, it is possible that you need more recent libraries for the kde trunk, and you may not want to mess up your system with other unstable libraries. You of course could also build and install unstable libraries into your home directory, but I like a closed environment better. Also, there have been problems installing kde to a custom directory lately, due to which is part of kdebase-workspace. To avoid this troubles, and even being able to push kde to /usr, I decided for a chroot.

Also, you’re able to do this on machines where you don’t even have root access, even though, you’ll have to change a few parts on the tutorial.

All this boils down to install Arch stable for your private use on your main installation, and a Arch “sandbox” in a chroot.
First Step – chroot environment

There is a great guide how to set up a chroot in Arch in the Arch Linux wiki. It is basically described how to build a Arch32 chroot in Arch64, but with some simple modifications you can also build a Arch64 chroot in Arch64. I’ll anyway describe here how to do it, if it doesn’t work out for you, you may want to read the guide. You’ll find the link in the Ressources section of this tutorial.
First of all, you can set up your arch chroot in any directory, and it must not be /opt, as in my case. This can depend on your partition settings – how ever you like.

mkdir /opt/arch-chroot
cp /etc/pacman.conf /opt/arch-chroot
cp /etc/pacman.d/mirrorlist /opt/arch-chroot
mkdir -p /opt/arch-chroot/var/{cache/pacman/pkg,lib/pacman}
pacman --root /opt/arch-chroot \
--cachedir /opt/arch-chroot/var/cache/pacman/pkg \
--config /opt/arch32/pacman.conf -Sy
pacman --root /opt/arch-chroot \
--cachedir /opt/arch-chroot/var/cache/pacman/pkg \
--config /opt/arch32/pacman.conf \
-S base base-devel sudo ttf-bitstream-vera ttf-ms-fonts
rm /opt/arch-chroot/{mirrorlist,pacman.conf}
cd /opt/arch-chroot/etc
ln -f /etc/passwd* .
ln -f /etc/shadow* .
ln -f /etc/group* .
ln -f /etc/sudoers .
ln -f /etc/resolv.conf .
ln -f /etc/localtime .
ln -f /etc/locale.gen .
ln -f /etc/profile.d/ profile.d
cp /etc/mtab .

There is a rc-script which is used to start the Arch chroot. You can add this to your rc.conf if you like.
You can save this file to /etc/rc.d/arch-chroot

You can also download it from >here<

. /etc/rc.conf
. /etc/rc.d/functions
case $1 in
stat_busy "Starting Arch chroot"
    mount --bind /proc /opt/arch-chroot/proc
    mount --bind /proc/bus/usb /opt/arch-chroot/proc/bus/usb
    mount --bind /dev /opt/arch32/dev
    mount --bind /dev/pts /opt/arch-chroot/dev/pts
    mount --bind /dev/shm /opt/arch-chroot/dev/shm
    mount --bind /sys /opt/arch-chroot/sys
    mount --bind /tmp /opt/arch-chroot/tmp
    mount --bind /home /opt/arch-chroot/home
    add_daemon arch-chroot
    stat_busy "Stopping Arch chroot"
    umount /opt/arch-chroot/proc/bus/usb
    umount /opt/arch-chroot/dev/pts
    umount /opt/arch-chroot/dev/shm
    umount /opt/arch-chroot/dev
    umount /opt/arch-chroot/sys
    umount /opt/arch-chroot/tmp
    umount /opt/arch-chroot/home
    rm_daemon arch-chroot
    $0 stop
    sleep 1
    $0 start
    echo "usage: $0 {start|stop|restart}"
exit 0

As the next step, you must chmod the file to be executable, and you should enter your chroot.

chmod +x /etc/rc.d/arch-chroot
chroot /opt/arch-chroot

Installing the KDE build dependencies

pacman -Sy subversion git bzip2 libxslt libxml2 libjpeg \
libungif shared-mime-info mesa boost dbus \
openssl pkgconfig xine-lib clucene redland \
gpgme hal cmake qt qca libical lcms xorg-server perl-libwww\
automoc4 akonadi eigen taglib soprano strigi qimageblitz phonon \
pulseaudio xorg-server gstreamer0.10-good-plugins

Setting up your KDE Development user

This step is important. You may not want to crash your local kde user and ~/.kde4 or ~/.kde directory. Since this needs environment variables, as well as your KDE install will need them, we create a own user for this.
Make sure your user is added to all groups you need for kde development, in example if you need sound or device plugging. This example includes my current configuration.

useradd -g users -G disk,wheel,games,kvm,dbus,hal,network,video,audio,optical,floppy,storage,power,policykit -m kde-devel -s /bin/bash
passwd kde-devel

Next, you should edit your ~/.bashrc
Add the following pathes and settings. This must be local user settings, if you use ksh or another shell for your user, make sure this settings work!

export QTDIR=$HOME/qt4
export PATH=$QTDIR/bin:$PATH
export PKG_CONFIG_PATH=$QTDIR/lib/pkgconfig:$PKG_CONFIG
export KDEDIR=$HOME/kde

export PATH=$KDEDIR/bin:$PATH
export QT_PLUGIN_PATH=$KDEDIR/lib/kde4/plugins

export DISPLAY=:1


kdesvn-build is a tool to easen installing the KDE repositories, this will easen up things a lot.
kdesvn-build must be configured properly. I provide my current configuration for you to download >here<

The configuration should work out of the box, I’ll describe shortly some of the options I edited.

First of all, the qt-kde git repository is fetched by http, because kdesvn-build has problems with git urls for some unknown reasons.

Also, you currently need to build phonon first form the qt repository, and then from the kdesupport library. I don’t really know why, maybe has to do with the qmake configuration parameters available building qt with phonon. This trick comes from Thiago Macieira, some KDE users on #kde-devel in IRC told me about that it’s currently needed.

Now, you should download the latest kdesvn-build from the official kdesvn-build page, extract it, and copy the kdesvn-build file in the user directory of the kde-devel user.

After you have this, kdesvn-build directory should include kdesvn-build and .kdesvn-buildrc

Now, run kdesvn-build, and ensure that you include the option –no-snapshots the first time you run it. This has the reason that kdesvn-build prefers snapshots on the first checkout, updating just a few files later. The problem is, that the svn snapshots of kde are broken (in example taglib/admin is fetched of 3.5 branch, which is just one of a lot of problems in the -svn snapshots I stumbled over).

Also note, that kdesvn-build obviously has problems building kde to /usr and qt to a custom directory at the same time. I talked Michael Pyne (maintainer of kdesvn-build) about this, but decided I will try to fix the bug myself, due to the reason that I can reproduce the problem. The version at the time writing (1.10) is still broken.

./kdesvn-build --no-snapshots

Afterwards, your ~/ direcotry should look like this:

drwx------ 9 kde-devel users 4096 2009-12-13 19:38 .
-rw-r--r-- 1 kde-devel users 6 2009-12-13 19:38 .kdesvn-lock
-rw-r--r-- 1 kde-devel users 12419 2009-12-13 19:37 .kdesvn-buildrc
-rw-r--r-- 1 kde-devel users 2786 2009-12-13 19:37 .kdesvn-build-data
drwxr-xr-x 2 kde-devel users 4096 2009-12-13 19:27 .config
drwxr-xr-x 10 kde-devel users 4096 2009-12-13 18:59 qt4
-rw------- 1 kde-devel users 2077 2009-12-13 17:58 .bash_history
drwxr-xr-x 3 kde-devel users 4096 2009-12-13 17:33 .subversion
drwx------ 3 kde-devel users 4096 2009-12-13 17:17 .dbus
drwx------ 3 kde-devel users 4096 2009-12-13 17:14 .kde4
-rw-r--r-- 1 kde-devel users 674 2009-12-13 17:13 .bashrc
drwxr-xr-x 7 kde-devel users 4096 2009-12-13 06:22 kde
drwxr-xr-x 19 kde-devel users 4096 2009-12-13 05:23 kdesvn
-rwxr-xr-x 1 kde-devel users 208941 2009-12-09 02:39 kdesvn-build
drwxr-xr-x 6 root root 4096 2009-12-08 20:15 ..
-rwxr-xr-x 1 kde-devel users 182 2009-02-09 22:43 .xinitrc
-rwxr-xr-x 1 kde-devel users 100 2009-02-09 22:43 .xsession
-rw-r--r-- 1 kde-devel users 16 2009-01-28 20:30 .bash_profile

Running kde trunk using Xephyr

Xephyr is a tool for nested sessions which shipping with the XOrg server. In archlinux the package xorg-server includes Xephyr, which was installed in the first chapter.
Switch into your chroot, and start Xephyr:

sudo chroot /opt/arch-chroot
/etc/rc.d/dbus start
Xephyr :1 -extension GLX -screen 1024x768 &;

You could also use a script to do this for you, and put it into your PATH (/bin/sbin) in your chroot environment:

if [ ! -e /var/run/ ]; then
    /etc/rc.d/dbus start
    /etc/rc.d/dbus stop
if [ -e /var/run/ ]; then
    rm /var/run/

/etc/rc.d/dbus start

Xephyr :1 -extension GLX -screen 1024x768 &

Start your KDE session

su - kde-devel

Keyboard Layouts

If you’re using evdev, you’ll very likely experience problems with your default keyboard layout and keymappings in Xephyr, since Xephyr uses XKB.
Now, there’s a simple workaround for this, run as user (in example on your primary system with your primary user)

xmodmap -pke > ~/.xmodmap
mv .xmodmap /home/kde-devel/.xmodmap

Now autostart xmodmap when your kde4 session starts:

su - kde-devel
vi .kde4/Autostart/xmodmap
xmodmap ~/.xmodmap
chmod +x ~/kde4/Autostart/xmodmap


ArchLinux Wiki (Arch32 chroot on Arch64)
KDE Techbase (Getting Started/Build/KDE4)
KDE Techbase (Getting Started/Increased productivity in KDE4 with Scripts)
KDE Techbase (Getting Started/Build/KDE4/ArchLinux)
KDE Techbase (Getting Started/Set up KDE4 for development)

QTiNT Refacturing

Yea, right. Not even the basic features are done, and I’m already refactored the first time. I’ve found some real issues with the former code design, and decided to refactor those parts.

This sadly means, that I havn’t been implementing any additional features. I personally like the code design better now, and it’s easier to handle http connections and parallel processing without a endless-queue effect.

I have completely implemented tree loading and the XML parsing of the tree for now. Actually, the displaying part is missing, even when the classes (TreeModel, TreeItem) are already written. Allthough I decided not to continue there, but do some additional http transfer stuff and parsing for the “basic claims data”, means the data defining things in the client like available severities and so on, which most likely also will include a rewrite of parts of the message frame (making it more dynamically).

oracle-xe on 64bit arch

Phew, what a run. I basically wanted to install oracle-xe with backwards compatibility libraries first. Since this didn’t work out, I installed kvm and installed it using ubuntu. Due to my wireless driver, I wasn’t able to bind ports correctly (or I was able, but I was pissed off by the long boot times :D).

Now, I decided to set up a arch32 chroot, and install oracle-xe there. It worked out quite fine on the 2nd try (made some mistakes on the first try).

So now, I have a nice system – up and running with oracle, still using 64 bit. That’s great, since I also can run some 32 bit apps which I always wanted to run on my 64bit box.

Something new in th Qt-World

Okay, in my last blog-post I told you I was working on a very new project. I called this Project “QTiNT”, based on the product we call “TiNT”, which has some german meaning (only trust in facts, roughly transalted).

So – what’s done so far? I’ve written a HTTP communication interface, my very own XML Parsing class for handling the XML data which I get from the server, I’ve finished my encryption class, which handles encryption and decryption of data, added some fancy configuration dialog for the proxy settings, and within that also the QSettings implementation.

The client basically has only two features: The login, and checking his first data (Views) being logged in. That’s so far not much, but I already prepared the Tree as a next step (Model, TreeItems are there, the HTTP communicaiton and XML parser for this one isn’t ready yet). Yet, it’s still a long ride until anybody could really work with it ;-).

Note: I’m only doing this project for my personal interest, and it will never be public in any way.

QTiNT Screenshot

QTiNT Screenshot

QtWvDialer – Port or KPPP?

Due to the reason that I use a very unstable and annoying dailup hsdpa connection (YeY, thanks to Orange for this, the carriers keep kicking me), I’ve been using QtWvDialer over KPPP.

What are the Reasons for this? KDE does provide a tool?

  • kppp only provides two initialization strings. That’s bad, because my modem requires 3 to switch from UMTS to HSDPA
  • It doesn’t reconnect on connection loss, rather showing an error message (annoying, due to being Orange customer)

Nowdays, I’m thinking on two things. I dislike the current code and structure of QtWvDialer, even the fact it’s still Qt3. I’ve thought about a kind of fork, or refactoring of QtWvDialer, switching it to Qt4, adding some nice things (as systemtray support, i hate windows floating around everywhere of things which should be in the background).

Also, I could just improve KPPP, adding some nice features (in example as many init strings as you like to have), improving the traffic log, improving the pppd error handling, supporting uatomatic reconnects.

Both options are challanging, and would be really great. I think people would love having KPPP improved rather. QtWvDialer is a perfect option for UMTS/HSDPA, but wouldn’t it be nice if a desktop shipped something capable?

How about KPPP shipping pre-set configurations (as init-strings) for companies? I think users would like to contribute there, making the desktop experience in KDE or generally Linux a lot of easier for users, who don’t want to be bugged with internals.

Tags: ,