An extremely ugly thing based on the fact that somehow rpaths
defined for python get propagated as runpaths for any eggs
including those that have explicit profiles.
Actually they override explicitly define rpaths making them useless
to define.
If it is possible to make pycurl link without this commit and/or eliminate
explicit rpaths being overwritten by propagation - it is better to do
this.

separate java-jdk to an own component
add patchelf
patchelf java-jdk's binaries

this is needed for linking tests to pass

gtk interface building resulted in no gtk interface being built for some
reason, and reference working set ups all used qt, so this commit
switches interface to qt and fights with bugs that come from this
change.

QGtkStyle (needed for Proview-R) support in qt4 needs glib.
However, config script for qt4 is written in such a way that if glib is not
requested but QGtkStyle is requested, it just silently turns QGtkStyle off.

On the other hand, if glib is requested explicitly, then if script fails to
enable QGtkStyle it reports an error properly.
I think it was a bug in qt4 error reporting mechanism that made compilation go
smoothly even though requested QGtkStyle was never provided.
Nobody will fix it now though.

Anyway, this commit properly enables glib and fights with everything that rises
up as a result of this, leading to successfull QGtkStyle support.

https:// link just works more reliably as I found after several timeouts.

But I still have no idea why the second db folder spawns right inside
the frst one.

straighetend the configs:
no more custom source packages
using xvfb for proview compilation (easier)
fix bug with pwr pre-build process killing itself