Version conflits

[Andreas Hanke]

Pascal Bleser schrieb:
>> The command is going to install 9 packages:
>>   1. rpm-4.4.2.43-40 at i586
>>   2. rpm-python-4.4.2-40 at i586
>>   3. smart-0.41-23.20 at i586
>>   4. libsvg-0.1.4-140 at x86_64
>>   5. libsvg-cairo-0.1.6-16 at x86_64
>>   6. python-cairo-1.0.2-14 at x86_64
>>   7. python-gtk2.8.2-21 at x86_64
>>   8. python-numeric-24.2-140 at x86_64
>>   9. smart-gui-0.41-23 at x86_64
> 
> 
> You must be picking the wrong package for your distribution or
> something, because there is no way that installing smart-gui requires
> installing rpm. rpm must already be installed on your system.

Look again:

The command is going to install 9 packages:
  1. rpm-4.4.2-43.4 at i586

Looking closer:

rpm at i586

It resolves the really existing dependencies by installing rpm at i586 on
an x86_64 system. rpm might be already there, but rpm at x86_64 and not
rpm at i586, which it is trying to install.

The dependency chain starts at the point where it picks smart at i586
instead of smart at x86_64. I think we just found where the people with
double-installed rpm come from!

Luckily, newer smart packages do not contain rpmhelper.so any more. This
should hopefully break the dependency chain:

rpmhelper.so is linked to _rpmmodule.so, which makes smart at i586 depend
on rpm-python at i586

_rpmmodule.so is linked to librpm-4.4.so, which makes rpm-python at i586
depend on rpm at i586

=> BOOM.

Without rpmhelper.so, smart at i586 depends on rpm-python at anyarchitecture,
so it should hopefully resolve it by picking rpm-python at x86_64 instead
of rpm-python at i586. While the initial problem - smart at i586 being picked
instead of smart at x86_64 - persists, the result should be less
problematic as before.

Rainer, in general, please try if installing packages like

smart install packagename at x86_64

works better for you.

Andreas