From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4a) Gecko/20030401
Description of problem:
The redhat-config-* python scripts don't function if the user has set PYTHONHOME.
The redhat-config-kickstart script, for example, explicitly sets PYTHONPATH so
that if the user sets a different PYTHONPATH the script will still function.
On our systems we have a custom python version installed in /usr/local and we
use PYTHONHOME=/usr/local to translate library paths. This breaks
I am able to solve the problem by adding the following to all the redhat python
I think this is a pretty failsafe solution.
Is it possible to have this added officially?
Version-Release number of selected component (if applicable):
Version 8.0 through 9
Steps to Reproduce:
1. set PYTHONHOME to other than /usr
2. start redhat-config-kickstart
Actual Results: Can't find necessary python modules.
Expected Results: Starts up.
msw, does this make sense to add?
This seems to be a pretty uncommon problem, so I'm reluctant to go change all
the redhat-config tools. Even if I make those changes, other Python apps that
do not manually export PYTHONHOME will still break. This makes me think that
this is not the right way to solve the problem.
I am going to close this as 'wontfix' but I will reconsider if we start to see
more of these reports in the future.
Let me try to explain a bit better. Here's the script I'm talking about:
One way to think about this is that PYTHONPATH and PYTHONHOME are coupled. If
you force a specific PYTHONPATH and python version (with /usr/bin/python) you
should also reset PYTHONHOME. Now if you did the following it would be a
> Even if I make those changes, other Python apps that
> do not manually export PYTHONHOME will still break.
Actually that isn't true. Most applications don't force a specific PYTHONPATH
(or override the user's settings) and python version. If they did then they
should reset PYTHONHOME.
It's fine if you don't want to do it but I don't think it is technically correct.
Sorry, one more thing. As a practical matter the support cost of users being
able to hose all of the redhat-config-* tools simply by setting one enviroment
variable is large. The risk of this patch == 0.