Hi,
Thanks for your suggestion and patch for generating an RPM for
Gnuplot.py. I included RPMs with the last release, and I realize that
distutils makes it quite easy. But:
1. For what Python version should I provide RPMs? The RPM is dependent
on the Python version because that determines where the files are
installed (/usr/lib/python1.5, /usr/lib/python2.2, etc). I am not
psyched about having to create RPMs for multiple Python versions, partly
because I don't have all the versions installed myself.
2. Creating an RPM for Python 2.2 under RedHat is not entirely trivial,
because under RedHat the Python 2.2 executable is called "python2", but
distutils calls "python" for one of its internal processing steps.
(Obviously it wouldn't be too hard to work around this problem.)
3. Since there are no standard RPMs for Numeric (that I know of), it is
not clear how to specify that dependency. For example, since I have
installed Numeric from source, I think that your suggested "requires =
_numpy.so" wouldn't work on my system.
4. It is really easy for somebody to download the source tar file and
create an RPM himself, then install the RPM. This gives the benefit of
RPM (e.g., being able to remove the package later) without some of the
problems. (I documented this procedure in README.txt.)
These arguments, and a tiny bit of laziness, led me to skip building RPMs.
How would you answer these questions (particularly the first one)?
Michael
Leonardo Milano wrote:
>Does it make sense to provide RPMS in the sourceforge page?
>The distutils allow for almost automatic rpm building.
>
>
--
Michael Haggerty
mh...@al...
|