Please do not reply directly to this email. All additional comments should be made in the comments box of this bug.
https://bugzilla.redhat.com/show_bug.cgi?id=476460
--- Comment #12 from Mamoru Tasaka mtasaka@ioa.s.u-tokyo.ac.jp 2009-01-25 01:22:32 EDT --- (In reply to comment #11)
Ok, I'm going to jettison start.sh. /var/run/milter, /var/log/milter and create a milter-base package for the python milters that use it. Is there a better way than having a package own one tiny shell script and two directories?
That will remove those objections to pymilter (and start.sh truly is cruft if you don't use it).
- Well, are /var/run/milter and /var/log/milter directories supposed to be used also by other applications than pymilter?
On Fedora there are already some srpms named *milter*. Would you know why none of these packages own these two directories? If you are unsure, I think it is better that these two directories should be owned by pymilter.
I define python because I need to build pymilter for python versions other than the system default. For instance, on EL4, the system python is 2.3, but the production milter requires python2.4. Both python and python modules are easily built to coexist in multiple versions. (Of course, that argues for cutting start.sh again - so it's history.) Would it be acceptable if an %ifdef excludes the python redefinition for fedora?
The version dependency is only redundant if building only for Fedora. Would it be acceptable if I %ifdef away the dependency for Fedora?
- If you want to maintain pymilter also on EL-4, okay.