On 08/01/2017 10:41 AM, Jason L Tibbitts III wrote:
One of things I've invested some time towards over the years is
cutting
down on the amount of %if spaghetti needed to share spec files between
Fedora and EPEL. Earlier in the discussion about the mass python-X ->
python2-X move I volunteered to maintain a number of "stub packages" in
EPEL which have the sole purpose of letting packagers uniformly have
dependencies on python2-X. A couple of people told me they thought I
was being facetious, so I wanted to try again and assure everyone that I
was indeed serious.
Here's a rough, in-progress draft:
https://fedoraproject.org/wiki/User:Tibbs/EPELPythonStubPackages
Basically: create some EPEL-specific python2-X packages that do nothing
besides depend on the base RHEL python-X packages.
If anyone can think of a reason why this would be problematic, please
let me know. I'm especially concerned about situations where the
existence of these packages could cause problems for RHEL. I've tried
to be careful about detailing how version and release would be
maintained and how the dependencies will be versioned so that RHEL can
be updated without the EPEL package being immediately updated to match.
A few comments:
"Since in RHEL it's possible for one package to have many different
versions in various channels, it makes sense to try to stick with the
lowest version that's in any channel. In any case, EPEL is limited in
what channels it supports. "
I don't think we have any way to find out the version of a package in
all channels. At least I don't know of such a way. So, I think we should
concern ourselves with only the channels that EPEL says it will try and
not conflict with.
Note that there is a case here if rhel and epel have a source package
with the same name, the epel one will be used in all buildroots no
matter what the versions. So, yeah, we would need to make sure and
retire the epel package as soon as rhel started providing a source
package with the same name. If they are still different source packages
then the newest one would win.
I'm in favor... lets give it a try with some of the common ones. :)
kevin