I suggest dropping the YUM code here entirely, as librepo (with Python bindings) is available all the way back to EL6.

This logic implies we prefer YUM. If we want to keep YUM code at all (which I think we shouldn't), we should flip this logic so it tries librepo first, and only if that fails, try yum.

You're right. I've deleted 'if yum_available' branch in that point. Nevertheless, dist-repo is still yum-dependent.

Rather than writing a config file and going that route, you can set these options through Python directly with dnf.conf objects. You can see an example of this here: https://pagure.io/releng/blob/master/f/scripts/spam-o-matic

I don't think you mean to use the dnf API for the yum implementation here?

I don't think you mean to use the dnf API for the yum implementation here?

This is more DNF API usage in the YUM functions...

Just a reminder YUM-3 is going away soon:
https://pagure.io/fesco/issue/2064

We shouldn't be using DNF APIs in the YUM function...

I don't think you mean to use the dnf API for the yum implementation here?

I concur. Perhaps this was a merge/rebase error along the way.

This change reverts that part, leaving it in line with the old yum code (apart from using koji.arch and with open().

https://github.com/mikem23/koji-playground/commits/pagure/pr/1117

@mikem The diff I see here looks good to me: https://github.com/mikem23/koji-playground/compare/pagure/pr/1117

At this point, I think it's good to merge.

:thumbsup:

@tkopecek This needs rebasing. It might be better to also clean up the commit history so that the changes are more logically separated and make sense.

Merging this with a manual rebase and two small fixes

• fixing a unit test that broke on py3
• accounting for the license of the included code from yum

Commit 4ad8a0ec fixes this pull-request

@mikem Commit 96d3ee5f was somehow missed during merge.