Description of problem: qemu-2.4.1-3.fc23 was pushed to stable updates on December, then shortly afterwards 2.4.1-2 was pushed to stable. As a result, a distro-sync wants to downgrade to 2.4.1-2. Version-Release number of selected component (if applicable): qemu-2.4.1-2 qemu-2.4.1-3
Should read "pushed to stable updates on December 14". See my comment in https://bodhi.fedoraproject.org/updates/FEDORA-2015-b2e8518b8e .
In this case bodhi should never have allowed qemu-2.4.1-2.fc23 to go stable
So basically a +1 of -2 triggered the autokarma threshold and queued it up for stable two days after -3 was pushed. Bodhi needs to be smart enough to handle this scenario by either skipping the tagging, or ejecting the update from the push entirely.
Something similar happened again - libpng-1.6.19-1.fc23 was downgraded to libpng-1.6.17-3.fc23 after the latter was submitted for stable. See https://bodhi.fedoraproject.org/updates/FEDORA-2015-4ad4998d00 (libpng-1.6.17-3.fc23) https://bodhi.fedoraproject.org/updates/FEDORA-2015-9199a1bfe1 (libpng-1.6.19-1.fc23) Dennis: please fix. Thanks.
Same thing just happened with bzip2 - the packager submitted an older version for stable and it downgraded a newer version. The version in stable is now -18. https://bodhi.fedoraproject.org/updates/FEDORA-2015-4a9c774398 (bzip2-1.0.6-18.fc23) https://bodhi.fedoraproject.org/updates/FEDORA-2015-be3a6f6ed8 (bzip2-1.0.6-19.fc23) Dennis: please fix. Thanks.
Some good discussion about what exactly happened is here: https://lists.fedoraproject.org/archives/list/test%40lists.fedoraproject.org/thread/K5X7652MWYS7NGOXTMQOLF57XPGP2Y25/
I opened this issue : https://github.com/fedora-infra/bodhi/issues/760 with, dnf list extras, I got bzip2-1.0.6-19.fc23 on the list .
A potential fix has been proposed. https://github.com/fedora-infra/bodhi/pull/768