I don't understand why this system of release-specific overrides which point to a release-specific background file is necessary. Shouldn't '/usr/share/backgrounds/default.xml', as in the
mate-rhel.gschema.override file, work just fine so long as we make sure the right backgrounds packages are installed?
This does indeed seem installable without conmon, which is the main issue here. No idea if it works, but it's better than before. :)
The test failures here are caused by an epoch issue (the first Fedora 30 Branched compose has a ceph build with epoch 2; the package's epoch was subsequently dropped to 1, which is not valid, so this build is not considered an update to the package currently in Fedora 30 stable and does not install during the tests, which openQA correctly flags as a failure). The epoch must be bumped back to 2. I am running a build with the epoch bumped back to 2 at present and will edit the update when it's done.
The openQA test failures here are caused by the updated gnome-boxes not being installed when it should be. It's not being installed because of a dependency issue involving libosinfo. The problem is that a new libosinfo in in the
f30 tag (and hence the buildroot, and so gnome-boxes was built against it) but has not been in a successful compose, so it is not in the F30 repositories for the openQA test to find.
Updates are tested as, and should be, independent units. Any update in updates-testing might be pushed to stable at any time; it should always be in a state where doing so would not cause problems.
This update should not have been created independently while that other update was still in testing. Either the new bind build from this update should have been edited into that update, or this update should not have been submitted before that one was pushed stable.
OK, so yes, there is a problem. This build comes with soname bumps. It provides
libisccfg-export.so.163. The current stable build - bind-9.11.4-13.P2.fc28 - provides
libisccfg-export.so.160. This update breaks the deps of anything built against bind - including dhcp-client - and it should not go out as-is. The soname bumps should be avoided for a stable release update if at all possible, if not possible, all dependent packages must be rebuilt and included in the update.