So the test failures here are to do with the lua change, I think. It seems like what happens is:
We set up a buildroot repo and a side repo with this rpm-6.0.1-4.fc45 build in it, and update the system; it correctly updates to this version of rpm and lua-libs 5.5.0-1.fc45
Later in the test we do dnf -y group install base-x and, for some reason, that decides to downgrade lua-libs to 5.4.8-5.fc44 and hence rpm to 6.0.1-3.fc45
The 'check no package from the update has an older version than what's in the update install' step of the test fails because rpm was downgraded
The tricky bit is why does dnf decide to downgrade lua-libs. I suspect it's because something in base-x requires lua 5.4 (probably liblua-5.4.so()(64bit)). dnf has two choices how to solve that:
Keep lua-libs-5.5.0 and rpm-6.0.1-4, install lua5.4-libs-5.5.0
Downgrade lua-libs to 5.4.8-5 (which it can still find in the 'rawhide' repo, currently, as there's been no compose since it was pushed stable) and rpm to 6.0.1-3
...and it picks 2) because it involves installing fewer packages, which dnf's (well, libsolv's) heuristics tend to like, I think.
If I'm right about that it's probably not a genuine problem, because once there's a new Rawhide compose and the older lua disappears from the rawhide repo, option 2 won't be available and it'll have to do option 1. So I think the thing to do would be to re-run the tests after the next Rawhide compose is done and synced, and see what happens.
This update's test gating status has been changed to 'waiting'.
This update was automatically created
This update's test gating status has been changed to 'waiting'.
This update's test gating status has been changed to 'failed'.
So the test failures here are to do with the lua change, I think. It seems like what happens is:
dnf -y group install base-xand, for some reason, that decides to downgrade lua-libs to 5.4.8-5.fc44 and hence rpm to 6.0.1-3.fc45The tricky bit is why does dnf decide to downgrade lua-libs. I suspect it's because something in base-x requires lua 5.4 (probably
liblua-5.4.so()(64bit)). dnf has two choices how to solve that:...and it picks 2) because it involves installing fewer packages, which dnf's (well, libsolv's) heuristics tend to like, I think.
If I'm right about that it's probably not a genuine problem, because once there's a new Rawhide compose and the older lua disappears from the rawhide repo, option 2 won't be available and it'll have to do option 1. So I think the thing to do would be to re-run the tests after the next Rawhide compose is done and synced, and see what happens.
This update's test gating status has been changed to 'waiting'.
I've re-run the tests now, let's see what happens.
This update's test gating status has been changed to 'failed'.
This update's test gating status has been changed to 'waiting'.
This update's test gating status has been changed to 'failed'.
This update's test gating status has been changed to 'waiting'.
This update's test gating status has been changed to 'passed'.
This update has been submitted for stable by bodhi