Closed miabbott closed 6 years ago
@giuseppe I think this was you; could you have a look?
@miabbott yes, that is a regression.
Could you check if https://github.com/projectatomic/atomic/pull/1219 solves the problem you have seen?
This is a quick fix to not regress on this, but I think the install data part needs some more work. Unfortunately it is quite difficult to handle correctly as the install.json
file gets out of sync very easily and in many cases I had to use -i
when uninstalling.
@giuseppe That PR worked for me. I did some simple tests with a single container installed and with a mix of non-system/system containers.
We should probably have some more rigorous testing of the various possible scenarios with containers installed, etc.
@miabbott I've added a test case for the reported issue
Our automated tests caught this when trying to uninstall the
cockpit
container:It looks like the uninstall script of the container was successfully run, but it blows up afterwards.
This was found using Fedora Atomic Host Continuous
The affected version was
atomic-1.22.20-6820f8f3b53a0a9fe89690b671624905ce3d0d30.665bce33cf99ea73047097d433cf652b7f2d9adc.fc27.x86_64
Previous good version was
atomic-1.22.17-56c32aeca382f2c68e5078679b7585837cb13716.665bce33cf99ea73047097d433cf652b7f2d9adc.fc27
The diffs between 56c32aeca and 6820f8f, lead me to think it was ba23e7692423dc66446f3c01e745598673ec2c4a that was the guilty commit.