#416 Errors in fedora-ci.koji-build.installability.functional due to dnf5
Closed 2 years ago by mvadkert. Opened 2 years ago by jcheca.

A recent installability test is failing in the prepare step with the following error:

Preparing the system for testing...
Ignoring "--fixrepo" option as it has no effect on Fedora profiles
Enabling buildroot repository...
Unknown argument "config-manager" for command "dnf5". Add "--help" for more information about the arguments.

yeah well, @msrb another problem to address for dnf5 :(

btw if this is a plugin, one need to install dnf5-plugins now for Rawhide

@msrb PTAL once you have time, I will be now out for 3 weeks :(

I disabled installability testing in Rawhide for now. It's not just the config-manager plugin that is missing, but also repository-packages:

Unknown argument "repo-pkgs" for command "dnf5". Add "--help" for more information about the arguments.

I added comments with my findings to the upsteam dnf issue tracker. We will also need to track this downstream as these problems are going to kick us in the butt in CentOS Stream 10.

Tracking internally as OSCI-5303.

For now, can you change the test to call dnf-3?

I initially reported this here, not sure if that's even the correct component though ...

Discussing zuul/rpm-install-test issue on https://pagure.io/releng/issue/11541

The issue is closed as resolved. Is there anything else that is still broken?

The report is for installability pipeline in Fedora CI, that seems to be disabled now.

@msrb we finally workarounded all the problems around dnf5 in Testing Farm.

What is needed to get installability back?

Sorry, this might take some time to fix properly on installability side. We have this tracked as OSCI-5303.

Here are the problems that we were hitting with installability and dnf5:

Note this was just the first level where things visibly blew up. It is possible that we will uncover more issues once we move pass this first level of issues.

I also suspect that we will hit some issues with rpm scriptlet problem detection code as it relies on specific output formatting:
https://github.com/fedora-ci/mini-tps/blob/main/mtps-pkg-test#L124-L125

The output from dnf5 seems to be a little bit different (?)

Why don't you use dnf 4 instead?

Thanks @astepano for looking into this.

Since dnf5 changes will be reverted in Rawhide, I've re-enabled the installability pipeline again.

Closing this, as this was resolved. More improvements are tracked in #448

Metadata Update from @mvadkert:
- Issue status updated to: Closed (was: Open)

2 years ago

Log in to comment on this ticket.

Metadata