Marek Blaha

Setting the `dbo.conf.arch` is an API equivalent of using `--forcearch=ppc64le` option in dnf CLI. The thing is that dnf needs to know the architecture (e.g. for variables substitutions in repository...

You can find integration tests in this repo: https://github.com/rpm-software-management/ci-dnf-stack Currently this patch causes these scenarios to fail: - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/comps-group.feature#L57 - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/comps-group.feature#L78 - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/comps-upgrade.feature#L102 - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/comps-upgrade.feature#L128 - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/comps-upgrade.feature#L230 - https://github.com/rpm-software-management/ci-dnf-stack/blob/1cccd09a11ec7bc2740c6cf4395bdf22523aaa48/dnf-behave-tests/dnf/history-rollback-group-upgrade.feature#L74 -...

Looks like modularity on ryzen machine (I do not have such a hardware, I just tweaked is_ryzen() to return always true) started to misbehave: ``` $ dnf module list Modular...

True: ``` # dnf install acpi Error: Problem: conflicting requests - package acpi-1.7-11.fc30.x86_64 does not have a compatible architecture ```

Thanks for the pull request! I'm afraid that this PR only hides the symptom. I'm really curious what the root cause of the issue might be. Are you able to...

There is only ordering setting `After=network-online.target` in dnf-makecache.service but according to systemd.unit(5) this does not implicate the requirement: ``` Before=, After= ... Note that those settings are independent of and...

I'm not sure that using --latest-limit with --duplicates does make sense. --duplicates is supposed to return packages installed in multiple versions, --latest-limit limits the number of versions taken into account...

Do we have any use case which requires using these two options together? In that case we need a test for it.

Blocked to prevent accidental merge. I will do more investigation whether this change of repoquery behaviour does not affect other use cases.

I investigated the whole history of repoquery command and latest-limit filter was executed after --whatXXX from the very beginning. It was never moved because of some user request or bugfix....