Add back hardcoded self-protection of dnf

[evan-goode]

Now that obsoletion of protected packages is allowed (#1610), it's best to hardcode the protection of dnf here instead of using the file in /etc/dnf/protected.d. This way, DNF 4 cannot uninstall itself, but DNF 5 can uninstall DNF 4, and likewise (given a similar change in DNF 5), DNF 5 cannot uninstall itself, but DNF 4 can uninstall DNF 5.

Related: https://bugzilla.redhat.com/show_bug.cgi?id=2221905
Related: https://bugzilla.redhat.com/show_bug.cgi?id=2221907

[evan-goode]

Requires: rpm-software-management/dnf#1991
Related: rpm-software-management/dnf5#906

[inknos]

just one small note. could you edit the commit message following the contribution format?
it is usually hard to describe changes like this one in the changelog when you are not the author. Other than this comment, it looks good. Thanks

[evan-goode]

Sure, I updated the commit message and bumped the version to 0.71.1. Pinging @jan-kolarik since he is currently doing the releases.

[jan-kolarik]

Thanks for the ping, the changes look OK to me, just the thing that in the PR, dnf hardcoded protection is added, but it's labeled as "allowing dnf to be removed by dnf5". That's a bit confusing to me, is it really allowed by just these PR changes?

[evan-goode]

Thanks for the ping, the changes look OK to me, just the thing that in the PR, dnf hardcoded protection is added, but it's labeled as "allowing dnf to be removed by dnf5". That's a bit confusing to me, is it really allowed by just these PR changes?

Well, the change in dnf is also needed for that. But I wrote that in the changelog since that is the main effect to the user of this change, and the motivating reason for it. The changelog could say "Added back hardcoded of protection of dnf", but whether the protection is done using /etc/protected.d/dnf.conf or in code is an implementation detail; IMO the changelog should rather communicate the effect and purpose of a change.

[jan-kolarik]

Thanks for the ping, the changes look OK to me, just the thing that in the PR, dnf hardcoded protection is added, but it's labeled as "allowing dnf to be removed by dnf5". That's a bit confusing to me, is it really allowed by just these PR changes?

Well, the change in dnf is also needed for that. But I wrote that in the changelog since that is the main effect to the user of this change, and the motivating reason for it. The changelog could say "Added back hardcoded of protection of dnf", but whether the protection is done using /etc/protected.d/dnf.conf or in code is an implementation detail; IMO the changelog should rather communicate the effect and purpose of a change.

Yeah, you're right. When I was looking at this first, I was quite narrow-minded.