Bug 1279185

I hit a strange behaviour, when using dnf with the --installroot option.

To explain it better, let's use yum for this example:

yum update --installroot=/<somepath>

When issuing this command, the following happens:

1) yum uses the repo files located in /<somepath>/etc/yum.repos.d/.*repo
2) yum uses the cache of /<somepath>/cache/yum.... to store the metadata
3) yum uses the cache of /<somepath>/cache/yum.... to store updated nevra
4) yum uses the cache of /<somepath>/lib/yumdb... to store the db entries

With other words, using --installroot does a full context switch (as I explain it) to a different "root". Not touching the underlaying systems yum.repos.d etc.

This has the advantage, that I can use the real host system (the true one in /) to update a different system (one for systems integration, different updates or other related tasks) in --installroot (which of course can have other .*repo files, other activated ones or disabled ones or other added ones.

When using dnf otoh, it does only steps 2 - 4... sadly dnf searches in the true / (root) directory for repo files. Thus ignoring the repo files found in /<somepath>/yum.repos.d/*.repo.

Temporarely (and contradicting to yum) I have to temporarely copy over the files found in /<somepath>/yum.repos.d/.*repo to /etc/yum.repos.d/ whenever issuing the --installroot command.

With other words... dnf is not correctly switching the context ... when I intend dealing with a different --installroot, then I not only want to store the metadata, cache, downloaded nevra in the installroot but also use the repo files found in the installroots /etc/yum.repos.d directory.