DnfRepo: fix module_hotfixes keyfile priority level

[jlebon]

We were setting the runtime value of the knob instead of the repoconfig
value. This meant that a user wanting to override the config via
dnf_repo_set_module_hotfixes (which also sets the runtime value) would
have its changes clobbered.

This should fix Silverblue Rawhide rpm-ostree composes:
https://pagure.io/releng/failed-composes/issue/717
https://pagure.io/releng/failed-composes/issue/929

where we use a hack to allow rpm-ostree to install modular packages
until we fully support modules (which requires #874) and it was getting
reset because dnf_repo_download_packages wants to reset the repo
config from the keyfile. For more information, see discussions at:

#885

[cgwalters]

@hughsie do you remember this one?

[cgwalters]

Looking at librepo history there's a whole pile of activity around then with possibly related bits, e.g.
rpm-software-management/librepo@a03aaa8 and https://github.com/rpm-software-management/librepo/commit//9f9bd95ffd36248891ffe64be1a494e38b7387b9

[jlebon]

Hmm, OK I think I follow this a bit more now. In dnf_repo_check_internal, we set LRO_LOCAL=1, clear LRO_MIRRORLIST and set LRO_URLS to our cached rpmmd so that librepo operates directly on it to determine if it's complete & up-to-date. So naturally, we want to set it back to the original values afterwards. This explains why dnf_repo_check_internal resets the knobs to the keyfile data at the end and why cc0f5aa added LRO_LOCAL=0 there.

But anyway, I realized the root issue here was actually that the module_hotfixes knob read from the keyfile was given priority RUNTIME instead of REPOCONFIG like the others. Just fixing that fixes the issue!

(Though again, it does seem to me the call to dnf_repo_set_keyfile_data in dnf_repo_update and dnf_repo_download_packages should be dropped since we already reset them in dnf_repo_check_internal which is needed for setting up the sack and it allows callers to modify knobs before actually downloading packages. Though the justification for changing these old behaviours at this point without a use case affected by this is low.)

[jlebon]

Would be good to have this in soon. Latest patch should be much simpler to review :). This will unblock Rawhide Silverblue composes, which hasn't been updated since almost a month and a half now.

[j-mracek]

LGTM!

[j-mracek]

@rh-atomic-bot r+

[rh-atomic-bot]

📌 Commit b4c9074 has been approved by j-mracek

[rh-atomic-bot]

⌛ Testing commit b4c9074 with merge 4a7ab08...

[jlebon]

Thanks for the review!

Hmm, looks like the CI build is frozen?

2020-02-04 14:35:21,538:DEBUG: Build #1220967: running
2020-02-04 14:35:26,809:INFO: Build #1220967: succeeded

Yet all the packages have built:
https://copr.fedorainfracloud.org/coprs/rpmsoftwaremanagement/rpm-gitoverlay-1580815068.819735/packages/

Can we retry?

[rh-atomic-bot]

☀️ Test successful - status-papr
Approved by: j-mracek
Pushing 4a7ab08 to master...