Somehow got stuck on a non-weekly version, can't update

Realized recently that my aurora install had stopped updating, and I’m not sure what happened at the time to cause it since I took so long to notice

rpm-ostree status reported no errors, and rpm-ostree upgrade –check reported no available updates, despite me being on a month old version:

original rpm-ostree status

ostree-image-signed:docker://ghcr.io/ublue-os/aurora-dx:stable (index: 1)
                   Digest: sha256:161c2a141446c76e3714b99b3422a2ad7340d2477adf1f9e46c7638fe5494469
                  Version: 43.20251119.1 (2025-11-19T16:00:52Z)
               BaseCommit: 39851ccc0ec4eb399bec0bc8fdc2d1a3b48e355fafb6f678f7a5f81d5a61f7b7
                   Commit: 08ec00174cababf03e435aaeab806bc30745aff74149db8c45030a1c19f179ba
                StateRoot: default

rpm-ostree upgrade conflict

if I ran rpm-ostree upgrade then I got an error about a hardlink conflict between aurora-logos and fedora-logos. I assumed aurora-logos was part of the OS, but when I checked another computer of mine (that was on latest aurora), it didn’t have either of those packages… I also couldn’t find any way to tell why aurora-logos was installed, or why fedora-logos was trying to be installed.

I figured I should try to see where those packages were introduced/removed, so I tried to use rpm-ostree db diff to compare the versions, but my out of date computer didn’t recognize any newer commit hashes, and my up to date computer didn’t recognize the commit 39851ccc0ec4eb399bec0bc8fdc2d1a3b48e355fafb6f678f7a5f81d5a61f7b7 which my out of date computer was stuck on?

I went into the github container version history and tried to find my 3985… release, and I discovered it didn’t seem to be one of the standard daily or weekly releases at all, and in fact has zero downloads, so I have no idea how I ended up on it.

Comparing the out of date and up to date rpm-ostree status again it became clear that there was no latest- at the beginning of my Version like there should (?) be.

It seemed like I was on a stale ‘branch’ or something (to use git terms), so I figured I needed to get back to the ‘main branch’ (the latest- tags), so I rebased to latest-43.20251119, figuring that would fix it.

current rpm-ostree status

ostree-image-signed:docker://ghcr.io/ublue-os/aurora-dx:latest-43.20251119 (index: 0)
                   Digest: sha256:196cbd40961dbe9ec9e4afb56ea3d3bc369c963afd9bb26f66beb90e57c80bff
                  Version: latest-43.20251119.1 (2025-11-19T19:54:19Z)
               BaseCommit: b6eed3e4aa44543de45ed9eb585ba6e1e16a30d717d3da4c5c67cf0e66aaca39
                   Commit: 87ddd23a24a9c629eeb3c16743e70ff7e282de9d0a1a1b400df7ff070407b638

That seems to have worked fine, however, it’s still not finding any updates, either via rpm-ostree update or via the automatic check. I’m assuming this is because my ‘branch’ is now the specific latest-43.20251119 commit instead of just aurora-dx:latest?

the main question

Now that I’m here, how can I get it back on the right track? Is there a way to correct my image ‘field’ to just aurora-dx:stable so that I can then run rpm-ostree upgrade normally? Or is my only option here to rebase directly up to the latest aurora-dx:stable ?

bonus questions

  1. normally, in what situation would rebasing to the latest version not work, but updating would? I’m having a bit of a hard time understanding what difference there is to this at least in terms of ‘git’ reasoning
  2. is there a way I can ‘bookmark’ commit that’s a known-good state before messing with rebases/etc? currently I only ever see the two versions and as soon as you make changes the old one is lost, but in a situation like this, the old one is technically the stable (if out of date) system I’ve been running fine for a month, while the new one, while it seems to be working, isn’t exactly battle tested. I don’t want to experiment with more operations to fix things right now because as soon as I do anything I’ll lose the stable old commit I have
  3. is there any way to tell when running into an upgrade conflict like this what is causing the new package to be installed? it feels like there must be way more information about the upgrade process that’s being hidden, resulting in getting this strange hardlinking error message with no context, but I don’t see any ‘verbose’ option and aren’t aware of any relevant log files
  4. is there a way to override the dnf error on atomic systems, at least for read-only actions, so that I can use it to query the state of the dependency tree/etc? I know I can’t use it to make modifications but when researching online there’s a lot of queries/etc that seem to only be possible using it
  5. is there an easy way to find the release in github by commit hash? I couldn’t find any search field and had to resort to clicking through pages one by one

Hey,

You have pinned your OS to a specific version for whatever reason you had.

Have you tried using ujust rebase-helper or sudo bootc switch --enforce-container-sigpolicy ghcr.io/ublue-os/aurora-dx:stable commands?

More information:

1 Like

Isn’t bootc switch basically going to do a rebase, but without handling layered packages/etc well? If so, then I think any of my concerns/questions about rebasing would also apply to it?

You propably need to remove those layers if you can’t rebase. That will always be an issue with layered packages. At some poiint it will break your updates because of package conflicts or something like that.

But ujust rebase-helper is the easieast way to do the rebase to non-date image.

it’ll probably rebase fine, (I haven’t tried it yet since I don’t want to lose my previous version history without knowing it’s the only option) but I didn’t want to risk it if there was a way to recover via more normal methods

would using ostree admin pin 1 be the proper way to make sure my original ‘stuck’ rollback commit is preserved before I give ujust rebase-helper a shot?

Yes, you can pin older image deployments with sudo ostree admin pin <x>

Those will always be preserved and never “overwritten” or wiped from the system.

Rebase with layers propably goes fine, most of the issues will show themselves when the base fedora version changes.

Okay, pinned the old version and tried ujust rebase-helper… still ran into the issue


ujust rebase-helper
Choose your action.
Current image: aurora-dx:latest (Fedora 43)
Root filesystem: overlay

Select your Aurora image:
The default selection is stable. Stable-daily (daily builds) are for enthusiasts, and latest is for testers
Would you like to pin to a specific build date?
NOTICE! You won't receive any updates after pinning to a specific build date.
Rebase target: ghcr.io/ublue-os/aurora-dx:stable
Confirm rebase?
Pulling manifest: ostree-image-signed:docker://ghcr.io/ublue-os/aurora-dx:stable
Importing: ostree-image-signed:docker://ghcr.io/ublue-os/aurora-dx:stable (digest: sha256:5e7f3ee96ed39ad9bb84de9c3ab7f110e8dfeb779b5ef91e77a4e29d4d376f2a)
ostree chunk layers already present: 53
ostree chunk layers needed: 20 (1.5 GB)
[0/20] Fetching ostree chunk 4cb386e8f63b94a63bb (129.8 MB)... done
[1/20] Fetching ostree chunk bd3ee20eceff2aa2d17 (26.3 MB)... done
[2/20] Fetching ostree chunk 71cf4cc1b107be8ea72 (240.8 MB)... done
[3/20] Fetching ostree chunk 8a36ec6a833a67df42e (228.3 MB)... done
[4/20] Fetching ostree chunk bc85eb41c3c7a9cf1e2 (64.9 MB)... done
[5/20] Fetching ostree chunk 2d55e1f9db653aae232 (16.3 MB)... done
[6/20] Fetching ostree chunk a9d4380a47dc2404e6c (43.7 MB)... done
[7/20] Fetching ostree chunk d5fb8de606d9fde9247 (21.7 MB)... done
[8/20] Fetching ostree chunk ce1ce401c261f70e919 (85.9 MB)... done
[9/20] Fetching ostree chunk 763fb74743bf2c0b352 (41.6 MB)... done
[10/20] Fetching ostree chunk 1005ec6377f6d2655bc (79.1 MB)... done
[11/20] Fetching ostree chunk 6717fce2f4405b11d92 (61.7 MB)... done
[12/20] Fetching ostree chunk 4058f8f61bb8440c83f (23.4 MB)... done
[13/20] Fetching ostree chunk e27b0a4ce3feedd2fff (43.8 MB)... done
[14/20] Fetching ostree chunk f1bf11642146a5a0f24 (62.1 MB)... done
[15/20] Fetching ostree chunk e75b83dbdd83ff00b9b (44.3 MB)... done
[16/20] Fetching ostree chunk e9208c1844052d70b57 (49.5 MB)... done
[17/20] Fetching ostree chunk 18700c227f8a8f314ae (15.9 MB)... done
[18/20] Fetching ostree chunk b0d2c6207d55399bed8 (168.6 MB)... done
[19/20] Fetching ostree chunk 653138440476e7ff448 (11.1 MB)... done
Checking out tree 8f72dcf... done
Enabled rpm-md repositories: updates fedora fedora-multimedia copr:copr.fedorainfracloud.org:alternateved:keyd updates-archive
Updating metadata for 'updates'... done
Updating metadata for 'fedora-multimedia'... done
Updating metadata for 'updates-archive'... done
Importing rpm-md... done
rpm-md repo 'updates'; generated: 2025-12-30T00:34:48Z solvables: 17531
rpm-md repo 'fedora' (cached); generated: 2025-10-23T03:37:20Z solvables: 77664
rpm-md repo 'fedora-multimedia'; generated: 2025-12-26T14:01:22Z solvables: 420
rpm-md repo 'copr:copr.fedorainfracloud.org:alternateved:keyd' (cached); generated: 2025-12-19T22:03:43Z solvables: 8
rpm-md repo 'updates-archive'; generated: 2025-12-30T00:48:06Z solvables: 19770
Resolving dependencies... done
Will download: 4 packages (10.4 MB)
Downloading from 'updates'... done
Downloading from 'fedora'... done
Importing packages... done
Relabeling... done
Checking out packages... done
error: Checkout fedora-logos-42.0.1-2.fc43.noarch: Hardlinking 93/9b502154717b9671b7cd572620f467bc1fcc296901640f71cb1f5c963f9231.file to start-here.svg: File exists

tried using bootc switch directly, which ‘worked’…except it got rid of my layered packages which I didn’t expect

reinstalled them again one by one after that and was able to figure out which one had a 4 layer deep transient dependency that was causing fedora-logos to be installed because aurora-logos was removed in early december. The logo files are now included directly in the image, but RPM doesn’t know that so it tried to fall back to fedora-logos and then conflicted with the aurora image.

for now, just going to do without that package