As I announced on the development mailing list, the ImageFactory stack is being retired. ImageFactory produces some of the release-blocking artifacts for Workstation, notably the disk images for ARM.
For Fedora Workstation, there are three options to choose:
Migrate the Lorax and ImageFactory produced images to KIWI. This is what Fedora KDE and Fedora LXQt have done. This option means that Workstation will only care about the definitions in fedora-kiwi-descriptions and no longer rely on fedora-kickstarts to define any Workstation artifacts. This will also resolve the problem where the ARM live ISO randomly fails to build due to inscrutable SIGKILL errors in Lorax.
Migrate only the ImageFactory produced images to KIWI. This option means that the live ISOs are still produced with Lorax, but disk images are produced with KIWI. Both fedora-kiwi-descriptions and fedora-kickstarts will be canonical definitions of the Workstation deliverable.
Stop producing disk images entirely. This option means that Fedora Workstation will no longer be easily available for ARM platforms. In theory the live ISO means ARM UEFI platforms are still available, but Lorax has been randomly failing to produce the ISO for a few cycles now. We have not figured out how to fix that yet.
My recommendation is option 1, but any of the options are valid choices. I would appreciate a decision from the WG as soon as reasonably possible.
My recommendation is option 1
This seems like surely the best option?
Note that options 1 or 2 will also reduce the delta between Fedora Workstation and Fedora Asahi Remix GNOME too, since Fedora Asahi Remix is produced using KIWI.
My recommendation is option 1 This seems like surely the best option?
I personally think so. Consolidating to one set of definitions makes it much easier to reason how things are produced. Moreover, I consider the kiwi definitions much easier for people to run locally to be able to do tests or changes. The Fedora kiwi definitions has CI to validate changes to actively used release-blocking artifacts to ensure they don't break as a result of a contributed change too.
In some respects, it's much more reliable than our other tools because it's coherent and tested. It also is easy for downstreams to build open to produce either their own lab spins or full remixes.
Metadata Update from @ngompa: - Issue set to the milestone: Fedora 42
Option 1 (all in on Kiwi) sounds right to me too.
Though the current Kiwi profiles look reasonably simple XML.
It's certainly not too early for F42 Workstation, but if we are doing this for F42, then I guess Rawhide should move to Kiwi asap.
The XML that defines Workstation deliverables is fairly simple too.
I've made a pull request for this now: https://pagure.io/pungi-fedora/pull-request/1419
We seem to have consensus, so closing.
Metadata Update from @catanzaro: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)
This is now merged and implemented for F42+.
Log in to comment on this ticket.