On Thu, Oct 10, 2024 at 6:07 AM Mikko Rapeli <[email protected]> wrote:
>
> Hi,
>
> On Wed, Oct 09, 2024 at 04:53:58PM -0300, Ricardo Salveti wrote:
> > Wouldn't it be better if this was a kernel class instead, similar to
> > how it is done with fitimage (via kernel-fitimage.bbclass)?
> >
> > I see a lot of similarities here, and it is confusing that one is done
> > as a kernel class and the other is added by including in the image
> > recipe instead.
>
> Interesting idea. I'll have a look. Maybe this could be an improvement
> on top of the uki.bbclass.

Yes, the current patch set seems to be quite functional and we can
work based on it.

> UKI combines kernel, initramfs image etc
> so the implementation can be an image class, kernel class or some
> magic post processing (e.g. wic) but needs to be generated using
> systemd tooling. Thus I think image class it is for now. I agree
> that setting this up is not nice. Need to configure initrd, kernel
> command line, kernel, systemd-boot etc and all the config switches
> are independent. At least the selftests will contain a fully
> working example so users can replicate that in their builds.

Right, for fitimage we have a similar build dependency chain, and why
I was wondering if it could be also made available via a kernel-class.

Thanks,
-- 
Ricardo Salveti
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#205407): 
https://lists.openembedded.org/g/openembedded-core/message/205407
Mute This Topic: https://lists.openembedded.org/mt/108906947/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to