PREFERRED_VERSION for custom linux kernels

PREFERRED_VERSION for custom linux kernels

[Jan Kiszka]

Hi all,

with the introduction of -kbuildtarget and the relevance of -native for
custom kernel recipes, we now have a recipe API change /wrt selecting
the kernel version (PREFERRED_VERSION_linux...) in case there are
multiple. This can look ugly downstream see [1]. I'm now also doing the
same for xenomai-images, and there I thought again about deriving the
preferred version of -native and -kbuildtarget from the one set for the
base target:

PREFERRED_VERSION_linux-xenomai-3-native ?=
"${PREFERRED_VERSION_linux-xenomai-3}"
PREFERRED_VERSION_linux-xenomai-3-kbuildtarget ?=
"${PREFERRED_VERSION_linux-xenomai-3}"

And that leads to my question: Update RECIPES-API-CHANGELOG with this
information or rather set similar defaults for
PREFERRED_VERSION_linux-${KERNEL_NAME}-native/kbuildtarget, obsoleting
the pattern above? Thoughts?

Jan

[1]
https://gitlab.com/cip-project/cip-core/isar-cip-core/-/commit/52508662349362f6e97528976f28a2cbf06e9d97
-- 
Siemens AG, Technology
Linux Expert Center

Re: PREFERRED_VERSION for custom linux kernels

[Jan Kiszka]

On 22.05.24 19:27, 'Jan Kiszka' via isar-users wrote:
> Hi all,
> 
> with the introduction of -kbuildtarget and the relevance of -native for
> custom kernel recipes, we now have a recipe API change /wrt selecting
> the kernel version (PREFERRED_VERSION_linux...) in case there are
> multiple. This can look ugly downstream see [1]. I'm now also doing the
> same for xenomai-images, and there I thought again about deriving the
> preferred version of -native and -kbuildtarget from the one set for the
> base target:
> 
> PREFERRED_VERSION_linux-xenomai-3-native ?=
> "${PREFERRED_VERSION_linux-xenomai-3}"
> PREFERRED_VERSION_linux-xenomai-3-kbuildtarget ?=
> "${PREFERRED_VERSION_linux-xenomai-3}"
> 
> And that leads to my question: Update RECIPES-API-CHANGELOG with this
> information or rather set similar defaults for
> PREFERRED_VERSION_linux-${KERNEL_NAME}-native/kbuildtarget, obsoleting
> the pattern above? Thoughts?
> 

No one commented, but I have some additional thought:

It's not that clear to me where such a generic assignment would best be
hosted if we decided to have it in isar. It likely does not help to put
those into the recipes as those are evaluated after the version choice,
at least based on my understanding. But having this in isar's
bitbake.conf or layer.conf looks weird as well.

Jan

> Jan
> 
> [1]
> https://gitlab.com/cip-project/cip-core/isar-cip-core/-/commit/52508662349362f6e97528976f28a2cbf06e9d97

-- 
Siemens AG, Technology
Linux Expert Center