Re: [PATCH] meta: Remove overrides from kernel filename selection

["'Jan Kiszka' via isar-users" <isar-users@googlegroups.com>]

On 08.06.26 11:25, Moessbauer, Felix (FT RPD CED OES-DE) wrote:
> On Mon, 2026-06-08 at 10:13 +0200, Jan Kiszka wrote:
>> On 08.06.26 09:34, 'Jan Kiszka' via isar-users wrote:
>>> On 08.06.26 08:43, Anton Mikanovich wrote:
>>>> 08.06.2026 09:26, Jan Kiszka wrote:
>>>>> On 28.05.26 16:20, 'MOESSBAUER, Felix' via isar-users wrote:
>>>>>> On Thu, 2026-05-28 at 16:56 +0300, Anton Mikanovich wrote:
>>>>>>> Because of the override processing logic values set by overrides can be
>>>>>>> changed only by overrides. It makes redifinition of KERNEL_FILE value
>>>>>>> made by simple assignment be ignored, which was already broken for the
>>>>>>> rpi-arm64-v8 machine.
>>>>>>>
>>>>>>> Rebuild selection logic to be done in bitbake.conf without overrides.
>>>>>>> Keep KERNEL_FILE redifinition for the rpi targets as an example of its
>>>>>>> usage for the downstreams.
>>>>>>>
>>>>>>> Signed-off-by: Anton Mikanovich <amikan@ilbers.de>
>>>>>>> Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
>>>>>>> ---
>>>>>>>   meta/classes-recipe/linux-kernel.bbclass | 3 ---
>>>>>>>   meta/conf/bitbake.conf                   | 5 +----
>>>>>>>   2 files changed, 1 insertion(+), 7 deletions(-)
>>>>>>>
>>>>>>> diff --git a/meta/classes-recipe/linux-kernel.bbclass b/meta/
>>>>>>> classes-recipe/linux-kernel.bbclass
>>>>>>> index 55ea6d63..e4ae356d 100644
>>>>>>> --- a/meta/classes-recipe/linux-kernel.bbclass
>>>>>>> +++ b/meta/classes-recipe/linux-kernel.bbclass
>>>>>>> @@ -211,9 +211,6 @@ def get_kernel_arch(d):
>>>>>>>     KERNEL_ARCH ??= "${@get_kernel_arch(d)}"
>>>>>>>   -# set KERNEL_FILE without depending on package arch used in
>>>>>>> bitbake.conf
>>>>>>> -KERNEL_FILE:forcevariable = "${@ 'vmlinux' if
>>>>>>> d.getVar('KERNEL_ARCH') in ['mipsel', 'riscv', 'arm64'] else
>>>>>>> 'vmlinuz'}"
>>>>>> Shall we remove the forcevariable override as well? At least in isar
>>>>>> this was the last remaining user.
>>>>>>
>>>>> forcevariable is a generic OE concept that we should continue to
>>>>> support. Our downstream users may expect it to work as before.
>>>>>
>>>>> But this removal above likely reintroduces the bug it fixed in
>>>>> fc99297b40eb. Did you actually check against that case? Sorry, seen this
>>>>> too late, but reading the git history is what I would have expected. And
>>>>> if you did, you must refer to it, explaining why a revert is fine.
>>>> I've tried to build native kernels and it looks like everything is fine.
>>>> Unfortunately fc99297 do not mention any real targets and don't have a test
>>>> case, so I can't be 100% sure it isn't broken.
>>>
>>> Did you check the resuling KERNEL_FILE for, say, riscv in case of
>>> building -native? If that is still correct, would be good to leave a
>>> reasoning behind, why.
>>>
>>
>> So, the patch is fine, though only by luck, not by analysis:
>>
>> Overrides work per PACKAGE_ARCH, and the -native package for a kernel
>> has HOST_DISTRO_ARCH here. That is why the old logic needed that fix-up.
>> The new one uses DISTRO_ARCH, thus is not affected by the -native
>> differences.
> 
> What is HOST_DISTRO_ARCH, I can't find this in any commit? Is that
> HOST_ARCH?
> 

My AI must have hallucinated this: HOST_ARCH.

> Thanks for the analysis. Still, I'm wondering what exactly must be
> built to expose this? Is a simple kernel crosscompile for riscv
> sufficient?

I built over fc99297b40eb0ac4b741926fa399d9b5eaaaf825^
linux-starfive-native, and KERNEL_FILE was vmlinu*z*.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center

-- 
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/isar-users/d7c43403-4366-4467-afb0-dc69eb7a8800%40siemens.com.