Re: [OE-core] [PATCH] module.bbclass: check whether CONFIG_MODULES set

[Kai <kai.kang@windriver.com>]

[-- Attachment #1: Type: text/plain, Size: 2957 bytes --]

On 2/4/26 23:53, Richard Purdie wrote:
> On Wed, 2026-02-04 at 09:10 -0500, Bruce Ashfield wrote:
>>
>> On Wed, Feb 4, 2026 at 9:05 AM Richard Purdie
>> <richard.purdie@linuxfoundation.org> wrote:
>>> On Wed, 2026-02-04 at 09:02 -0500, Bruce Ashfield wrote:
>>>> On Wed, Feb 4, 2026 at 8:58 AM Richard Purdie
>>>> <richard.purdie@linuxfoundation.org> wrote:
>>>>> On Wed, 2026-02-04 at 08:46 -0500, Bruce Ashfield via
>>>>> lists.openembedded.org wrote:
>>>>>> On Tue, Feb 3, 2026 at 9:09 PM Kai Kang via
>>>>>> lists.openembedded.org
>>>>>> <kai.kang=windriver.com@lists.openembedded.org> wrote:
>>>>>>> From: Kai Kang<kai.kang@windriver.com>
>>>>>>>
>>>>>>> Check whether kernel config CONFIG_MODULES set or not
>>>>>>> before do_compile
>>>>>>> and do_install in module.bbclass. If not set, it cannot
>>>>>>> build and
>>>>>>> install external modules.
>>>>>>>
>>>>>>
>>>>>> This isn't the place to do that check.
>>>>>>
>>>>>> If modules are disabled get modules_do_compile out of the
>>>>>> tasks.
>>>>> Does it make sense to have something including module.bbclass
>>>>> which
>>>>> doesn't have CONFIG_MODULES set?
>>>> Yes, that's what I'm trying to say.
>>>>
>>>> Don't even include it if you can't build modules. That puts
>>>> the conditional in a single place, and keeps the code cleaner
>>>> in the class.
>>> I think we're talking cross purposes a bit.
>>>
>>> The issue is that if you ever reach do_compile in something using
>>> module.bbclass which does not have CONFIG_MODULES set, that is a
>>> problem and a hard error.
>>>
>>> So I'd argue that:
>>>
>>> * we only need the check in do_compile
>>> * it should be a fatal exit hard error, not a warning
>>>
>>> which means the patch needs tweaking but not as you're suggesting.
>>>
>>>
>>
>> I don't have a strong opinion, so won't object.
>>
>> I just don't think adding any conditionals for module support in
>> moduldes.bbclass makes sense. So one versus two checks
>> is a wash for me.
> I agree the warn conditional isn't quite right. I think it should be:
>
> if [ CONFIG_MODULES not set in config ];
>      bberror Broken config, CONFIG_MODULES not set
>      exit 1
> fi
>
> so it is basically a sanity test that the config is valid.
>
> I'm a bit worried about what I'm missing though :/
The original problem is kernel config_modules is disabled, then external 
modules such as lttng-modules can't be built.

Recipe lttng-tools RRECOMMENDS on lttng-modules. If install lttng-tools 
into image, it fails to build lttng-modules.

Kernel config CONFIG_MODULES couldn't be checked whether enabled or 
disabled after bitbake parse. Then there is no chance to remove 
lttng-module from the dependencies. That why I check the config during 
the do_compile and do_install with bbnote and generate empty rpm package.

Regards,
Kai

>
> Cheers,
>
> Richard
>
>
>

-- 
Kai Kang
Wind River Linux

[-- Attachment #2: Type: text/html, Size: 4792 bytes --]