[OE-core] serial-console: IMAGE_FEATURE, MACHINE_FEATURE or DISTRO_FEATURE?

[Chris Larson]

On Mon, Oct 24, 2011 at 1:27 AM, Frans Meulenbroeks
<fransmeulenbroeks at gmail.com> wrote:
> 2011/10/24 Koen Kooi <koen at dominion.thruhere.net>
>>
>> Hi,
>>
>> Otavio is cleaning up systemd in meta-oe to work with his initramfs setup
>> and we stumbled onto a challenge that we'd like some more feedback on.
>>
>> The old sysvinit recipes always hardcode a serial getty into inittab,
>> systemd installs a getty@<port>.service file instead. Otavio doesn't need
>> (or want) a getty on serial in his setup, but I need (and want) it on all my
>> setups. So we agreed that it needs to be a feature :)
>>
>> I nitially suggested MACHINE_FEATURE, which Otavio implemented here:
>> http://patchwork.openembedded.org/patch/13713/
>>
>> , but I'm having second thoughs about it and am leaning toward an
>> IMAGE_FEATURE instead. Since this involves *_FEATURE and we want systemd to
>> move into oe-core eventually I'm moving the discussion from OE-devel to
>> OE-core. Not much discussion has been going on, so I hope we'll get some
>> more feedback here. I'll try to buttonhole some more OE developers at ELC
>> and report back as well.
>>
>
>
> Technically it is probably both.
> I've seen embedded systems without a console. That seems to indicate a
> machine feature.
> Then again even if you have a console you might not want to enable it (e.g.
> in order not to give the customer root access or even info wrt printk etc).
> That seems to indicate a machine feature.
> Probably the answer is inbetween: A machine feature HAS_CONSOLE or something
> like that and an image feature CONSOLE_ENABLED.

That sounds analogous to
MACHINE_FEATURES+DISTRO_FEATURES==COMBINED_FEATURES, but by image
rather than distro?
-- 
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics