[meta-freescale] [meta-fsl-arm-extra][PATCH] nitrogen6x.conf: Allow kernel provider override

[Eric Nelson]

Hi Otavio,

On 10/30/2013 07:28 AM, Otavio Salvador wrote:
> Hello Eric,
>
> On Wed, Oct 30, 2013 at 12:25 PM, Eric Nelson
> <eric.nelson at boundarydevices.com> wrote:
>> On 10/30/2013 07:10 AM, Otavio Salvador wrote:
>>>
>>> On Wed, Oct 30, 2013 at 11:41 AM, Gary Thomas <gary at mlbassoc.com> wrote:
>>>>
>>>> This change lets the user override the choice of kernel in local.conf
>>>> Without it, there is no way to build any kernel, e.g. linux-imx, other
>>>> than the linux-boundary version.
>>>>
>>>> Signed-off-by: Gary Thomas <gary at mlbassoc.com>
>>>
>>>
>>> I really dislike this.
>>>
>>> I understand your need, and it is a valid one, but it'd be better you
>>> to make a new machine (which includes this one) and override it there.
>>>
>>
>> I'm not sure I understand why. It seems perfectly reasonable to me
>> to build a new kernel recipe and use it, while retaining other
>> information from the machine configuration.
>>
>> I'm thinking specifically of a user who may have a custom kernel
>> tree with pin-muxing based on their usage.
>
> I see the point and I do have this case internally here; the point is
> if we have it 'soft' and someone reports:
>
> Nitrogen6X is not working. Next question needs to be, are you using
> linux-boundary or another?
>

I understand your point, but it's going to happen anyway. The beauty
(and curse) of flexible hardware is that it's flexible ;)

I think the best we can hope for in terms of support is to always
ask if a failure occurs with a "stock" build... Fresh "repo sync",
specified U-Boot, et cetera.

I don't know if hard-coding this value will meaningfully change
things.

>>> If we start allow all kind of override in machine configuration it
>>> loses its meaning and complicates the support.
>>>
>>> Eric? comments?
>>>
>>
>> I'm not sure I understand the concern. This seems pretty
>> straightforward. The default is there, but a user can
>> over-ride it.
>
> If user forks the kernel, adding a machine .conf  file is the easiest
> part of it. I'd to avoid support uncertainty regarding settings.
>

That's true, too (machine configurations aren't very complicated), but
my understanding of that is recent.

Regards,


Eric