Re: [RFC][PATCH] Attempt to unify predefined package feed handling

["David Nyström" <david.nystrom@enea.com>]

On 2014-03-01 19:10, Otavio Salvador wrote:
> Hello David,
>
> On Thu, Feb 27, 2014 at 4:59 PM, David Nyström <david.c.nystrom@gmail.com>wrote:
>
>>
>> On 2014-02-27 16:18, Laurentiu Palcu wrote:
>>
>>> On Thu, Feb 27, 2014 at 03:35:50PM +0100, David Nyström wrote:
>>>
>>>> On 2014-02-27 15:24, Laurentiu Palcu wrote:
>>>>
>>>>> Hi David,
>>>>>
>>>>> My comments/questions below.
>>>>>
>>>> Thank you for your detailed comments on the RFC.
>>>> I will return with V2, where your comments are addressed when the
>>>> general approach is Acked.
>>>>
>>>> What do you think about the general approach ?
>>>>
>>> I think the approach is ok. I don't know a simpler way than this to add
>>> package feeds into the image.
>>>
>>
>> OK, there seems to be no screaming objections from other parties either.
>>
>> I'll resend v2
>>
>
> I understand why you are adding this here but I think it'd be good to the
> settings to be put in a package so in a product we may add extra
> repositories updating the package.
>

Good point.

> I am trying to think now how this could be done (or optionally done).
>
> What you think?

Yes, how would we determine the repo subdirs for a specific machine 
without too much
anonymous python magic in a recipe, considering multilib setups?

Example for qemux86-64:
-- 
all
x86_64 = ${TARGET_ARCH}
core2-64 = ${??}
qemu86_64 = ${MACHINE}
-- 

If we can get a sane way to retrieve them, modifying the repo feeds via 
install/postinstall should be the easy part.


> Other point, PACKAGE_FEED_URIS could be change to PACKAGE_FEED_URI as it
> matches SRC_URI for example.

Agreed.