[OE-core] [PATCH] mime,mime-xdg: Conflict with allarch

[Khem Raj]

On 2/25/20 6:07 AM, Richard Purdie wrote:
> On Tue, 2020-02-25 at 13:24 +0100, Andreas Müller wrote:
>> On Tue, Feb 25, 2020 at 11:21 AM Richard Purdie
>> <richard.purdie at linuxfoundation.org> wrote:
>>> On Mon, 2020-02-24 at 21:44 -0800, Khem Raj wrote:
>>>> mime and mime-xdb classes add dependencies ( both runtime and
>>>> buildtime
>>>> ) on target packages, therefore recipes using allarch should not
>>>> inherit
>>>> mime/mime-xdg or vice-versa, this fixes subtle sstate mismatches
>>>>
>>>> [YOCTO #13805]
>>>>
>>>> Signed-off-by: Khem Raj <raj.khem at gmail.com>
>>>> Cc: Andreas Müller <schnitzeltony at gmail.com>
>>>> ---
>>>>   meta/classes/mime-xdg.bbclass | 5 +++++
>>>>   meta/classes/mime.bbclass     | 5 +++++
>>>>   2 files changed, 10 insertions(+)
>>>
>>> Is this the right way to fix this? mime info probably is arch
>>> independent, so should we be whitelisting some of the dependencies?
>>>
>>> Cheers,
>>>
>>> Richard
>>>
>> Khem and I had a short discussion about this [1]. His first approach
>> was to remove DEPENDS. As far as I remember the deps were added to
>> avoid QA warning.
>>
>> Honestly I never understood why these are spawned. If there is a fix /
>> workaround this could be reduced to rdeps which an be handled for
>> allarch.
>>
>> [1] https://github.com/YoeDistro/openembedded-core/commit/5605a9cf78c3731e2cc055cd4f8cc88a0b3d7e26
> 
> I'd guess something like:
> 
> SIGGEN_EXCLUDERECIPES_ABISAFE += "\
>    shared-mime-info \
>    desktop-file-utils \
> "
> 
> in layer.conf might be the correct fix. The other approach is just a
> workaround.

yeah, I am testing something like this.

> 
> Cheers,
> 
> Richard
> 
> 
>