[meta-intel] [PATCH 0/4] misc meta-intel Fixes

[Kamble, Nitin A]

> -----Original Message-----
> From: Richard Purdie [mailto:richard.purdie at linuxfoundation.org]
> Sent: Wednesday, August 14, 2013 1:32 PM
> To: Kamble, Nitin A
> Cc: Burton, Ross; meta-intel at yoctoproject.org; Zanussi, Tom; Hart, Darren
> Subject: Re: [meta-intel] [PATCH 0/4] misc meta-intel Fixes
> 
> On Wed, 2013-08-14 at 20:17 +0000, Kamble, Nitin A wrote:
> >
> > > -----Original Message-----
> > > From: Burton, Ross [mailto:ross.burton at intel.com]
> > > Sent: Wednesday, August 14, 2013 1:06 PM
> > > To: Kamble, Nitin A
> > > Cc: Zanussi, Tom; Hart, Darren; meta-intel at yoctoproject.org
> > > Subject: Re: [meta-intel] [PATCH 0/4] misc meta-intel Fixes
> > >
> > > On 14 August 2013 20:46,  <nitin.a.kamble at intel.com> wrote:
> > > > Recently an issue was observed and fixed on meta-minnow, which was
> > > > breaking builds when DISTRO is not poky. The same issue also exist
> > > > on the crownbay, emenlow, fri2 & sys940x BSPs.
> > > >   Here are the commits to fix the issue for these BSPs.
> > >
> > > Not sure I like changing DISTRO_FEATURES from within a BSP, that's a
> > > layering violation in my opinion.
> > >
> > > There was a patch a week or so ago to add a
> > > distro_features_check.bbclass, which let you do this in recipes:
> > >
> > >   inherit distro_features_check
> > >   REQUIRED_DISTRO_FEATURES = "x11"
> > >
> > > There was some debate on the patch but it's fairly trivial to do
> > > this sort of check in an anonymous Python function without the
> > > class, systemd does this to ensure it only gets built in a distribution with
> the systemd feature.
> > >
> > > Ross
> >
> > Hi Ross,
> >   Good to see similar fix for systemd, and thanks for bringing in the
> reference.
> >
> >   EMGD being the required driver for hardware, this case is different
> compared to systemd case.
> > In this case the feature is required, and not optional. The BSP is
> > broken and can not even build without the DISTRO_FEATURE. So IMO
> > making sure the feature is enabled is better that just checking weather the
> feature is enabled or not.
> >   As the oecore layer is one of the needed layer for these BSPs, I do not see
> it as a layering violation.
> > Most of MACHINE_FEATURES are also defined in the oecore layer.
> 
> DISTRO_FEATURES is something the distro configuration is responsible for
> setting up. Merging this change would put meta-intel's yocto project
> compatible status at risk.

Hi RP,
  I understand the point now. For everyone's information, I think you are mentioning this part from the compatible agreement.
"Are hardware support, configuration (distro) policy, and recipe metadata separated into different layers which do not depend on each other?"

So we can only check and set the DISTRO_FEATURES var from meta-intel layers.

Nitin

> 
> The system needs to error if the configuration doesn't match the
> requirements and simply state the machine cannot be built without that
> feature.
> 
> Cheers,
> 
> Richard