[OE-core] [RFC] recipes-efl inside meta-oe or meta-efl next to meta-oe?

Graeme Gregory dp at xora.org.uk
Wed Mar 23 23:53:16 UTC 2011


On Wed, Mar 23, 2011 at 11:02:17PM +0000, Richard Purdie wrote:
> On Wed, 2011-03-23 at 17:52 +0000, Graeme Gregory wrote:
> > On Wed, Mar 23, 2011 at 04:03:04PM +0000, Joshua Lock wrote:
> > > On Wed, 2011-03-23 at 15:40 +0000, Graeme Gregory wrote:
> > > > On Wed, Mar 23, 2011 at 08:19:31AM -0700, Khem Raj wrote:
> > > > > On Wed, Mar 23, 2011 at 4:49 AM, Koen Kooi <koen at dominion.thruhere.net> wrote:
> > > > > > Hi,
> > > > > >
> > > > > > I'd like to import the EFL recipes Martin did from meta-shr into the meta-openembedded repo. Before I go bothering Martin about it, what would be the best place to put them?
> > > > > >
> > > > > > Inside meta-oe:
> > > > > >
> > > > > > meta-openembedded/
> > > > > >        meta-oe/
> > > > > >                recipes-efl
> > > > > 
> > > > > Can efl we layered directly on top of oe-core  ? i.e. without needing meta-oe
> > > > > in that case its better to be an independent layer. otherwise I would say put
> > > > > them under recipes-efl
> > > > > 
> > > > That is certainly not how I see layers being built up. I am expecting
> > > > in future us to have things like.
> > > > 
> > > > meta-gnome
> > > > meta-kde
> > > > meta-efl
> > > > meta-lxde
> > > > meta-xorg
> > > 
> > > Me too, in fact I think this image from the Yocto Project website
> > > succinctly portrays the goal:
> > > http://www.yoctoproject.org/sites/default/files/yocto-layers_1.png
> > > 
> > > Possibly less the meta-yocto but you get the point.
> > > 
> > I hope not as thats a truly horrible diagram.
> > 
> > I dont think there is anyway in modern linux to produce a nice neat stack
> > like that. Once you move beyond trivial images.
> 
> The diagram is idealised but I think the concepts there are valid, for
> example, bring in the meta-linaro layer to use the linaro toolchain,
> bring in a BSP layer for a particular piece of hardware support and so
> forth. Certainly, a developer is likely to have local tweaks in a layer
> of their own on top.
> 
> Why wouldn't that work?
> 
The diagram demonstrates a nice stack, but what about if you want two
gui layers. Real life example, gnome and kde have cross dependencies on
each other.

In a real system you just cant pile up a nice little stack of components. It
works more like lego where you have multiple components that fit together
depending on multiple components to hold them up.

The diagram is overly idealised to the point of zero information, fit for
marketing but not engineering.

Graeme





More information about the Openembedded-core mailing list