[meta-ti] building Yocto for Pandaboard

[Denys Dmytriyenko]

On Thu, Feb 09, 2012 at 08:59:51AM -0800, Philip Balister wrote:
> On 02/09/2012 08:36 AM, Denys Dmytriyenko wrote:
> > On Thu, Feb 09, 2012 at 11:23:14AM -0500, Brian Hutchinson wrote:
> >>  On Wed, Feb 8, 2012 at 8:16 PM, William Mills <wmills at ti.com> wrote:
> >>> As Gary said there has not been too many end user questions on meta-ti yet.
> >>
> >>  All I care about is meta-ti as that is what all our products are based
> >>  on. I've been watching subject for a while now trying to discern all
> >>  the issues and make a wise choice.
> > 
> > Brian,
> > 
> >>  I'm wanting to switch from Arago to whatever TI supports next as I
> >>  supply the rest of our development team with tools and images that
> >>  they build applications on for our products and I can't jerk them
> >>  around changing distros.
> > 
> > As you are aware, Arago is not going away - there is work going on in 
> > meta-arago layer to update/port it to the new Yocto infrastructure.
> > 
> > Arago/meta-arago is still going to be the official platform distribution for 
> > TI SDK products. But, a separate meta-ti layer was created early in the 
> > process to detach and unify the BSP layer and allow people to use TI hardware 
> > with different distributions. And that's actually part of the problem, as 
> > distributions like religions conflict with each other in a single layer... :)
> > 
> 
> Denys, from my point of view, there are two issues we need to solve:

Philip,

These are very good points you raise!

> 1) Defining the meta-ti toolchain dependencies. Angstrom uses gcc-4.5
> for various reasons. Will the TI programs work against all gcc versions
> available from oe-core/meta-oe?

We've discussed it internally and with some of the other involved parties. 
Right now gcc-4.5 with Linaro patches from meta-oe is still the recommended 
toolchain. We need to validate gcc-4.6 from oe-core though. Few people (Bill 
Mills and Khem Raj among others) are currently at Linaro Connect trying to 
convince Linaro guys to get serios and provide an official meta-linaro layer 
for Yocto.

> 2) Image construction pieces in oe-core are not all there yet. This is
> what leads to angstrom specific bits creeping into the BSP layer. I've
> talked with Paul Eggleton and Koen about this at FOSDEM. We should sit
> down next week at ELC and see if we can come up with a set os tasks we
> can push into oe-core that let all layer/distro combinations produce
> working images.

Some work has already started in regards to tasks and images in meta-ti and 
meta-arago to clean them up and simplify the dependencies. I agree, there is a 
need for a better framework from OE-Core to define base tasks and avoid 
cluttering BSP layers with distro-specific dependencies. Let's talk next week 
at ELC.

-- 
Denys