It sounds like a good direction, but the biggest issue now is gumstix OE is forked. It won't see these improvements.

On Saturday, March 24, 2012, Richard Purdie <> wrote:
> I just wanted to give some feedback from an OpenEmbedded perspective on
> the topic of this thread.
> I don't know how closely you've followed what has been happening in the
> OE world over the past 18 months but there have been some big changes.
> We were and are all aware OE is a complex thing. Its extremely versatile
> and powerful but also hard to use, understand and work with at times.
> The learning curve to new users is steep.
> I believe we can improve OE and make it more accessible to people. There
> is no single magic bullet to fix this but are many different changes
> which when combined should help. These things take time and resources
> and a combined will. This is partly why the Yocto project was created.
> Some of the changes:
> * A push to split what was a monolithic OE into layers. OE-Core is the
>  shared common core with hardware support for specific hardware and
>  other features being added in layers on top of this.
> * An effort to document more of the system. There is a tech writer
>  working on this full time and the Yocto Project has some manuals
>  which are a start at the documentation problem. They're not perfect,
>  they are a work in progress but they're certainly better than
>  nothing.
> * Focus on usability:
>  - We recently had a drive of "nominate your most confusing error
>    message" and then went and improved them.
>  - A new graphical interface to building images called hob (still
>    using bitbake behind the scenes)
> * Defined QA testing on the core of the system. We test that all 4.5
>  architectures we support build in several different combinations for
>  OE-Core and added automated qemu booting so we can test images at
>  least have basic runtime functionality. This is to address the
>  problem that people pick it up and it doesn't build. Now people
>  should have a starting point they can work from.
> * Focus on consistency and cleanup of the metadata. We've also gone
>  through and fixed warning messages and QA type issues found in the
>  resulting packages.
> * Additions and improvements to tools like standalone SDK/toolchains
>  and application developer functionality like Eclipse integration.
> * Fetching causes a lot of problems (I noticed mention of 404s in this
>  thread). We drove a rewrite of the bitbake fetcher code and made
>  mirroring of source code easier and more deterministic.
> * Work on improving build speed and also allow reusing of prebuilt
>  objects (the sshared-state or sstate cache).
> * Implementation of a bugzilla for OE-Core related issues with a good
>  track record on resolving reported problems.
> So my point is that OE is trying to evolve and that changes are
> happening to try and make things easier and more accessible. Gumstix is
> still using OE-Classic and hasn't seen many of the improvements
> Yocto/OE-Core have brought. Please keep these things in mind when
> thinking about OE.
> I would also mention that if there are things we could do to improve OE
> further beyond what has happened already, the Yocto Project would be
> interested to know about them.
> As it happens and unrelated to this, I do have a gumstix board I've been
> experimenting with recently too :)
> Cheers,
> Richard
> Yocto Project Architect
> OpenEmbedded Technical Steering Committee Member
> Linux Foundation Fellow
> ------------------------------------------------------------------------------
> This SF email is sponsosred by:
> Try Windows Azure free for 90 days Click Here
> _______________________________________________
> gumstix-users mailing list

Brad Midgley