[Arm-netbook] [review] SoC proposal

[Bari Ari]

On 02/09/2012 08:49 AM, Gordan Bobic wrote:
>
>
> And it isn't just about GCC - it's about the entire package set. There
> are a lot of things that still don't build/work on ARM, not even
> armv5tel target, despite that having been around for years.
>
> GCC + kernel + glibc is a great start, but if you are coming up with a
> whole new arch from scratch, it's still going to take quite a long time
> before package maintainers acquire the hardware and the inclination to
> support the architecture.
>
> I'm not saying this is unachievable (even if it means sending a dev kit
> to maintainers of all 2000-4000+ packages in a typical Linux distro),
> but I don't think the issue should be trivialized.
>
> It took me 3-4 months of evenings and weekends to just get (at least all
> the important things) RHEL6 to build on ARM. It took patching 71
> packages. Most (but certainly not all) of what was required to be fixed
> already had at least provisional patches upstream, but a lot required
> back/forward porting patches, and some things I had to fix myself.
>
> And that's on an arch where the version of Fedora (12) that RHEL6 is
> based on was already (mostly) working on ARM.
>
> But a whole new architecture, with all the debugging and testing that
> will take? Sure it's possible, and I'll stick my ore in and help if it
> looks like it's going to help, but the scale of the task seems
> impossible to over-estimate.
>
>
It certainly is a train wreck when you look back on how the ARM support 
has evolved. Even the ARM kernel maintainers had to take a break from 
the madness they were getting from all the ARM vendors with their own 
agendas.

I agree that most of the work will be in porting packages (with x86 in 
mind) to this.