ARMv4 vs ARMv6

[Graeme Gregory]

On Thu, 2008-10-16 at 10:14 +0100, Andy Green wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Somebody in the thread at some point said:
> 
> | If someone is making OpenMooCow 2.0 and they have a GTA02 and GTA02-only
> | toolchain, it would be good if GTA03 folks are able to install it OK.
> | Or, we should send out a single toolchain+(2 x libs) that is v4 and v6
> | ready, and make it easy for the MooCow guy to generate v4 and v6 binary.
> | ~ Or, we should have the toolchain issue source packages, which Moocow
> | guy will offer with his v4 binary, and make it easy for anyone to recook
> | a source package built on v4 toolchain+libs against v6 toolchain+libs.
> |
> |> Or just use OE?
> 
> The toolchain is for people who don't want to have to deal with OE, it's
> a valid and supported way.  "Just" use OE isn't a good enough
> proposition if we are going to enable casual devs to contribute.
> 
> Can you take me through what you think the toolchain situation would
> look like after we decided that we support v4 and v6 separately?  What's
> in that tarball then?  There are two tarballs now for v4 or v6, or there
> are both sets of libs in there to link against, or what?  When a guy
> builds his package, he can generate v4 and v6 binary packages simply,
> and it is still valid to generate only v4 to satisfy use on both platforms?
> 
I beleive this is where it gets a little complex.

gcc in the toolchain would be quite happy making armv6 or armv4
binaries. But the libraries shipped would only be one or the other. This
is one of the things OE hides from you. 80% of the time linking against
the wrong arch of libraries works anyway. But knowing users they will
find that 20% with regular annoyance.

Graeme