[Libosinfo] [PATCH libosinfo] test-isodetect: add Debian's arm flavours

[Christophe Fergeau]

On Thu, Jul 13, 2017 at 04:34:28PM +0100, Daniel P. Berrange wrote:
> On Thu, Jul 13, 2017 at 05:18:24PM +0200, Christophe Fergeau wrote:
> > On Thu, Jul 13, 2017 at 02:39:19PM +0200, Guido Günther wrote:
> > > On Wed, Jul 12, 2017 at 09:35:16AM +0200, Christophe Fergeau wrote:
> > > > 
> > > > However for this one I'd expect arch = "armv7l" as this is what is in
> > > > the RNG schema? My understanding of
> > > > https://bugzilla.redhat.com/show_bug.cgi?id=719609#c13 is that armv7l is
> > > > a soft floating point arch, armv7hl would be hard floating point. And
> > > 
> > > 
> > > 
> > > > armhf is hard floating point as well. So I don't think we can consider
> > > > the 2 are the same here. You probably need to add a new ARM arch to the
> > > > schema file, and use that for armhf.
> > > 
> > > On my A20 Olinuxino I see
> > > 
> > > $ uname -m
> > > armv7l
> > > 
> > > $ dpkg --print-architecture
> > > armhf
> > > 
> > > I assumed that the armv7l refers to an architecture supporting hard
> > > float but your link suggests otherwise. In this case we need to fixup
> > > osinfo-db as well. I couldn't find any other good references on this.
> > > 
> > 
> > Oh well, I'm utterly confused by ARM arch names, so maybe armv7l and
> > armhf are the same, I would not know :(
> 
> So the architecture is basically "arm7" and QEMU emulates an 'arm7'
> architecture.
> 
> hard float vs soft float is essentially referring to how the guest
> software and kernel is compiled - whether it was built to use software
> emulated floating point vs hardware floating point.
> 
> So as long as you tell QEMU to emulate a CPU model for arm7 that can
> do hard float, then QEMU can support guests OS which use hard or soft
> float.
> 
> This is a long winded way of saying that in terms of representing
> architectures in libvirt and libosinfo, we don't need a distinction
> between hard & soft float arm. They're all just arm7.
> 
> The distinction is more akin having an i686 guest, which is compiled
> to assume a CPU model with SSE instructions. This is a concept we
> don't track in libosinfo right now. We could, but I'd just ignore
> until its critical (if ever).

Yup, makes a lot of sense, thanks for the detailed explanation!

Christophe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/libosinfo/attachments/20170713/cef4eea0/attachment.sig>