Semantics of kernel ABI-nums

Andy Whitcroft apw at
Tue Nov 15 11:27:52 UTC 2011

On Sat, Nov 12, 2011 at 01:55:12PM +0000, Colin Ian King wrote:
> On 11/11/11 09:47, Stefan Bader wrote:
> >>> linaro-mx51, linaro-omap, and linaro-xexpress? Is there documentation on how
> > I have to agree with Paolo that I have no idea about a n900 flavour ever done by
> > us. And the things that get referenced strongly point to linaro.
> >
> > For the arm topic branches we _have_ been doing, the ABI number usually starts
> > at a different base. I am not sure I can remember all the reasoning. I think
> > without that there was some naming clash of some packages. And at least in some
> > cases the arm kernel of the next release was the same version as the previous
> > which would not work without the offset. The ABI numbers we know of should be
> > all documented in [1].
> >
> > -Stefan
> >
> > [1]
> >
> Makes me wonder if we should consider building into the kernel the name
> of originating git repo in the same way we've added
> /proc/version_signature in fs/proc/version_signature.c

I don't think we can guarentee to have that information at the time it
is built, ie. it would be something the configurer of the package would
have to get right.

There is a VCS-Git: tag in the control file, that would be the obvious
place for the reference as that ends up on the machine when the package
is installed.


More information about the kernel-team mailing list