svn commit: r237883 - in head/sys/arm: at91 conf

[Warner Losh]

On Jul 1, 2012, at 9:27 AM, Marcel Moolenaar wrote:

> 
> On Jun 30, 2012, at 11:56 PM, Warner Losh wrote:
> 
>> Author: imp
>> Date: Sun Jul  1 06:56:41 2012
>> New Revision: 237883
>> URL: http://svn.freebsd.org/changeset/base/237883
>> 
>> Log:
>> Create a pseudo-lint kernel for all at91 SoCs.  This kernel will not
>> currently boot, but will serve as a good linting.  make universe could
>> now be altered to skip building all the other at91 kernels...
> 
> BTW: I've implemented the LINT kernel for ARM. It combines all
> SoCs. It does have a lot of duplicate definitions, but by not
> having the linker fail on that, you get a successful build of
> something we already understand does not boot. It's good coverage
> with a single kernel and can help to bring the "make universe"
> time down by only building LINT for ARM.

I was thinking of adding NOUNIVERSE tags to the kernels that we didn't want routinely built.  I'd rather have a single ARM kernel that can be built for testing purposes.  Don't like the multiple defined error being suppressed, but short of some uglyish macros, I can't think of a better way.  I've written those 'ugly' macros for my multi-board work, and plan on re-using them for the multi-soc work I intend to do to replace the current "selected too late" SoC support for Atmel.  I was thinking we could expand the current set of platform/MD calls (initarm, etc), wrap them in some macros so they could all be compiled together.  Not sure if you did this or not...

> I can port that to FreeBSD. Shall I make some patches for people
> to look at?

Sure.  I'd love to see it.  I'd be happy to preview any partial work if you want early feedbac.

Warner