On Tue, Jul 03, 2012 at 05:35:34PM +0200, Guennadi Liakhovetski wrote:
> On Tue, 3 Jul 2012, Mark Brown wrote:

> > But in general I'm not sure we're requiring that .configs
> > that people can generate do more than build.

> Well, you could have a kernel configuration with REGULATOR=y for some 
> reason with fixed regulator disabled. It would work with earlier kernels. 
> Now it would begin to fail... Do we want to allow that?

Well, it does happen from time to time; for example when moving from
forcing some GPIOs on to making them software managed.  It's only going
to affect boards which were previously using regulators but not the
fixed one, I'm not sure if that is common for these systems...

> Well, I suspect, other boards must have had the same exact problem, as 
> they added regulator support. It seems some boards to that, but not all...

Handling this really needs some kind of "set on upgrade" thing but I'm
not sure that's worth the trouble, especially since it's not that common.