Re: [PATCHv2 1/1] RTC:driver for Cortina's SOC

[Arnd Bergmann <arnd@arndb.de>]

On Tuesday 12 May 2015 11:05:34 Paul Bolle wrote:
> On Mon, 2015-05-11 at 18:25 +0200, Hans Ulli Kroll wrote:
> > --- a/drivers/rtc/Kconfig
> > +++ b/drivers/rtc/Kconfig
> 
> > +config RTC_DRV_GEMINI
> > +     tristate "Gemini SoC RTC"
> > +     help
> > +       If you say Y here you will get support for the
> > +       RTC found on Gemini SoC's.
> > +
> > +       This driver can also be built as a module. If so, the module
> > +       will be called rtc-gemini.
> 
> This has, by the look of it, very little dependencies, probably just
> RTC_CLASS. So I think that this symbol can be enabled in allyesconfig,
> allmodconfig, and randconfig for most architectures. (I just tested this
> - what a novel idea! - and it will indeed be set by allyesconfig and
> allmodconfig on x86_64.)
> 
> > --- /dev/null
> > +++ b/drivers/rtc/rtc-gemini.c
> 
> > +#include <mach/hardware.h>
> 
> Except for arm, builds using a .config containing
>     CONFIG_RTC_DRV_GEMINI=[my]
> 
> will fail here. (*/mach/hardware.h is mainly used for arm.) That is
> bound to generate, well, fanmail for you. Perhaps it's better if
> RTC_DRV_GEMINI depends on, say, ARCH_GEMINI.

Well spotted, thanks for the report.

I have a patch that will turn mach-gemini into multiplatform, and
at that point it will also fail there. The correct solution I think
is to move the GEMINI_RTC_SECOND etc definitions into rtc-gemini.c
itself and remove the #include.

It still makes sense to have a dependency, but I'd express it
as

	depends on ARCH_GEMINI || COMPILE_TEST

in order to let the driver get compiled for an allyesconfig kernel
on all architectures, but not appear as an option for normal builds
that do not set COMPILE_TEST.

	Arnd