Re: [PATCH 2/2] video: omap2: Compile omap2 support only when needed

[Tomi Valkeinen <tomi.valkeinen-l0cyMroinI0@public.gmane.org>]

On Tue, 2011-06-21 at 10:18 +0530, Tushar Behera wrote:
> Hi,
> 
> On Monday 20 June 2011 06:36 PM, Tomi Valkeinen wrote:
> > On Mon, 2011-06-20 at 16:16 +0530, Tushar Behera wrote:
> >> Currently display support for omap2 is selected by default and
> >> it gets built for all the configurations.
> >>
> >> Instead of it being a built-in feature, it's compilation should
> >> depend on the config option CONFIG_FB_OMAP2.
> >
> > No, I don't think so. omap2 directory contains vram, vrfb and omapdss,
> > all of which can be used without omapfb driver. vram and vrfb can be
> > even used without omapdss driver.
> Even if I build the kernel with i386_defconfig, I get some compiled 
> files within drivers/video/omap2.
> 
> $ make ARCH=x86 i386_defconfig O=out_dir
> $ make ARCH=x86 O=out_dir
> 
> $ ls out_dir/drivers/video/omap2
> built-in.o  displays  modules.builtin  modules.order
> 
> IMHO, drivers/video/omap2/ should not be compiled if the kernel is not 
> built for omap2.

Ok. Yes, that's a known "problem". We could have a OMAPx check there,
but then again, the driver is a driver for the DSS HW, which could, at
least in theory, used in other SoCs.

I don't think any driver should normally depend on ARCH_something or
MACH_something.

I think this is more of a "problem" with the build system. The same
thing can be seen with, for example, drivers/video/backlight/ and
drivers/video/display/ directories. In practice the created files do not
affect the kernel in any way, as far as I see.

 Tomi