[U-Boot] [PATCH 2/6] mmc: complete the unfinished move of CONFIG_MMC

[Tom Rini <trini@konsulko.com>]

On Tue, Dec 06, 2016 at 03:25:46PM +0900, Masahiro Yamada wrote:
> Hi Tom,
> 
> 2016-12-06 1:02 GMT+09:00 Tom Rini <trini@konsulko.com>:
> > On Sun, Dec 04, 2016 at 10:23:10PM +0900, Masahiro Yamada wrote:
> >
> >> Commit 7a777f6d6f35 ("mmc: Add generic Kconfig option") created
> >> a Kconfig entry for this option without any actual moves, then
> >> commit 44c798799f66 ("sunxi: Use Kconfig CONFIG_MMC") moved the
> >> instances only for SUNXI.
> >>
> >> We generally do not like such partial moves.  This kind of work
> >> is automated by tools/moveconfig.py, so it is pretty easy to
> >> complete this move.
> >>
> >> This commit was created as follows:
> >>
> >> [1] remove "depends on ARCH_SUNXI || SANDBOX" and
> >>     reword the prompt and help message (copied from Linux)
> >> [2] run 'tools/moveconfig MMC'
> >>
> >> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> > [snip]
> >>  801 files changed, 630 insertions(+), 197 deletions(-)
> >
> > OK.  I think this means that we want MMC to be either default y or
> > default y if ARM || POWERPC || SANDBOX, to cut down on the size here.
> 
> 
> Nit.   s/POWERPC/PPC/
> (Unfortunately, U-Boot historically uses a different
> config for PowerPC.)

Ah right.

> I tried your suggestion.
> It is true that 625 touched defconfigs were decreased to 389,
> but do you think this is a really good idea?
> 
> Theoretically, nothing of MMC
> is related to ARM, PowerPC, Sandbox.
> 
> I think this is shortening defconfigs
> for the purpose of shortening.

Yes, it is for the purpose of shortening the configs.

> BTW, do you know that
> Kbuild subsystem is discussing a new keyword "imply" ?
> 
> https://patchwork.kernel.org/patch/9422463/
> 
> 
> I am thinking it, if accepted, will be a better solution than
> such unrelated "depends on <ARCH>" or "depends on <SOC>".

Adding "imply" could help here I think, yes.  The problem is that I want
to make the defaults in Kconfig produce a reasonable defconfig rather
than require each new defconfig to require a lot of information to
produce a reasonable binary.  Looking at the imply examples I think that
yes, we could probably get away with doing things like:
arch/Kconfig:
config ARM
	bool "ARM Architecture"
	...
	imply MMC
	imply DISPLAY_BOARDINFO
	imply DISPLAY_CPUINFO

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20161206/b89b465d/attachment.sig>