On Mon, Nov 16, 2015 at 4:40 AM, Masahiro Yamada
[off-list ref] wrote:

2015-11-05 22:40 GMT+09:00 Linus Walleij [off-list ref]:

quoted

On Thu, Nov 5, 2015 at 12:15 PM, Masahiro Yamada
[off-list ref] wrote:

quoted

ARCH_SUNXI selects RESET_CONTROLLER.
The dependency "depends on RESET_CONTROLLER" is already met.

Signed-off-by: Masahiro Yamada <redacted>

NACK, this pin controller needs that as dependency, I don't worry
that it's overspecified.

They both do this:
#include <linux/reset.h>

And the object should specify its dependencies, not rely on them
being implicitly available.

The dep is there for human readers too, not just machines. And to
them, this is helpful information.

How strict we should follow this rule?

Many drivers depend on OF, but do they all have "depends on OF"?

I think they should, and if they are in my subsystems I consider
it a bug if they don't have this dependency. Also it is a very common
reason why test compiles fail.

Yours,
Linus Walleij