Re: [PATCH 2/2] Enable 'make CONFIG_FOO=y oldconfig'

Hi,

On Wed, Aug 10, 2011 at 1:59 PM, David Woodhouse [off-list ref] wrote:

On Wed, 2011-08-10 at 13:44 -0400, Arnaud Lacombe wrote:

quoted

Exactly my point, you just successfully created an half-backed config
which is different than what Aunt Tillie wanted you to generate. This
should be an hard error, same for "all*.config", not to mention that
the error message is far from being helpful.

You are whining about something that has been true of the kernel config
system for at least the last 16 years that I've been working on it,

s/16/6/; the all.config logic has been added in:

commit 90389160efc2864501ced6e662f9419eb7a3e6c8
Author: Roman Zippel [off-list ref]
Date:   Tue Nov 8 21:34:49 2005 -0800

    [PATCH] kconfig: preset config during all*config

so, at best, this buggy behavior is ~ 6 years old. Before that, I'd
assume that the internal namespace was not accessible by any other
mean than the front-ends.

You're *right*, of course, but you're getting on my tits by whining
about it only *now*, in this context.

well... sorry for your tits, 'hope you're enjoying it ;-)

At least I have offered *an* error
message reporting that the request was not honoured, which is a whole
lot better that we've been used to.

s/error/warning/, but indeed, that's a step forward.

Please, if this offends you then by all means go and fix it. A sane way
of handling dependencies would give a way to say "do what you need to do
in order to enable CONFIG_SATA_MV", and should remove the abomination of
'select', which was introduced purely to work around that lack.

But none of that is directly relevant in *this* thread.

to paraphrase you, I'd say, this might looks "cute but might give
behavior that people will come to depend on in their scripts and then
we take it away again", "that's why I'd kind of like to see it done
*once*, *properly*".

 - Arnaud

--
dwmw2