Re: [RFC PATCH] kbuild: Do not select symbols with unmet dependencies

[Catalin Marinas <catalin.marinas@arm.com>]

On Wed, 2009-08-12 at 14:36 +0200, Arnd Bergmann wrote:
> On Wednesday 12 August 2009, Catalin Marinas wrote:
> > The "select" statement in Kconfig files allows the enabling of options
> > even if they have unmet direct dependencies (i.e. "depends on" expands
> > to "no"). Currently, the "depends on" clauses are used in calculating
> > the visibility but they do not affect the reverse dependencies in any
> > way.
> > 
> > The patch introduces additional tracking of the "depends on" statements
> > and does not allow selecting an option if its direct dependencies are
> > not met, also printing a warning.
> > 
> > Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> > Cc: Sam Ravnborg <sam@ravnborg.org>
> 
> I guess this will change the behaviour of a number of subsystems,
> likely causing unexpected regressions. I think your change
> makes sense, but we need to be much more careful.

It would indeed cause regressions, that's why I'm only asking for
comments currently.

> Can you extract a list of configuration symbols that are
> impacted by your patch?

I can generate a list with allyesconfig but it looks like it breaks some
common usage in Linux. For example, the VIDEO_* entries in
drivers/media/video/Kconfig under the "Encoders/decoders and other
helper chips" menu automatically inherit a dependency on
!VIDEO_HELPER_CHIPS_AUTO. This option is enabled to allow other config
options to select whatever they need. But it would fail with my patch
because of the direct dependency of the VIDEO_* options on
!VIDEO_HELPER_CHIPS_AUTO.

It needs a bit more thinking here and maybe writing something like:

menu "..." if !VIDEO_HELPER_CHIPS_AUTO

rather than

menu "..."
	depends on !VIDEO_HELPER_CHIPS_AUTO

though the parser (after modifying the zconf.y to handle this) seems to
consider both dependencies at the same level

-- 
Catalin