Re: [PATCH 0/1] mspec driver: compile error

[Andrew Morton <akpm@osdl.org>]

On Wed, 08 Nov 2006 18:45:30 +0900
Fernando Luis Vázquez Cao <fernando@oss.ntt.co.jp> wrote:

> On Wed, 2006-11-08 at 10:42 +0100, Jes Sorensen wrote:
> > Fernando Luis Vázquez Cao wrote:
> > > On Tue, 2006-11-07 at 13:35 -0800, Andrew Morton wrote:
> > > The problem occurs because i386 (as expected) does not define
> > > IA64_UNCACHED_ALLOCATOR. I thought that making the select expression
> > > depend on IA64 as shown below might silence allmodconfig:
> > > 
> > >   select IA64_UNCACHED_ALLOCATOR if IA64
> > > 
> > > But my guess was wrong and the same warning appeared. It seems that "if"
> > > expressions do not prevent allmodconfig from checking the symbol
> > > indicated by the select the "if" is conditioning. By the way, is this
> > > the expected behaviour? If so, we need to get rid of the reverse
> > > dependency, modify the "depends on" line accordingly, and make
> > > IA64_UNCACHED_ALLOCATOR selectable. I may be missing the whole point so
> > > please correct if I am wrong.
> > 
> > This patch is a bad solution as it requires people to manually select
> > the uncached allocator. It should be enabled automatically by MSPEC,
> > not the other way round.
> > 
> > Given that MSPEC is clearly marked as depending on IA64, it seems bogus
> > for i386 allmodconfig to barf over it and the problem should be fixed
> > there instead IMHO.
> Agreed. That is why I asked if that was allmodconfig's expected
> behaviour. Andrew?
> 

kconfig's `select' isn't very smart.  This is one of the reasons why one
should avoid using it.