Re: [PATCH -mm] ssb: Make pcmciahost depend on PCMCIA=y

[Paul Mundt <lethal@linux-sh.org>]

On Wed, Sep 12, 2007 at 12:09:09PM +0200, Michael Buesch wrote:
> On Wednesday 12 September 2007 04:11:00 Paul Mundt wrote:
> > SSB uses a bool (SSB_PCMCIAHOST_POSSIBLE) to determine whether to
> > build in PCMCIA support or not, as the PCMCIA host code itself is
> > also only a bool, make SSB_PCMCIAHOST_POSSIBLE depend on PCMCIA=y.
> > 
> > Without this, SSB_PCMCIAHOST_POSSIBLE evaluates to y when PCMCIA
> > is built as a module, which results in link errors due to the
> > pcmcia_access_configuration_register() accesses, where the symbol
> > is only defined in a module.
> > 
> > --- linux-2.6.23-rc4-mm1.orig/drivers/ssb/Kconfig	2007-09-11 15:15:52.000000000 +0900
> > +++ linux-2.6.23-rc4-mm1/drivers/ssb/Kconfig	2007-09-12 10:51:53.000000000 +0900
> > @@ -37,7 +37,7 @@
> >  
> >  config SSB_PCMCIAHOST_POSSIBLE
> >  	bool
> > -	depends on SSB && PCMCIA && EXPERIMENTAL
> > +	depends on SSB && PCMCIA=y && EXPERIMENTAL
> >  	default y
> >  
> >  config SSB_PCMCIAHOST
> > 
> > 
> 
> There we go. The usual SELECT dependency hell again...
> Would changing SSB_PCMCIAHOST_POSSIBLE to tristate also fix it?
> What would be the sideeffects?
> 
I tried that first, if you do that you have to change the default to
SSB && PCMCIA, and then anything that depends on it also has to be a
tristate. That worked ok for SSB_PCMCIAHOST, but it didn't work ok for
the b43 wireless + PCMCIA, which is why I opted for the PCMCIA=y thing
instead, which makes sure that SSB_PCMCIAHOST can't be enabled if PCMCIA
is modular.