[linux-dvb] Re: refactoring

[Holger Waechtler <holger@qanu.de>]

Kenneth Aafløy wrote:

> On Friday 01 October 2004 13:46, Gerd Knorr wrote:
>  
> > On Fri, Oct 01, 2004 at 10:59:52AM +0100, Andrew de Quincey wrote:
> >    
> > > On Friday 01 Oct 2004 08:49, you wrote:
> > >      
> > > > For the cx22702 you should be able to simply take the code from the
> > > > v4l cvs ...
> > > >        
> > > Doh! I wish I'd remembered that.. I've already done that one.
> > >      
> > - rework Kconfig: don't ask the users which FE they want to build, but
> >   enable the ones needed via "select" [1].  That makes it easier for the
> >   users (you can't forget to enable the frontend, and you don't need to
> >   know which one your card has) and also avoids the source being
> >   cluttered up with "#ifdef CONFIG_DVB_<FRONTEND>".
> >    
>  
> > config VIDEO_CX88_DVB
> >        tristate "DVB Support for cx2388x based TV cards"
> >        depends on VIDEO_CX88 && DVB_CORE
> >        select DVB_CX22702
> >        ---help---
> >        <help text here>
> >    
> Please, don't use select, that will force a user with a Nexus-S with the stv0299
> frontend to have all other frontends compiled in. This will create bloated
> drivers, and actually would hurt the embedded platform (Holger argued that the
> i2c core was huge, don't think he wants to be forced to include more frontends)
>  
stripped-down frontend drivers are a few 100bytes ... 2.5kB which is 
kind of tiny compared to the i2c core. Most card drivers will get linked 
together with less than 3 frontend drivers.

> Use the default construct instead, which makes it possible to deselect, but
> 'reminds' the user that it's probably a good idea to select this item.
>  
This pushes the responsibility of analyzing the hardware to the user 
back again. Since most cards don't have printed on their metal can which 
demodulator is inside this is quite a hard task for many users (and I 
admid that I would not accept this as user too - ).

Holger