Re: [v4l-dvb-maintainer] Re: [PATCH 08/18] V4L/DVB (4734): Tda826x: fix frontend selection for dvb_attach

[Trent Piepho <xyzzy@speakeasy.org>]

On Sun, 15 Oct 2006, Adrian Bunk wrote:
> > > To be honest and after looking deeper at it, I don't like this
> > > CONFIG_DVB_FE_CUSTOMIZE approach at all since it adds that much
> > > complexity for not much gain.
> > Yes, it adds some complexity. The gain, however, is to allow having
> > smaller kernel size on embedded systems and DVR using MythTV or Freevo.
> > There's a similar feature for V4L (Autoselect pertinent
> > encoders/decoders and other helper chips), that allows selecting just
> > the needed stuff.
>
> I understand this, but I still don't like the solution.
> But unless I can send a patch with a better solution, this is nothing I
> can complain about...

There used to be a different way of selecting which front-ends get
built/loaded, that was recently replaced with DVB_FE_CUSTOMISE.  It was
even more complex and only worked for a few card drivers that specially
supported it.

What is nice about the new DVB_FE_CUSTOMISE way is that there is very
little code needed.  There is a boilerplate #if in the font-end header file
for the _attach function, and that's it.

> There are two things that might be changed even now:
>
> The "default m if DVB_FE_CUSTOMISE" is something I do not like that
> much, especially considering that embedded users caring about kernel
> size are likely to use CONFIG_MODULES=n.

A front-end could default to the highest level of the card drivers which
utilizes it.  I'm not sure how to express that in a Kconfig file, and it
would be more work to maintain.  One has to go into the FE customisation
menu to turn on DVB_FE_CUSTOMISE anyway, at which point it seems like there
is little difference between disabling the drivers one doesn't want vs
enabling the drivers one does.