Re: [PATCH] parport_pc to use pnpbios_register_driver()

[Thomas Hood <jdthood@mail.com>]

On Mon, 2001-11-12 at 03:30, Kai Germaschewski wrote:
> On Mon, 12 Nov 2001, Keith Owens wrote:
> > Does this combination make sense?  If you are building a pnpbios driver
> > into the kernel then the configuration should force pnpbios support to
> > be built in as well.  We don't allow this combination for things like
> > scsi or ide, they require the common support to be built in if any
> > drivers are built in.  IMHO this problem should be fixed in .config,
> > not in driver source.
> 
> The affected drivers usually support different interfaces to the hardware, 
> like ISA, ISAPnP, PCI, so it makes sense to build them w/o ISAPnP support.

True.

> For instance, people may want to have built-in serial support (w/o ISAPnP 
> support is fine, because they've only got a standard SuperI/O chip), but 
> also modular isapnp and sound driver support, so they can demand load 
> sound support if necessary.

Fine.

> Making ISAPnP mandatory only because you build a driver which supports
> ISAPnP hardware is not an option IMO. This would force people to build
> ISAPnP support who don't even have an ISA bus (only because the driver for
> their PCI card also supports an ISAPnP variant).

You misunderstand.  Obviously we don't want to force anyone
to build in isa-pnp or pnpbios support.  What Keith is saying
is that IF isa-pnp code is enabled in a driver then the isa-pnp
driver should also be built.  Furthermore if isa-pnp code is
enabled in an _integral_ (i.e., non-modular) driver, then the
isa-pnp driver should also be built integrally.  The combination
of integral-driver-with-isa-pnp-support-enabled and modular
isa-pnp driver should be disallowed by the configurator.

Thomas