Re: Promised patch 2.1.108

dalecki (dalecki@cs.net.pl)
Wed, 8 Jul 1998 15:37:25 -0700 (MST)


On Wed, 8 Jul 1998, Philip Blundell wrote:

> dalecki@cs.net.pl said:
> > -ifeq ($(CONFIG_NET),y)
> > +ifeq ($(CONFIG_INET),y)
> > L_OBJS += Space.o net_init.o loopback.o
> > endif
>
> That's not really correct. It should be possible to build with devices but no
> TCP/IP in an ideal world, say on an Appletalk only machine. That's a pretty
> unlikely case though I grant you. What you should really do is have a top
> level "Network device support" option and make those files depend on it.

I don't quite agree. There is already an CONFIG_NETDEVICE option, which I
didn't mess with, since anything else from CONFIG_UNIX depends on
CONFIG_INET beeing enabled.

>
> > -ifdef CONFIG_NET
> > +ifdef CONFIG_INET
> > O_OBJS += dev.o dev_mcast.o dst.o neighbour.o rtnetlink.o utils.o
>
> Same there. This undoes all the work that was done to make the net core less
> dependent on INET.

No. This is mainly a workaroud for an binutils, or to be more exact: ar
bug, which still includes objects with cross refferences into each other,
even in the case that both of them are not reachable from the main
execution path. I have already described this situtaion some time ago in
an exhausting manner. (Please see the script I posted tose times, which I
used to detect unrefferenced global objects in the kernel.)

Marcin
=========================================================================
In real life: System Programmer at AIS AXON GmbH

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu