Subject: Re: "Fixing" the SB_MAX option and binary kernel kits.
To: Simon Burge <simonb@netbsd.org>
From: Ignatios Souvatzis <is@jocelyn.rhein.de>
List: tech-kern
Date: 04/21/1999 21:49:13
On Wed, Apr 21, 1999 at 01:34:11PM +1000, Simon Burge wrote:
> Folks,
> 
> <sys/socketvar.h> includes "opt_sb_max.h", and then further down defines
> SB_MAX if it isn't already defined.  This has the effect of causing
> 40-odd files to be recompiled if one simply added "options SB_MAX=foo"
> to their config file.
> 
> The only place SB_MAX is used is to set the value of the variable
> sb_max in kern/uipc_socket2.c (with a comment saying "patchable").
> I was simply going to remove the inclusion of "opt_sb_max.h" in
> <sys/socketvar.h> and add it to the top of kern/uipc_socket2.c, but then
> started wondering where we're headed overall with "binary kits" for
> kernels.  Since every kernel builds kern/uipc_socket2.c, would it make
> more sense to initalise sb_max in conf/param.c?

sounds reasonable to me...
	-is