On Fri, Oct 26, 2007 at 06:33:17PM +0100, Iain Hibbert wrote:
> On Tue, 23 Oct 2007, Tom Spindler wrote:
> 
> > Given that many of these pieces of sh^W^W^Wlovely source files
> > don't actually bother using 'struct queue', and far more often
> > monkey around with the affected structures themselves - it might
> > be better to just embed local copies of the routines into the
> > affected sources. (As the defs are all of five lines of code each,
> > I don't think it'd be a horrible burden.)
> 
> Having looked at it in a bit more depth, I think that if no testers can be
> found for ISO/EON/TPIP then just hardcoding it is the cheapest option and
> is least likley to break anything not already broken, eg:
> 
> --- /usr/src/sys/netiso/iso_pcb.c	2007-07-20 22:24:52.000000000 +0100
> +++ iso_pcb.c	2007-10-26 12:30:30.000000000 +0100
> @@ -118,7 +118,13 @@
>  		return ENOBUFS;
>  	isop->isop_head = head;
>  	isop->isop_socket = so;
> -	insque(isop, head);
> +
> +	/* insert at head of pcb list */
> +	isop->isop_next = head->isop_next;
> +	isop->isop_prev = head;
> +	head->isop_next = isop;
> +	isop->isop_next->isop_prev = isop;
> +

Please, don't do that.  Use queue(3) macros.  They work, developers
believe they work, and LIST_INSERT_HEAD() will reduce the lines of code
and comments in that example from five to one.

Dave

-- 
David Young             OJC Technologies
dyoung@ojctech.com      Urbana, IL * (217) 278-3933 ext 24