From owner-freebsd-hackers  Wed Apr 12 12:41:01 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id MAA21552
          for hackers-outgoing; Wed, 12 Apr 1995 12:41:01 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id MAA21537
          for <hackers@FreeBSD.org>; Wed, 12 Apr 1995 12:40:36 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id FAA24223; Thu, 13 Apr 1995 05:37:54 +1000
Date: Thu, 13 Apr 1995 05:37:54 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199504121937.FAA24223@godzilla.zeta.org.au>
To: elh@p5.spnet.com, hackers@FreeBSD.org
Subject: malloc
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

>	a while ago i encountered an 'application almost bug' wherein
>	sbrk(1) returns (old_mem+1).  Some other unix's
>	(e.g., SunOS) return a double word aliged value.

I think sbrk() should do what it is told, and callers of sbrk() should
do any necessary rounding and aligning.  malloc() should align page-sized
objects to a page boundary and doubleword alignment is irrelevant for
this.  malloc() doesn't do this.  Try:

	printf(""%p\n", malloc(0x1000));
	printf(""%p\n", malloc(0x1000));
	printf(""%p\n", sbrk(1));
	printf(""%p\n", malloc(0x1000));

We should support i486 alignment checking so that misaligned pointers
can be found easily.  Last time I enabled it (4-5 years ago) it found
too many problems, so I stopped using it.  gcc generated misaligned
"rep movsd" instructions for structs that don't require word alignment...

Bruce