[ blah - this got lost in my inbox ]

Jason R Thorpe wrote:

> On Tue, Feb 13, 2001 at 11:58:37PM +1100, Simon Burge wrote:
> 
>  > Index: rominfo.h
>  > ===================================================================
>  > RCS file: /cvsroot/syssrc/sys/arch/hp300/stand/common/rominfo.h,v
>  > retrieving revision 1.1
>  > diff -d -p -u -r1.1 rominfo.h
>  > --- rominfo.h	1997/02/04 03:52:49	1.1
>  > +++ rominfo.h	2001/02/13 12:54:17
>  > @@ -47,7 +47,7 @@
>  >  struct jmpvec {
>  >  	short op;	/* jmp instruction */
>  >  	long  addr;	/* address */
>  > -};
>  > +} __attribute__((__packed__));
> 
> This one is obviously correct.

Cool.

>  >  struct rominfo {
>  >  	char p1[0xDC0];
>  > @@ -65,4 +65,4 @@ struct rominfo {
>  >  	char  p6;		/* ??				(FFFFFEDB) */
>  >  	long  msus;		/* ??				(FFFFFEDC) */
>  >  	struct jmpvec jvec[48];	/* jump vectors			(FFFFFEE0) */
>  > -};
>  > +} __attribute__((__packed__));
> 
> You might want to see how the a.out compiler padded this one (not sure
> what happens to the "char" -- I'm pretty sure it's packed in a.out, but
> you might want to make sure...)

Judging by the comment it is packed, but I don't have any m68k a.out
boxes on which to check.  Can anyone else verify this before I commit
the change?  Something like:

	struct foo {
		char c1;
		int i1;
	} /* __attribute__((__packed__)) */;

	main()
	{
		printf("sizeof = %d\n", sizeof(struct foo));
		exit(0);
	}

should print "5" for a packed structure and "6" (or "8"?) for an padded
structure.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD CDs, Support and Service:    http://www.wasabisystems.com/