Re: CVS commit: syssrc/sys/sys

["Perry E. Metzger" <perry%wasabisystems.com@localhost>]

Todd Vierling <tv%wasabisystems.com@localhost> writes:
> Yes, I know.  However, PR kern/13369 was created because of kernel source
> code which used bcopy under the assumptions listed in bcopy(3), since we
> don't have a bcopy(9).

Well, they made the wrong assumption.

If you feel we need it, we can add a bcopy(9) that just says "do not
use bcopy. it is a legacy interface in the kernel. use mem*".

> 1. reintroduce ovbcopy() as a macro,

No, we don't want more of it.

> 2. eliminate all uses of bcopy() in the kernel (incl. imported sources),

We are trying to do that. We should do it more agressively.

> I'd rather see most instances of bcopy() eliminated (partial #2) here, and
> then Just Deal with the slight overhead of the bcopy() compatibility
> #define.

The problem is that you really do want to know when people are
differentiating bcopy vs. ovbcopy in kernel code. You don't want to
encourage that at all. Document, yes. Remove bcopys, yes. Do not,
however, make bcopy do overlapping regions in the kernel even as a
stopgap.

--
Perry E. Metzger                perry%wasabisystems.com@localhost
--
NetBSD Development, Support & CDs. http://www.wasabisystems.com/