Peter Memishian wrote:
> > I'm not sure I understand what you're suggesting.
> >
> > Are you suggesting that I just drop this issue on the floor, and abandon
> > the attempt to clean this up?
>
> I'm suggesting that we placate lint through whatever means necessary, but
> keep the existing signature and semantics of the macros. When we have
> more time, other things like having the macros check for malformed
> messages could be investigated. That work would need to include either
> collecting data demonstrating that adding those checks will still yield a
> stable DEBUG system, or fixing bugs so that the system remains stable.
> Only with the checks in-place would I feel comfortable changing the macros
> to return unsigned values, but even then I question the point, as I'm
> unaware of a pressing need to have a single mblk with 2^31 bytes of data.
>
So, doing the math in unsigned integers, and then casting back to a
signed type would be OK, right? E.g.
#define MBLKL(mp) (intptr_t)((uintptr_t)(mp)->b_wptr -
(uintptr_t)(mp)->b_rptr))
Despite what the man page says, I don't want to cast to an (int),
because that would represent loss of data, where the old code didn't
lose it.
-- Garrett
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
