Re: CVS commit: src/sys/kern

[David Laight <david%l8s.co.uk@localhost>]

On Tue, Jan 31, 2012 at 07:50:26PM +0000, Alexander Nasonov wrote:
> Joerg Sonnenberger wrote:
> > On Tue, Jan 31, 2012 at 07:32:52PM +0000, Alexander Nasonov wrote:
> > > #define sizeof_fam(s, m, n) (sizeof(s) + sizeof(((s *)NULL)->m[0]) * (n))
> > 
> > That's still not necessarily optimal, depending on the padding rules of
> > the platform. You want to do offsetof(s, m[0]) + n * sizeof((s*)NULL->m[0]).
> 
> I'm aware of this but I was merely following examples from the standard.
> 
> More specifically I was following DR 282 (flexible array members &
> struct padding) which was incorporated into TC2.
> 
> http://std.dkuug.dk/jtc1/sc22/wg14/www/docs/dr_282.htm
> 
> I can add a comment about padding to sizeof_fam. If there is no padding,
> my and your expression should return same values.

And if there is padding, and it is after the last field (and a few
other clauses) your scheme will allocate a longer buffer than needed.

The advantage of Joerg's is that you don't have to 'know' the type of
the member.

Sudden barin explosion - how about (untested):
#define sizeof_var_struct(s, m, c) \
        offsetof(s, m[0]) + (c) * (offsetof(s, m[1]) -  offsetof(s, m[0]))

        David

-- 
David Laight: david%l8s.co.uk@localhost