Lars Ingebrigtsen <larsi@gnus.org> schrieb am So., 14. Mai 2017 um
12:28 Uhr:

> Philipp Stephani <p.stephani2@gmail.com> writes:
>
> > +                    case AF_INET:
> > +                      port = ((struct sockaddr_in6 *) (struct sockaddr
> *) &sa1)->sin6_port;
> > +                      has_port = true;
> > +                      break;
> > +                    case AF_INET6:
> > +                      port = ((struct sockaddr_in *) (struct sockaddr
> *) &sa1)->sin_port;
> > +                      has_port = true;
>
> Aren't these two cases in reverse?  If it's AF_INET6, it's an in6
> struct, not the other way around.
>

Oops, thanks!


>
> Not that it matters, since (as Eli said) the sizes of the first elements
> in the structs are identical...
>
>
I don't know whether we can rely on that (i.e. whether Posix guarantees it)
or whether it's an implementation detail.
(This also needs __attribute__((may_alias)), type punning, or memcpy due to
aliasing.)