Hi Stefan,, On Wed, 18 Sep 2019 at 15:58, Stefan Schmidt <stefan@xxxxxxxxxxxxxxxxxx> wrote: > > It fails on my machine and likely fails on others, better > check for it. > --- > send.c | 4 +--- > 1 file changed, 1 insertion(+), 3 deletions(-) > > diff --git a/send.c b/send.c > index 97c566d..d026f6a 100644 > --- a/send.c > +++ b/send.c > @@ -53,11 +53,9 @@ static int really_send(int sock, const struct iface *iface, > pkt_info->ipi6_ifindex = iface->ifindex; > memcpy(&pkt_info->ipi6_addr, iface->ifaddr_src, sizeof(struct in6_addr)); > > -#if 1 > -//#ifdef HAVE_SIN6_SCOPE_ID > +#ifdef HAVE_SIN6_SCOPE_ID This HAVE_ stuff is from autoconf where I took the code (radvd), so this will mostly end in #ifdef 0 because we use meson now and don't ask me how to check on this in meson but it's simple I think. :-) I think we can remove this code and see if we break something. The point here radvd is ported to BSD, Linux, etc. and this is some system thing. _Maybe_ it is really not necessary for Linux systems _because_ we set above "pkt_info->ipi6_ifindex = iface->ifindex;" which is maybe enough on a Linux system... but it could also be to solve some UAPI messup.... just remove it. In a scope of link-local we must set the interface somehow, I am not sure how the correct API works here, as I said... look if it makes problems. Sorry but it means: I am too lazy too look it up why this is necessary in radvd (git blame is maybe enough?). Note: That it compiles on my Linux machine and not on your Linux machine, means there is something in Linux API headers. - Alex
