Actually… we don't need #3931 at all… The if-return in https://github.com/RIOT-OS/RIOT/pull/3935/files#diff-74976d45a42eaa063e3561b5fd36f615R536-538 assures already, that the function is exited before the next hop determination of standard IPv6 is called....

Some fine reviewing I did there 😞

The if-return is only for the error-case. #3931 is required - I actually ran into problems with this.

Okay let's write the code from #3931 with the assumption that all #ifdefs are true:

    (void)pkt;
    iface = gnrc_sixlowpan_nd_next_hop_l2addr(l2addr, l2addr_len, iface, dst);
    if (iface <= KERNEL_PID_UNDEF) {
        return iface;
    }
    if (iface <= KERNEL_PID_UNDEF) {
        iface = gnrc_ndp_node_next_hop_l2addr(l2addr, l2addr_len, iface, dst, pkt);
    }

Do you see what I'm getting at? Two times the same check, the first results in returning. So how could iface be <= KERNEL_PID_UNDEF in the code introduced in #3931?

(except if it was given on input).

It is about the case where iface > KERNEL_PID_UNDEF.

We're setting the interface twice in this case and apparently the later setting overwrites the first one in a wrong manner.

Ahhh… now I see what you mean… okay I revert this PR to its original state.

Done

Please squash

Squashed

Technically there was no ACK, but should be fine anyway.

Sorry 😞