Hi Ricardo, On Thu, Mar 28, 2019 at 10:15:41PM -0300, Ricardo Nabinger Sanchez wrote: > Hello, > > We have been chasing a segfault for a few weeks and today we were able > to track it down. There is a null-pointer dereferencing when using > tcp-check connect; although we don't know yet as to why the protocol > family might be unknown during tcpcheck_main().
Many thanks for this fix! I think I'm seeing how this can happen, it's probably since the introduction of the DNS to set addresses on servers. In the past a server always had an address, but now a server may also not have an address at all. It's unclear to me exactly how a check could be scheduled on an addressless server but I think it could happen if the check is already scheduled and the server loses its address just before the check actually runs. I'm taking it just before issuing 1.9.6, thanks! Willy
