Log tls handshake timeouts about rabbitmq-server

[lukebakken]

@ansd you might find this interesting. When the TLS handshake times out, this is what rabbit_mqtt_reader:init/1 returns:

1> redbug:start("rabbit_mqtt_reader:init->return,stack",[{time,60000}]).
{529,1}

% 10:21:43 <0.791.0>({rabbit_mqtt_reader,init,1})
% rabbit_mqtt_reader:init({acceptor,{0,0,0,0,0,0,0,0},8883})
%   proc_lib:init_p_do_apply/3
                                                                                                                                                                                    % 10:21:43 <0.791.0>(dead)
% rabbit_mqtt_reader:init/1 -> {exit,normal}

I'm assuming this is why nothing is logged ... I'm continuing my investigation. I'm kind of surprised a badmatch is getting turned into {exit, normal} somehow 🤔

UPDATE: oh, here's why:

(rabbit@PROKOFIEV)4> redbug:start("rabbit_networking:handshake->return,stack",[{time,60000}]).
{529,1}

% 10:29:05 <0.774.0>({rabbit_mqtt_reader,init,1})
% rabbit_networking:handshake({acceptor,{0,0,0,0,0,0,0,0},8883}, false)
%   rabbit_mqtt_reader:init/1
%   proc_lib:init_p_do_apply/3

% 10:29:05 <0.774.0>(dead)
% rabbit_networking:handshake/2 -> {exit,normal}

UPDATE 2: yeah, it seems like the whole supervisor setup for a TCP/SSL ranch listener swallows errors like TLS handshake timeout, with nary a log message or crash anywhere.