TCAP ISM: trap the EXIT signal to avoid getting killed too early

When we receive a TC-END with RESULT-L, the process hierarchy might
kill us before we received the RESSULT-L, or aftger we received it,
but before we had a chance to deliver it to the TC User.

Trapping the EXIT flag solves the problem.  It also means that the EXIT
signal will not propagate to other linked processes.  However, there
should be no other processes, as ISM is not starting any.
Harald Welte 10 years ago
parent cb55abc0f7
commit 3a58450d11
  1. 9

@ -89,6 +89,9 @@ start_link(Usap, DialogueId, InvokeId, OpClass, InvTimeout) ->
% DHA needs to tell us: USAP, DialogueID, InvokeID, CCO-PID, OpClass, InvTimer
init([Usap, DialogueId, InvokeId, CcoPid, OpClass, InvTimeout]) ->
% we need to trap EXIT signals, as otherwise we might be terminated by
% the dialogue END before we have delivered the components to the TC-User.
process_flag(trap_exit, true),
State = #state{usap = Usap, dialogueId = DialogueId,
invokeId = InvokeId, cco = CcoPid,
op_class = OpClass, inv_timeout = InvTimeout},
@ -230,7 +233,11 @@ handle_info(Info, StateName, State) ->
{next_state, StateName, State}.
%% handle a shutdown request
terminate(_Reason, _StateName, _State) -> ok.
terminate(Reason, StateName, State) ->
error_logger:format("~w (~w) terminating Reason: ~w in state ~w~n",
[?MODULE, self(), Reason, StateName]).
%% handle updating state data due to a code replacement
code_change(_OldVsn, StateName, State, _Extra) ->