[erlang-questions] question on 'EXIT' messages delivered to exit-trapping gen_servers
Adam Kocoloski
into some behavior that I don't understand. We have a standalone
gen_server (let's call it A) that traps exits and was started by
another gen_server (B) using start_link. B does not trap exits, and
is linked to some other process C. Now we have the following sequence
of events:
* exit(C, kill)
* B exits immediately with Reason killed
* A calls terminate with Reason killed
That last line surprised me. I was expecting to receive {'EXIT', B,
killed} in A's mailbox, and in fact I had written a handle_info to
process all 'EXIT' messages, but none arrived. I checked that
trap_exit is still true in A's terminate. {trap_exit, true} clearly
has some effect, as without it terminate isn't even called.
I tried changing the exit signal to some other atom, but I still get
the same basic results (just with goodbye_cruel_world instead of kill/
killed in the Reasons). I'm stumped. Regards, Adam
_______________________________________________
erlang-questions mailing list
erlang-q...@erlang.org
http://www.erlang.org/mailman/listinfo/erlang-questions
Chandru
It looks like if a gen_server process was spawned using proc_lib:spawn and an EXIT signal is received indicating that the Parent process has died, the terminate() function is called!
The documentation should probably clarify this.
Here is the extract from gen_server.erl in R11B-5
loop(Parent, Name, State, Mod, Time, Debug) ->
Msg = receive
Input ->
Input
after Time ->
timeout
end,
case Msg of
{system, From, Req} ->
sys:handle_system_msg(Req, From, Parent, ?MODULE, Debug,
[Name, State, Mod, Time]);
{'EXIT', Parent, Reason} ->
terminate(Reason, Name, Msg, Mod, State, Debug);
_Msg when Debug =:= [] ->
handle_msg(Msg, Parent, Name, State, Mod, Time);
_Msg ->
Debug1 = sys:handle_debug(Debug, {?MODULE, print_event},
Name, {in, Msg}),
handle_msg(Msg, Parent, Name, State, Mod, Time, Debug1)
end.
cheers
Chandru
Adam Kocoloski
Adam