Merge branch 'anders/diameter/connect_timer/OTP-12281' into maint

* anders/diameter/connect_timer/OTP-12281:
  Tweak reason in closed event
  Fix ignored connect timer
  Check {connect,watchdog}_timer distinction in event testcases
  Rename reconnect_timer to connect_timer in examples and suites

6 files changed, 47 insertions, 36 deletions

diff --git a/lib/diameter/examples/code/peer.erl b/lib/diameter/examples/code/peer.erl
index b4ee17e4b7..7519abfb2c 100644
--- a/lib/diameter/examples/code/peer.erl
+++ b/lib/diameter/examples/code/peer.erl
@@ -74,7 +74,7 @@ start(Name, Opts)
     | {error, term()}.
 
 connect(Name, T) ->
-    diameter:add_transport(Name, {connect, [{reconnect_timer, 5000}
+    diameter:add_transport(Name, {connect, [{connect_timer, 5000}
                                             | client(T)]}).
 
 %% listen/2
diff --git a/lib/diameter/src/base/diameter_peer_fsm.erl b/lib/diameter/src/base/diameter_peer_fsm.erl
index 86fc43cdc5..ee6e7dd89e 100644
--- a/lib/diameter/src/base/diameter_peer_fsm.erl
+++ b/lib/diameter/src/base/diameter_peer_fsm.erl
@@ -225,8 +225,8 @@ start_transport(Addrs0, T) ->
             erlang:monitor(process, TPid),
             q_next(TPid, Addrs0, Tmo, Data),
             {TPid, Addrs};
-        No ->
-            exit({shutdown, No})
+        {error, No} ->
+            exit({shutdown, {no_connection, No}})
     end.
 
 svc(#diameter_service{capabilities = LCaps0} = Svc, Addrs) ->
@@ -368,11 +368,8 @@ transition({diameter, {TPid, connected}},
 %% message. This may be followed by an incoming message which arrived
 %% before the transport was killed and this can't be distinguished
 %% from one from the transport that's been started to replace it.
-transition({diameter, {_, connected}}, _) ->
-    {stop, connection_timeout};
-transition({diameter, {_, connected, _}}, _) ->
-    {stop, connection_timeout};
-transition({diameter, {_, connected, _, _}}, _) ->
+transition({diameter, T}, _)
+  when tuple_size(T) < 5, connected == element(2,T) ->
     {stop, connection_timeout};
 
 %% Connection has timed out: start an alternate.
diff --git a/lib/diameter/src/base/diameter_watchdog.erl b/lib/diameter/src/base/diameter_watchdog.erl
index eff5096745..b7f2d24941 100644
--- a/lib/diameter/src/base/diameter_watchdog.erl
+++ b/lib/diameter/src/base/diameter_watchdog.erl
@@ -255,11 +255,15 @@ close({'DOWN', _, process, TPid, {shutdown, Reason}},
 close(_, _) ->
     ok.
 
-event(_, #watchdog{status = T}, #watchdog{status = T}) ->
-    ok;
-
-event(_, #watchdog{transport = undefined}, #watchdog{transport = undefined}) ->
+event(_,
+      #watchdog{status = From, transport = F},
+      #watchdog{status = To, transport = T})
+  when F == undefined, T == undefined;  %% transport not started
+       From == initial, To == down;     %% never really left INITIAL
+       From == To ->                    %% no state transition
     ok;
+%% Note that there is no INITIAL -> DOWN transition in RFC 3539: ours
+%% is just a consequence of stop.
 
 event(Msg,
       #watchdog{status = From, transport = F, parent = Pid},
@@ -411,7 +415,7 @@ transition({'DOWN', _, process, TPid, _Reason},
     stop;
 
 %% ... or not.
-transition({'DOWN', _, process, TPid, _Reason},
+transition({'DOWN', _, process, TPid, _Reason} = D,
            #watchdog{transport = TPid,
                      status = T,
                      restrict = {_,R}}
@@ -422,20 +426,14 @@ transition({'DOWN', _, process, TPid, _Reason},
 
     %% Close an accepting watchdog immediately if there's no
     %% restriction on the number of connections to the same peer: the
-    %% state machine never enters state REOPEN in this case. The
-    %% 'close' message (instead of stop) is so as not to bypass the
-    %% sending of messages to the service process in handle_info/2.
-
-    if T /= initial, M == accept, not R ->
-            send(self(), close),
-            S#watchdog{status = down};
-       T /= initial ->
-            set_watchdog(S#watchdog{status = down});
-       M == connect ->
-            set_watchdog(S);
-       M == accept ->
-            send(self(), close),
-            S
+    %% state machine never enters state REOPEN in this case.
+
+    if T == initial;
+       M == accept, not R ->
+            close(D, S0),
+            stop;
+       true ->
+            set_watchdog(S#watchdog{status = down})
     end;
 
 %% Incoming message.
diff --git a/lib/diameter/test/diameter_config_SUITE.erl b/lib/diameter/test/diameter_config_SUITE.erl
index d10ee83ba4..ad5b3f9420 100644
--- a/lib/diameter/test/diameter_config_SUITE.erl
+++ b/lib/diameter/test/diameter_config_SUITE.erl
@@ -157,7 +157,7 @@
          {length_errors,
           [[exit], [handle], [discard]],
           [[x]]},
-         {reconnect_timer,
+         {connect_timer,
           [[3000]],
           [[infinity]]},
          {watchdog_timer,
diff --git a/lib/diameter/test/diameter_event_SUITE.erl b/lib/diameter/test/diameter_event_SUITE.erl
index 94b4967921..f43f111d20 100644
--- a/lib/diameter/test/diameter_event_SUITE.erl
+++ b/lib/diameter/test/diameter_event_SUITE.erl
@@ -107,29 +107,38 @@ start_server(Config) ->
 %% Connect with matching capabilities and expect the connection to
 %% come up.
 up(Config) ->
-    {Svc, Ref} = connect(Config, []),
+    {Svc, Ref} = connect(Config, [{connect_timer, 5000},
+                                  {watchdog_timer, 15000}]),
     start = event(Svc),
-    {up, Ref, {_,_Caps}, _Config, #diameter_packet{}} = event(Svc),
-    {watchdog, Ref, _, {initial, okay}, _} = event(Svc).
+    {up, Ref, {TPid, Caps}, Cfg, #diameter_packet{}} = event(Svc),
+    {watchdog, Ref, _, {initial, okay}, _} = event(Svc),
+    %% Kill the transport process and see that the connection is
+    %% reestablished after a watchdog timeout, not after connect_timer
+    %% expiry.
+    exit(TPid, kill),
+    {down, Ref, {TPid, Caps}, Cfg} = event(Svc),
+    {watchdog, Ref, _, {okay, down}, _} = event(Svc),
+    {reconnect, Ref, _} = event(Svc, 10000, 20000).
 
 %% Connect with non-matching capabilities and expect CEA from the peer
 %% to indicate as much and then for the transport to be restarted
-%% (after reconnect_timer).
+%% (after connect_timer).
 down(Config) ->
     {Svc, Ref} = connect(Config, [{capabilities, [{'Acct-Application-Id',
                                                    [?DICT_ACCT:id()]}]},
                                   {applications, [?DICT_ACCT]},
-                                  {reconnect_timer, 5000}]),
+                                  {connect_timer, 5000},
+                                  {watchdog_timer, 20000}]),
     start = event(Svc),
     {closed, Ref, {'CEA', ?NO_COMMON_APP, _, #diameter_packet{}}, _}
         = event(Svc),
-    {reconnect, Ref, _} = event(Svc).
+    {reconnect, Ref, _} = event(Svc, 4000, 10000).
 
 %% Connect with matching capabilities but have the server delay its
 %% CEA and cause the client to timeout.
 cea_timeout(Config) ->
     {Svc, Ref} = connect(Config, [{capx_timeout, ?SERVER_CAPX_TMO div 2},
-                                  {reconnect_timer, 2*?SERVER_CAPX_TMO}]),
+                                  {connect_timer, 2*?SERVER_CAPX_TMO}]),
     start = event(Svc),
     {closed, Ref, {'CEA', timeout}, _} = event(Svc).
 
@@ -165,6 +174,13 @@ uniq() ->
 event(Name) ->
     receive #diameter_event{service = Name, info = T} -> T end.
 
+event(Name, TL, TH) ->
+    T0 = now(),
+    Event = event(Name),
+    DT = timer:now_diff(now(), T0) div 1000,
+    {true, true, DT, Event} = {TL < DT, DT < TH, DT, Event},
+    Event.
+
 start_service(Name, Opts) ->
     diameter:start_service(Name, [{monitor, self()} | Opts]).
 
diff --git a/lib/diameter/test/diameter_transport_SUITE.erl b/lib/diameter/test/diameter_transport_SUITE.erl
index 9408fae62c..fcffa69c24 100644
--- a/lib/diameter/test/diameter_transport_SUITE.erl
+++ b/lib/diameter/test/diameter_transport_SUITE.erl
@@ -194,7 +194,7 @@ reconnect({connect, Ref}) ->
     true = diameter:subscribe(SvcName),
     ok = start_service(SvcName),
     [{{_, _, LRef}, Pid}] = diameter_reg:wait({?MODULE, Ref, '_'}),
-    CRef = ?util:connect(SvcName, tcp, LRef, [{reconnect_timer, 2000},
+    CRef = ?util:connect(SvcName, tcp, LRef, [{connect_timer, 2000},
                                               {watchdog_timer, 6000}]),
 
     %% Tell partner to kill transport after seeing that there are no