Merge branch 'john/tools/fix-cover-register-race/ERL-943/OTP-15813' into maint

* john/tools/fix-cover-register-race/ERL-943/OTP-15813:
  cover: Fix register/2 race on startup

2 files changed, 53 insertions, 14 deletions

diff --git a/lib/tools/src/cover.erl b/lib/tools/src/cover.erl
index 8fe866cb69..2b3af417b6 100644
--- a/lib/tools/src/cover.erl
+++ b/lib/tools/src/cover.erl
@@ -196,6 +196,8 @@ start() ->
 		receive 
 		    {?SERVER,started} -> 
 			{ok,Pid};
+		    {?SERVER,{error,Error}} -> 
+			{error,Error};
 		    {'DOWN', Ref, _Type, _Object, Info} -> 
 			{error,Info}
 		end,
@@ -645,19 +647,31 @@ remote_reply(MainNode,Reply) ->
 %%%----------------------------------------------------------------------
 
 init_main(Starter) ->
-    register(?SERVER,self()),
-    ?COVER_MAPPING_TABLE = ets:new(?COVER_MAPPING_TABLE,
-                                   [ordered_set, public, named_table]),
-    ?COVER_CLAUSE_TABLE = ets:new(?COVER_CLAUSE_TABLE, [set, public,
-                                                        named_table]),
-    ?BINARY_TABLE = ets:new(?BINARY_TABLE, [set, public, named_table]),
-    ?COLLECTION_TABLE = ets:new(?COLLECTION_TABLE, [set, public,
-                                                    named_table]),
-    ?COLLECTION_CLAUSE_TABLE = ets:new(?COLLECTION_CLAUSE_TABLE, [set, public,
-                                                                  named_table]),
-    ok = net_kernel:monitor_nodes(true),
-    Starter ! {?SERVER,started},
-    main_process_loop(#main_state{}).
+    try register(?SERVER,self()) of
+        true ->
+            ?COVER_MAPPING_TABLE = ets:new(?COVER_MAPPING_TABLE,
+                                           [ordered_set, public, named_table]),
+            ?COVER_CLAUSE_TABLE = ets:new(?COVER_CLAUSE_TABLE, [set, public,
+                                                                named_table]),
+            ?BINARY_TABLE = ets:new(?BINARY_TABLE, [set, public, named_table]),
+            ?COLLECTION_TABLE = ets:new(?COLLECTION_TABLE, [set, public,
+                                                            named_table]),
+            ?COLLECTION_CLAUSE_TABLE = ets:new(?COLLECTION_CLAUSE_TABLE,
+                                               [set, public, named_table]),
+            ok = net_kernel:monitor_nodes(true),
+            Starter ! {?SERVER,started},
+            main_process_loop(#main_state{})
+    catch
+        error:badarg ->
+            %% The server's already registered; either report that it's already
+            %% started or try again if it died before we could find its pid.
+            case whereis(?SERVER) of
+                undefined ->
+                    init_main(Starter);
+                Pid ->
+                    Starter ! {?SERVER, {error, {already_started, Pid}}}
+            end
+    end.
 
 main_process_loop(State) ->
     receive
diff --git a/lib/tools/test/cover_SUITE.erl b/lib/tools/test/cover_SUITE.erl
index ee58fd7a10..462767f430 100644
--- a/lib/tools/test/cover_SUITE.erl
+++ b/lib/tools/test/cover_SUITE.erl
@@ -37,7 +37,7 @@ all() ->
                  dont_reconnect_after_stop, stop_node_after_disconnect,
                  export_import, otp_5031, otp_6115,
                  otp_8270, otp_10979_hanging_node, otp_14817,
-                 local_only],
+                 local_only, startup_race],
     case whereis(cover_server) of
         undefined ->
             [coverage,StartStop ++ NoStartStop];
@@ -1775,7 +1775,32 @@ local_only(Config) ->
     {ok,Name} = test_server:start_node(?FUNCTION_NAME, slave, []),
     {error,local_only} = cover:start([Name]),
     test_server:stop_node(Name),
+    ok.
 
+%% ERL-943; We should not crash on startup when multiple servers race to
+%% register the server name.
+startup_race(Config) when is_list(Config) ->
+    PidRefs = [spawn_monitor(fun() ->
+                                     case cover:start() of
+                                         {error, {already_started, _Pid}} ->
+                                             ok;
+                                         {ok, _Pid} ->
+                                             ok
+                                     end
+                             end) || _<- lists:seq(1,8)],
+    startup_race_1(PidRefs).
+
+startup_race_1([{Pid, Ref} | PidRefs]) ->
+    receive
+        {'DOWN', Ref, process, Pid, normal} ->
+            startup_race_1(PidRefs);
+        {'DOWN', Ref, process, Pid, _Other} ->
+            ct:fail("Cover server crashed on startup.")
+    after 5000 ->
+            ct:fail("Timed out.")
+    end;
+startup_race_1([]) ->
+    cover:stop(),
     ok.
 
 %%--Auxiliary------------------------------------------------------------