[cover] Allow reconnection if node has been disconnected or down

OTP-10523

Earlier, if the connection to a remote cover node was lost, all cover
data was lost and the cover_server on the remote node would die. This
would cause problems if there were cover compiled modules that would
still be executed since they would attempt to write to the no longer
existing ets tables belonging to the cover_server.

This commit changes this behavior so that the cover_server on the
remote node will survive connection loss and continue collecting cover
data. If the connection is re-established then the main node will sync
with the remote node again and cover data will not be lost (unless the
node was down).

1 files changed, 107 insertions, 9 deletions

diff --git a/lib/tools/test/cover_SUITE.erl b/lib/tools/test/cover_SUITE.erl
index 950c4ddbcb..3bf1b44af8 100644
--- a/lib/tools/test/cover_SUITE.erl
+++ b/lib/tools/test/cover_SUITE.erl
@@ -23,7 +23,7 @@
 	 init_per_group/2,end_per_group/2]).
 
 -export([start/1, compile/1, analyse/1, misc/1, stop/1, 
-	 distribution/1, export_import/1,
+	 distribution/1, reconnect/1, die_and_reconnect/1, export_import/1,
 	 otp_5031/1, eif/1, otp_5305/1, otp_5418/1, otp_6115/1, otp_7095/1,
          otp_8188/1, otp_8270/1, otp_8273/1, otp_8340/1]).
 
@@ -45,7 +45,8 @@ suite() -> [{ct_hooks,[ts_install_cth]}].
 all() -> 
     case whereis(cover_server) of
 	undefined ->
-	    [start, compile, analyse, misc, stop, distribution,
+	    [start, compile, analyse, misc, stop,
+	     distribution, reconnect, die_and_reconnect,
 	     export_import, otp_5031, eif, otp_5305, otp_5418,
 	     otp_6115, otp_7095, otp_8188, otp_8270, otp_8273,
 	     otp_8340];
@@ -408,20 +409,117 @@ distribution(Config) when is_list(Config) ->
     ?line true = is_unloaded(LocalBeam),
     ?line true = is_unloaded(N2Beam),
 
-    %% Check that cover_server on remote node dies if main node dies
+    %% Check that cover_server on remote node does not die if main node dies
     ?line {ok,[N1]} = cover:start(N1),
-    ?line true = is_pid(rpc:call(N1,erlang,whereis,[cover_server])),
+    ?line true = is_pid(N1Server = rpc:call(N1,erlang,whereis,[cover_server])),
     ?line exit(whereis(cover_server),kill),
-    ?line timer:sleep(10),
-    ?line undefined = rpc:call(N1,erlang,whereis,[cover_server]),    
-    
+    ?line timer:sleep(100),
+    ?line N1Server = rpc:call(N1,erlang,whereis,[cover_server]),
+
     %% Cleanup
     ?line Files = lsfiles(),
     ?line remove(files(Files, ".beam")),
     ?line ?t:stop_node(N1),
     ?line ?t:stop_node(N2).
 
-    
+%% Test that a lost node is reconnected
+reconnect(Config) ->
+    DataDir = ?config(data_dir, Config),
+    ok = file:set_cwd(DataDir),
+
+    {ok,a} = compile:file(a),
+    {ok,b} = compile:file(b),
+    {ok,f} = compile:file(f),
+
+    {ok,N1} = ?t:start_node(cover_SUITE_reconnect,peer,
+			    [{args," -pa " ++ DataDir},{start_cover,false}]),
+    {ok,a} = cover:compile(a),
+    {ok,f} = cover:compile(f),
+    {ok,[N1]} = cover:start(nodes()),
+
+    %% Some calls to check later
+    rpc:call(N1,f,f1,[]),
+    cover:flush(N1),
+    rpc:call(N1,f,f1,[]),
+
+    %% This will cause a call to f:f2() when nodes()==[] on N1
+    rpc:cast(N1,f,call_f2_when_isolated,[]),
+
+    %% Disconnect and check that node is removed from main cover node
+    net_kernel:disconnect(N1),
+    [] = cover:which_nodes(),
+    timer:sleep(500), % allow some time for the f:f2() call
+
+    %% Do some add one module (b) and remove one module (a)
+    code:purge(a),
+    {module,a} = code:load_file(a),
+    {ok,b} = cover:compile(b),
+    cover_compiled = code:which(b),
+
+    [] = cover:which_nodes(),
+    check_f_calls(1,0), % only the first call - before the flush
+
+    %% Reconnect the node and check that b and f are cover compiled but not a
+    net_kernel:connect_node(N1),
+    timer:sleep(100),
+    [N1] = cover:which_nodes(), % we are reconnected
+    cover_compiled = rpc:call(N1,code,which,[b]),
+    cover_compiled = rpc:call(N1,code,which,[f]),
+    ABeam = rpc:call(N1,code,which,[a]),
+    false = (cover_compiled==ABeam),
+
+    %% Ensure that we have:
+    %% * one f1 call from before the flush,
+    %% * one f1 call from after the flush but before disconnect
+    %% * one f2 call when disconnected
+    check_f_calls(2,1),
+
+    cover:stop(),
+    ?t:stop_node(N1),
+    ok.
+
+%% Test that a lost node is reconnected - also if it has been dead
+die_and_reconnect(Config) ->
+    DataDir = ?config(data_dir, Config),
+    ok = file:set_cwd(DataDir),
+
+    {ok,f} = compile:file(f),
+
+    NodeName = cover_SUITE_die_and_reconnect,
+    {ok,N1} = ?t:start_node(NodeName,peer,
+			    [{args," -pa " ++ DataDir},{start_cover,false}]),
+    %% {ok,a} = cover:compile(a),
+    {ok,f} = cover:compile(f),
+    {ok,[N1]} = cover:start(nodes()),
+
+    %% Some calls to check later
+    rpc:call(N1,f,f1,[]),
+    cover:flush(N1),
+    rpc:call(N1,f,f1,[]),
+
+    %% Kill the node
+    rpc:call(N1,erlang,halt,[]),
+    [] = cover:which_nodes(),
+
+    check_f_calls(1,0), % only the first call - before the flush
+
+    %% Restart the node and check that cover reconnects
+    {ok,N1} = ?t:start_node(NodeName,peer,
+			    [{args," -pa " ++ DataDir},{start_cover,false}]),
+    timer:sleep(100),
+    [N1] = cover:which_nodes(), % we are reconnected
+    cover_compiled = rpc:call(N1,code,which,[f]),
+
+    %% One more call...
+    rpc:call(N1,f,f1,[]),
+
+    %% Ensure that no more calls are counted
+    check_f_calls(2,0),
+
+    cover:stop(),
+    ?t:stop_node(N1),
+    ok.
+
 export_import(suite) -> [];
 export_import(Config) when is_list(Config) ->
     ?line DataDir = ?config(data_dir, Config),
@@ -1253,4 +1351,4 @@ is_unloaded(What) ->
     end.
 
 check_f_calls(F1,F2) ->
-    {ok,[{{f,f1,0},F1},{{f,f2,0},F2}]} = cover:analyse(f,calls,function).
+    {ok,[{{f,f1,0},F1},{{f,f2,0},F2}|_]} = cover:analyse(f,calls,function).