diff options
Diffstat (limited to 'lib/tools/src')
| -rw-r--r-- | lib/tools/src/cover.erl | 188 |
1 files changed, 137 insertions, 51 deletions
diff --git a/lib/tools/src/cover.erl b/lib/tools/src/cover.erl index e21bd1b88c..10f14b0a49 100644 --- a/lib/tools/src/cover.erl +++ b/lib/tools/src/cover.erl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 2001-2011. All Rights Reserved. +%% Copyright Ericsson AB 2001-2012. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -26,12 +26,17 @@ %% ARCHITECTURE %% The coverage tool consists of one process on each node involved in %% coverage analysis. The process is registered as 'cover_server' -%% (?SERVER). All cover_servers in the distributed system are linked -%% together. The cover_server on the 'main' node is in charge, and it -%% traps exits so it can detect nodedown or process crashes on the -%% remote nodes. This process is implemented by the functions -%% init_main/1 and main_process_loop/1. The cover_server on the remote -%% nodes are implemented by the functions init_remote/2 and +%% (?SERVER). The cover_server on the 'main' node is in charge, and +%% it monitors the cover_servers on all remote nodes. When it gets a +%% 'DOWN' message for another cover_server, it marks the node as +%% 'lost'. If a nodeup is received for a lost node the main node +%% ensures that the cover compiled modules are loaded again. If the +%% remote node was alive during the disconnected periode, cover data +%% for this periode will also be included in the analysis. +%% +%% The cover_server process on the main node is implemented by the +%% functions init_main/1 and main_process_loop/1. The cover_server on +%% the remote nodes are implemented by the functions init_remote/2 and %% remote_process_loop/1. %% %% TABLES @@ -81,15 +86,17 @@ export/1, export/2, import/1, modules/0, imported/0, imported_modules/0, which_nodes/0, is_compiled/1, reset/1, reset/0, + flush/1, stop/0, stop/1]). --export([remote_start/1]). +-export([remote_start/1,get_main_node/0]). %-export([bump/5]). -export([transform/4]). % for test purposes -record(main_state, {compiled=[], % [{Module,File}] imported=[], % [{Module,File,ImportFile}] stopper, % undefined | pid() - nodes=[]}). % [Node] + nodes=[], % [Node] + lost_nodes=[]}). % [Node] -record(remote_state, {compiled=[], % [{Module,File}] main_node}). % atom() @@ -497,6 +504,19 @@ stop(Node) when is_atom(Node) -> stop(Nodes) -> call({stop,remove_myself(Nodes,[])}). +%% flush(Nodes) -> ok | {error,not_main_node} +%% Nodes = [Node] | Node +%% Node = atom() +%% Error = {not_cover_compiled,Module} +flush(Node) when is_atom(Node) -> + flush([Node]); +flush(Nodes) -> + call({flush,remove_myself(Nodes,[])}). + +%% Used by test_server only. Not documented. +get_main_node() -> + call(get_main_node). + %% bump(Module, Function, Arity, Clause, Line) %% Module = Function = atom() %% Arity = Clause = Line = integer() @@ -541,7 +561,10 @@ remote_call(Node,Request) -> Return = receive {'DOWN', Ref, _Type, _Object, _Info} -> - {error,node_dead}; + case Request of + {remote,stop} -> ok; + _ -> {error,node_dead} + end; {?SERVER,Reply} -> Reply end, @@ -569,40 +592,14 @@ init_main(Starter) -> ets:new(?BINARY_TABLE, [set, named_table]), ets:new(?COLLECTION_TABLE, [set, public, named_table]), ets:new(?COLLECTION_CLAUSE_TABLE, [set, public, named_table]), - process_flag(trap_exit,true), + net_kernel:monitor_nodes(true), Starter ! {?SERVER,started}, main_process_loop(#main_state{}). main_process_loop(State) -> receive {From, {start_nodes,Nodes}} -> - ThisNode = node(), - StartedNodes = - lists:foldl( - fun(Node,Acc) -> - case rpc:call(Node,cover,remote_start,[ThisNode]) of - {ok,RPid} -> - link(RPid), - [Node|Acc]; - Error -> - io:format("Could not start cover on ~w: ~p\n", - [Node,Error]), - Acc - end - end, - [], - Nodes), - - %% In case some of the compiled modules have been unloaded they - %% should not be loaded on the new node. - {_LoadedModules,Compiled} = - get_compiled_still_loaded(State#main_state.nodes, - State#main_state.compiled), - remote_load_compiled(StartedNodes,Compiled), - - State1 = - State#main_state{nodes = State#main_state.nodes ++ StartedNodes, - compiled = Compiled}, + {StartedNodes,State1} = do_start_nodes(Nodes, State), reply(From, {ok,StartedNodes}), main_process_loop(State1); @@ -707,8 +704,13 @@ main_process_loop(State) -> {From, {stop,Nodes}} -> remote_collect('_',Nodes,true), reply(From, ok), - State1 = State#main_state{nodes=State#main_state.nodes--Nodes}, - main_process_loop(State1); + Nodes1 = State#main_state.nodes--Nodes, + main_process_loop(State#main_state{nodes=Nodes1}); + + {From, {flush,Nodes}} -> + remote_collect('_',Nodes,false), + reply(From, ok), + main_process_loop(State); {From, stop} -> lists:foreach( @@ -788,14 +790,30 @@ main_process_loop(State) -> end, main_process_loop(S); - {'EXIT',Pid,_Reason} -> - %% Exit is trapped on the main node only, so this will only happen - %% there. I assume that I'm only linked to cover_servers on remote - %% nodes, so this must be one of them crashing. - %% Remove node from list! - State1 = State#main_state{nodes=State#main_state.nodes--[node(Pid)]}, + {'DOWN', _MRef, process, {?SERVER,Node}, _Info} -> + %% A remote cover_server is down, mark as lost + Nodes = State#main_state.nodes--[Node], + Lost = [Node|State#main_state.lost_nodes], + main_process_loop(State#main_state{nodes=Nodes,lost_nodes=Lost}); + + {nodeup,Node} -> + State1 = + case lists:member(Node,State#main_state.lost_nodes) of + true -> + sync_compiled(Node,State); + false -> + State + end, main_process_loop(State1); + + {nodedown,_} -> + %% Will be taken care of when 'DOWN' message arrives + main_process_loop(State); + {From, get_main_node} -> + reply(From, node()), + main_process_loop(State); + get_status -> io:format("~p~n",[State]), main_process_loop(State) @@ -850,7 +868,16 @@ remote_process_loop(State) -> {remote,stop} -> reload_originals(State#remote_state.compiled), unregister(?SERVER), - remote_reply(State#remote_state.main_node, ok); + ok; % not replying since 'DOWN' message will be received anyway + + {remote,get_compiled} -> + remote_reply(State#remote_state.main_node, + State#remote_state.compiled), + remote_process_loop(State); + + {From, get_main_node} -> + remote_reply(From, State#remote_state.main_node), + remote_process_loop(State); get_status -> io:format("~p~n",[State]), @@ -961,6 +988,36 @@ unload([]) -> %%%--Handling of remote nodes-------------------------------------------- +do_start_nodes(Nodes, State) -> + ThisNode = node(), + StartedNodes = + lists:foldl( + fun(Node,Acc) -> + case rpc:call(Node,cover,remote_start,[ThisNode]) of + {ok,_RPid} -> + erlang:monitor(process,{?SERVER,Node}), + [Node|Acc]; + Error -> + io:format("Could not start cover on ~w: ~p\n", + [Node,Error]), + Acc + end + end, + [], + Nodes), + + %% In case some of the compiled modules have been unloaded they + %% should not be loaded on the new node. + {_LoadedModules,Compiled} = + get_compiled_still_loaded(State#main_state.nodes, + State#main_state.compiled), + remote_load_compiled(StartedNodes,Compiled), + + State1 = + State#main_state{nodes = State#main_state.nodes ++ StartedNodes, + compiled = Compiled}, + {StartedNodes, State1}. + %% start the cover_server on a remote node remote_start(MainNode) -> case whereis(?SERVER) of @@ -984,6 +1041,30 @@ remote_start(MainNode) -> {error,{already_started,Pid}} end. +%% If a lost node comes back, ensure that main and remote node has the +%% same cover compiled modules. Note that no action is taken if the +%% same {Mod,File} eksists on both, i.e. code change is not handled! +sync_compiled(Node,State) -> + #main_state{compiled=Compiled0,nodes=Nodes,lost_nodes=Lost}=State, + State1 = + case remote_call(Node,{remote,get_compiled}) of + {error,node_dead} -> + {_,S} = do_start_nodes([Node],State), + S; + {error,_} -> + State; + RemoteCompiled -> + {_,Compiled} = get_compiled_still_loaded(Nodes,Compiled0), + Unload = [UM || {UM,_}=U <- RemoteCompiled, + false == lists:member(U,Compiled)], + remote_unload([Node],Unload), + Load = [L || L <- Compiled, + false == lists:member(L,RemoteCompiled)], + remote_load_compiled([Node],Load), + State#main_state{compiled=Compiled, nodes=[Node|Nodes]} + end, + State1#main_state{lost_nodes=Lost--[Node]}. + %% Load a set of cover compiled modules on remote nodes, %% We do it ?MAX_MODS modules at a time so that we don't %% run out of memory on the cover_server node. @@ -1094,7 +1175,6 @@ remove_myself([Node|Nodes],Acc) -> remove_myself(Nodes,[Node|Acc]); remove_myself([],Acc) -> Acc. - %%%--Handling of modules state data-------------------------------------- @@ -2254,7 +2334,13 @@ do_reset2([]) -> do_clear(Module) -> ets:match_delete(?COVER_CLAUSE_TABLE, {Module,'_'}), ets:match_delete(?COVER_TABLE, {#bump{module=Module},'_'}), - ets:match_delete(?COLLECTION_TABLE, {#bump{module=Module},'_'}). + case lists:member(?COLLECTION_TABLE, ets:all()) of + true -> + %% We're on the main node + ets:match_delete(?COLLECTION_TABLE, {#bump{module=Module},'_'}); + false -> + ok + end. not_loaded(Module, unloaded, State) -> do_clear(Module), @@ -2307,7 +2393,7 @@ pmap(Fun, [E | Rest], Pids, Limit, Cnt, Acc) when Cnt < Limit -> pmap(Fun, Rest, Pids ++ [Pid], Limit, Cnt + 1, Acc); pmap(Fun, List, [Pid | Pids], Limit, Cnt, Acc) -> receive - {'DOWN', _Ref, process, _, _} -> + {'DOWN', _Ref, process, X, _} when is_pid(X) -> pmap(Fun, List, [Pid | Pids], Limit, Cnt - 1, Acc); {res, Pid, Res} -> pmap(Fun, List, Pids, Limit, Cnt, [Res | Acc]) @@ -2316,6 +2402,6 @@ pmap(_Fun, [], [], _Limit, 0, Acc) -> lists:reverse(Acc); pmap(Fun, [], [], Limit, Cnt, Acc) -> receive - {'DOWN', _Ref, process, _, _} -> + {'DOWN', _Ref, process, X, _} when is_pid(X) -> pmap(Fun, [], [], Limit, Cnt - 1, Acc) end. |