inside gen_server call

动机
前段时间的两个工作.一个是entity集群库, 可以通过entity_id调用任意节点上的entity. 一个是名字服务, 可以为一系列pid注册名字, 并可以以这些名字调用对应的pid.都会遇到同一些问题: 当我们使用GenServer.call/2时, 发生了什么, 会有什么异常情况发生? 哪些异常应该捕获? 以什么样的方式处理这些异常/错误?当call的pid所在的node崩溃时, 会有什么异常? 在调用开始前/中崩溃, 有什么不同.当call的pid所在的node网络突然中断呢? 会有什么表现?当call的pid崩溃时呢?是否应该捕获timeout?这些问题在文档中并没有答案. 所以, 探索一下.
深挖实现
源码版本
erlang: OTP-21.0.9
inside gen_server.erl call
gen_server.erl:203
%% —————————————————————–
%% Make a call to a generic server.
%% If the server is located at another node, that node will
%% be monitored.
%% If the client is trapping exits and is linked server termination
%% is handled here (? Shall we do that here (or rely on timeouts) ?).
%% —————————————————————–
call(Name, Request) ->
case catch gen:call(Name, ‘$gen_call’, Request) of
{ok,Res} ->
Res;
{‘EXIT’,Reason} ->
exit({Reason, {?MODULE, call, [Name, Request]}})
end.

call(Name, Request, Timeout) ->
case catch gen:call(Name, ‘$gen_call’, Request, Timeout) of
{ok,Res} ->
Res;
{‘EXIT’,Reason} ->
exit({Reason, {?MODULE, call, [Name, Request, Timeout]}})
end.
gen.erl:160
do_call(Process, Label, Request, Timeout) when is_atom(Process) =:= false ->
Mref = erlang:monitor(process, Process),

%% OTP-21:
%% Auto-connect is asynchronous. But we still use ‘noconnect’ to make sure
%% we send on the monitored connection, and not trigger a new auto-connect.
%%
erlang:send(Process, {Label, {self(), Mref}, Request}, [noconnect]),

receive
{Mref, Reply} ->
erlang:demonitor(Mref, [flush]),
{ok, Reply};
{‘DOWN’, Mref, _, _, noconnection} ->
Node = get_node(Process),
exit({nodedown, Node});
{‘DOWN’, Mref, _, _, Reason} ->
exit(Reason)
after Timeout ->
erlang:demonitor(Mref, [flush]),
exit(timeout)
end.
可以看到, call一个process的过程:

monitor process
send_msg to process and receive for reply.

可能的情况有

正常返回, demonitor
noconnection
pid down for any reason
timeout

那么, 前面的各种异常, 会对应到哪些情况呢? 有没有意外?先看看monitor一个process时到底做了什么.
inside monitor
erlang.erl:1291

-type registered_name() :: atom().
-type registered_process_identifier() :: registered_name() | {registered_name(), node()}.
-type monitor_process_identifier() :: pid() | registered_process_identifier().
-type monitor_port_identifier() :: port() | registered_name().

%% monitor/2
-spec monitor
(process, monitor_process_identifier()) -> MonitorRef
when MonitorRef :: reference();
(port, monitor_port_identifier()) -> MonitorRef
when MonitorRef :: reference();
(time_offset, clock_service) -> MonitorRef
when MonitorRef :: reference().

monitor(_Type, _Item) ->
erlang:nif_error(undefined).
在monitor process时, 可以是一个pid, name, name node tuple. 但这里没有具体实现, 找一下nif.
inside receive