If a client call times out, the server raises a std::system_error when trying to write the answer.

I'm not sure whether this should be silently caught by the server or if an interface to handle this type of error should be created.

A temporary workaround is to increase the client timeout, which doesn't resolve the problem if the client quits. When the run function of the server is used, the exception can be caught and the server restarted, which is not the case with the asyn_run function.

Note: I tried using rpc::server::suppress_exceptions but this only catches exceptions from within the bound callbacks.

The following is a minimal example.

Compilation and run

Running on Linux (Mint, 18.2).

server

$ clang++ server.cpp -o server -Iinclude -Llib -std=c++14 -lrpc -lpthread
$ ./server
terminate called after throwing an instance of 'std::system_error'
  what():  shutdown: Bad file descriptor
Abandon

client

$ clang++ client.cpp -o client -Iinclude -Llib -std=c++14 -lrpc -lpthread
$ ./client 
rpc::timeout: Timeout of 50ms while calling RPC function 'bug'

Sources
server.cpp

#include <thread>
#include <chrono>

#include "rpc/server.h"

int main() {

  rpc::server server(9999);

  server.bind("bug", []{
      using namespace std::chrono_literals;
      std::this_thread::sleep_for(2s);
    });

  server.run();
}

client.cpp

#include <iostream>

#include "rpc/client.h"
#include "rpc/rpc_error.h"

int main() {

  rpc::client client("localhost", 9999);
  client.set_timeout(50);

  try {
    client.call("bug");
  } catch(rpc::timeout& e) {
    std::cerr << e.what() << '\n';
  }
}