CMake is building 32-bit build even when executed from "Visual Studio 2015 x64 native Tools" command prompt. I'd to do following to get 64-bit build. It might be nice to include this in documentation.

1. Go to "Visual Studio 2015 x64 native Tools" command prompt. Make sure you have VS2015 Update 3.
2. cmake -G"Visual Studio 14 2015 Win64" ..
3. cmake --build .

This will generate Debug build. To get the release build do cmake --build . --config Release.

I've only recently learned about this library and initial tests look very promising. However, given that the last commit seems to be over a year ago I was wondering if this project is still alive.

I'm running into a build error due to the format.cc dependency found here:

dependencies/src/format.cc:478:20: error: expected unqualified-id before numeric constant const unsigned CHAR_WIDTH = 1;

I think the bug has been addressed in a new version of this dependency.

Oh, I should mention this is on GCC 6.3.1.

Currently, I'm trying to add rpclib to hunter package manager: Add rpclib to hunter-packages

This pull request looks not okay for me. Is namespace renaming really needed? Why we can't just write this code inside rpclib's header file (we DON'T need to modify dependencies)?:

namespace clmdep_asio
{
	using namespace asio;
};

rpclib will be forked to hunter-packages, because hunter somehow must manage dependencies by itself (no need in "dependency" folder)

I see that making a blocking call to a function when the server is not running will block forever.

rpc::client rpc_client{port, addr};
rpc_client.get_connection_state(); // returns "initial"
rpc_client.call("echo", "arguments"); // hangs if the server isn't running

This is because the future.wait() of a blocking call will wait forever. Would you accept a PR adding the ability to set a timeout? If so, what would your preferred response to an error be? Throwing an rpc_error with the what indicating timeout?

• Replaces chrono_literals with std::chrono::milliseconds()
• Define make_unique if not available
• Replace index_sequence with manual indexing of tuple elements (limits to 5 args per call on C++11)
• Replace decltype(auto) function types with C++11-compatible notation
• Replace named captures in lambdas with regular captures
• Replaced decay_t with decay<>::type

Hi, the main goal of this PR is to solve issue #175. Here is the description of what each commit does:

a0e6123 : Add a unit test reproducting issue #175 b18569a : fix thread safety in session management 8a98779 : Prevent the server_session to be destroyed while its closing code is still queued, causing a segfault b134678 : Fixed socket double close. If (exit_) is true, then the session is already closed and the socket_.close() is queued

We use your library actively, please give me a feedback on this PR so it could be accepted in the official repository. It would be beneficial for everyone to avoid another fork.

Hello rpclib team, How can I build rpclib for Android ? I am following this: https://kvurd.com/blog/compiling-a-cpp-library-for-android-with-android-studio/ But, where should I make changes in rpclib CMakeLists.txt ?

Thanks

@sztomi

a simple rpclib server will eventially crash after ~1000 consecutive connections. Looks like the server doesn't properly release the connection to a client after client goes out of scope. How would I properly close the client?

test_leak crashes, test_no_leak doesn't crash.

void test_leak(){
  for(int counter=0; ; counter++){
    rpc::client client(host, port);
    auto response = client.call("fun", "arg").as<string>();
  }
}

void test_no_leak(){
   rpc::client client(host, port);
   for(int counter=0; ; counter++){
     auto response = client.call("fun", "arg").as<string>();
  }
}

• ls -l /proc/$pid/fd | wc -l keeps increasing
• htop shows memory keeps increasing
• lsof -p $pid keeps increasing, showing entries such as:

binary $pid user  ...     IPv4 ...      0t0      TCP localhost:18700->localhost:39692 (ESTABLISHED)

EDIT: for the server, use this for eg:bind("fun", []() { return ""; });

Here's where it crashes:

https://github.com/rpclib/rpclib/blob/master/lib/rpc/detail/server_session.cc#L98

When server_session tries to execute write_strand_.post with the lambda it just segfaults. Here are the last three logs before the segfault happens:

2017-11-17 06:29:38.176  DEBUG   dispatcher      Dispatching call to 'Recognize' (/tmp/rpclib-2.2.0/lib/rpc/dispatcher.cc:44)
2017-11-17 06:29:38.219  INFO    session         Client disconnected
2017-11-17 06:29:38.219  INFO    session         Closing session.

I'm sorry I don't have an example which could be used to reproduce the bug.

When a very large string (80k bytes) to the echo example, the client only displays the first 4102.

To reproduce:

compile the echo_client and echo_server example and input a ridiculously large string. Only a subset of the string will be output.

Hi,

How to use without the server?

I want transfer the data to rpclib thought a string, like you receive from server, but i will put it manually.

I need communicate from my mobile app without the server, so i will send from mobile a string, receive on cxx side, send to rpclib and rpclib will work like it receive from http server.

How can i do this?

Thanks.

Hello,

thanks for the great work. I ran both the client and the server examples successfully.

however, I need many more features for my server-side so I can't use c++ on my server. I need c# .net core. on the other hand, I couldn't find any compatible library for .net that supports msgpack RPC

my client is c++ and my server is C# .net core

any recommendation? #272 #263

In my case I need the address/port of the remote endpoint (client) for connection tracking purposes.

I noticed the underlying server::impl logs the address obtained from the socket_.remote_endpoint().

Is there any way of accessing this information inside a functor?

Thanks in advance. LL