Re: [Boost-users] boost::asio::asyn_accept handler

On 29/03/2017 12:02, Christopher Pisz via Boost-users wrote:

Sigh. I know that code wouldn't work. That is my interpretation of what you gave me. Now you are talking about two arguments and placeholders, but why would I give it a place holder, when I am giving it the argument directly? the tcp_connection::pointer argument is new_connection. The placeholder is for the error code which I am not supplying directly. Either way, it still fails to compile if you add a std::placeholders::_2 on the end.

You need _1 (and only that) in the functor that is passed to async_accept, since that expects a callback that takes exactly one error_code argument. You might need other placeholders if elsewhere you're using functors that take more than one argument, as in one previous instance where you were using multiple binds or getting a callback passed in externally to then rebind locally.

Can we just get a compilable working example of the following method while maintaining the use of the std::function variable? I don't want to bind directly, I don't want to use lambda, I want to have an std::function variable that I can pass to others. It's really a simple 6 line problem....

Just be careful with passing functions outside of the class, as they can have hidden dependencies (eg. the below will UB or crash if something tries to call the callback after tcp_server has been destroyed).

The following code still fails to compile:

void start_accept() { tcp_connection::pointer new_connection = tcp_connection::create(acceptor_.get_io_service());

std::function callback = std::bind(&tcp_server::handle_accept, this, new_connection, std::placeholders::_1);

acceptor_.async_accept(new_connection->socket(), callback); }

I still can't see your monitor from where I'm sitting, so I don't know what errors it produces. If you want anything more helpful than guesswork, then tell people what the errors are; don't just say that it fails. (http://www.catb.org/esr/faqs/smart-questions.html) First off, is that & actually in your source code or is it only being added when you email? You need to fix that if it's actually in your code. Secondly, you shouldn't specify parameter names in the type signature of std::function; just use types. Otherwise, that code looks like it should work, assuming handle_accept has a compatible signature.