[asio-users] Should async_accept be retried on failure?

[John S. <js...@gm...>]

Take the following code snippet:

void Acceptor::StartNextAccept()
{
    // _acceptor is of type:
std::shared_ptr<boost::asio::ip::tcp::acceptor> _acceptor;
    _acceptor->async_accept([this](const boost::system::error_code& ec,
net::ip::tcp::socket sock) {
        if (ec) {
            LogErrorCode(ec);  // log on failure
        }
        else {
            _callback(sock);  // pass socket on to the next component
            StartNextAccept(); // enqueue next accept
        }
    });
}

In the fail case, the code just logs that an error happened and stops
posting new async_accept calls.  The socket effectively stops processing
new connections.

Are there classes of errors for async_accept, outside of the listen socket
being {misconfigured, bad, closed, cancelled}, such that retrying the
async_accept call would make sense?  Or would that just be an infinite loop
of errors?

Or put another way, can accept calls fail transiently?  If so, what error
codes to pay attention to?

Thanks,
jrs