Keepalive

[akashihi]

2056

#2056

A keepalive will be a nice addition to the current http server.

[gardster]

Please add keep_alive information and corresponding constants (5 seconds, 512 connections to the documentation)

[akashihi]

Please add keep_alive information and corresponding constants (5 seconds, 512 connections to the documentation)

done

[gardster]

Code looks good to me.
Please add the corresponding changelog entry.
Also there are several commits about node flags. Do they relevant for your change?

[akashihi]

Please add the corresponding changelog entry.
Done
Also there are several commits about node flags. Do they relevant for your change?
Rebase glitch, should be fixed now

[gardster]

LGTM

[hy05190134]

void Connection::start()
29 {
30 TCP_socket.async_read_some(
31 boost::asio::buffer(incoming_data_buffer),
32 strand.wrap(boost::bind(&Connection::handle_read,
33 this->shared_from_this(),
34 boost::asio::placeholders::error,
35 boost::asio::placeholders::bytes_transferred)));
36
37 if (keep_alive)
38 {
39 // Ok, we know it is not a first request, as we switched to keepalive
40 timer.cancel();
41 timer.expires_from_now(boost::posix_time::seconds(keepalive_timeout));
42 timer.async_wait(std::bind(
43 &Connection::handle_timeout, this->shared_from_this(), std::placeholders::_1));
44 }
45 }
46
47 void Connection::handle_read(const boost::system::error_code &error, std::size_t bytes_transferred)
48 {
49 if (error)
50 {
51 return;
52 }
53
54 if (keep_alive)
55 {
56 timer.cancel();
57 timer.expires_from_now(boost::posix_time::seconds(0));
58 }

I wonder the keepalive whether work since you set expire_from_now in start() then expires_from_now in handle_read(), but every start() will execute handle_read()

[hy05190134]

@akashihi @gardster

[hy05190134]

void Connection::handle_read(const boost::system::error_code &error, std::size_t bytes_transferred)
48 {
49 if (error)
50 {
51 return;
52 }
53

when read some bytes error, why not execute handle_shutdown right now?

[mjjbell]

when read some bytes error, why not execute handle_shutdown right now?

This was added recently to fix a file exhaustion issue - see #6112