Reordered netplay and NAT structs to follow the coding guidelines more closely.
Moved part of the chat struct out of the program's image and into the heap.
* Netplay Stuff
## PROTOCOL FALLBACK
In order to support older clients a protocol fallback system was introduced.
The host will no longer send its header automatically after a TCP connection is established, instead, it awaits for the client to send his before determining which protocol this connection is going to operate on.
Netplay has now two protocols, a low protocol and a high protocol; the low protocol is the minimum protocol it supports, while the high protocol is the highest protocol it can operate on.
To fully support older clients, a hack was necessary: sending the high protocol in the unused client's header salt field, while keeping the protocol field to the low protocol. Without this hack we would only be able to support older clients if a newer client was the host.
Any future system can make use of this system by checking connection->netplay_protocol, which is available for both the client and host.
## NETPLAY CHAT
Starting with protocol 6, netplay chat is available through the new NETPLAY_CMD_PLAYER_CHAT command.
Limitations of the command code, which causes a disconnection on unknown commands, makes this system not possible on protocol 5.
Protocol 5 connections can neither send nor receive chat, but other netplay operations are unaffected.
Clients send chat as a string to the server, and it's the server's sole responsability to relay chat messages.
As of now, sending chat uses RetroArch's input menu, while the display of on-screen chat uses a widget overlay and RetroArch's notifications as a fallback.
If a new overlay and/or input system is desired, no backwards compatibility changes need to be made.
Only clients in playing mode (as opposed to spectating mode) can send and receive chat.
## SETTINGS SHARING
Some settings are better used when both host and clients share the same configuration.
As of protocol 6, the following settings will be shared from host to clients (without altering a client's configuration file): input latency frames and allow pausing.
## NETPLAY TUNNEL/MITM
With the current MITM system being defunct (at least as of 1.9.X), a new system was in order to solve most if not all of the problems with the current system.
This new system uses a tunneling approach, which is similar to most VPN and tunneling services around.
Tunnel commands:
RATS[unique id] (RetroArch Tunnel Session) - 16 bytes -> When this command is sent with a zeroed unique id, the tunnel server interprets this as a netplay host wanting to create a new session, in this case, the same command is returned to the host, but now with its unique session id. When a client needs to connect to a host, this command is sent with the unique session id of the host, causing the tunnel server to send a RATL command to the host.
RATL[unique id] (RetroArch Tunnel Link) - 16 bytes -> The tunnel server sends this command to the host when a client wants to connect to the host. Once the host receives this command, it establishes a new connection to the tunnel server, sending this command together with the client's unique id through this new connection, causing the tunnel server to link this connection to the connection of the client.
RATP (RetroArch Tunnel Ping) - 4 bytes -> The tunnel server sends this command to verify that the host, whom the session belongs to, is still around. The host replies with the same command. A session is closed if the tunnel server can not verify that the host is alive.
Operations:
Host -> Instead of listening and accepting connections, it connects to the tunnel server, requests a new session and then monitor this connection for new linking requests. Once a request is received, it establishes a new connection to the tunnel server for linking with a client. The tunnel server's address and port are obtained by querying the lobby server. The host will publish its session id together with the rest of its info to the lobby server.
Client -> It connects to the tunnel server and then sends the session id of the host it wants to connect to. A host's session id is obtained from the json data sent by the lobby server.
Improvements (from current MITM system):
No longer a risk of TCP port exhaustion; we only use one port now at the tunnel server.
Very little cpu usage. About 95% net I/O bound now.
Future backwards compatible with any and all changes to netplay as it no longer runs any netplay logic at MITM servers.
No longer operates the host in client mode, which was a source of many of the current problems.
Cleaner and more maintainable system and code.
Notable functions:
netplay_mitm_query -> Grabs the tunnel's address and port from the lobby server.
init_tcp_socket -> Handles the creation and operation mode of the TCP socket based on whether it's host, host+MITM or client.
handle_mitm_connection -> Creates and completes linking connections and replies to ping commands (only 1 of each per call to not affect performance).
## MISC
Ping Limiter: If a client's estimated latency to the server is higher than this value, connection will be dropped just before finishing the netplay handshake.
Ping Counter: A ping counter (similar to the FPS one) can be shown in the bottom right corner of the screen, if you are connected to a host.
LAN Discovery: Refactored and moved to its own "Refresh Netplay LAN List" button.
## FIXES
Many minor fixes to the current netplay implementation are also included.
* Remove NETPLAY_TEST_BUILD
Unsurprisingly, netplay and runahead are wildly incompatible; both rely
on internal rewinding, without communicating this fact to each other.
Somewhat more surprisingly, netplay already has all the infrastructure
for negative input latency, as it's structurally the same as receiving
delayed input from a peer. This patch makes the two features
"compatible" by disabling runahead per se when netplay is active, and
using runahead's configuration to adjust netplay's own input latency
feature, which is now allowed to be negative. The effect is mostly the
same (modulo the second core support), and it doesn't confuse netplay
peers.
An unfortunate hack for cores that translate the keyboard input device
into keydown/keyup events without saving that in the savestate. We
simply replay the previous frame's input before loading the rewound
frame's state, to assure that both the state and the keyup/down state
are correct.
Ideally, cores would save this as part of the state, but it's a bit
proximal and a fairly significant change for those that use it, so it's
easier to fix in netplay.
This abstracts away the details of particular input devices for netplay,
and adds support for mice and (similar) lightguns. Unfortunately, in
practice, no core handles mice or lightguns in a savestate-safe way, so
they need to be used in stateless mode to be reliable, but they do work.
(1) All mode change code unified, so server mode changes and client mode
changes and announcements go through the same functions
(2) New messages which are translateable and work with multiple input
devices
The new input handling makes adding and removing players more
complicated, since data can be present that's not expected from the
connected clients list, or absent that's expected in the list but
actually shouldn't be there.
Netplay state demotions, i.e. changes from playing to spectating or
disconnected states, could cause chain disconnections of all other
clients. This was due to a bug in when MODE change messages were sent.
Clients rely on the server sending all messages in its own order, and as
a consequence, the server typically holds messages for retransmission
until they can be retransmitted at the correct time. MODE messages were
not held, so could be sent early. When they were sent early, this caused
other clients to panic and disconnect.
A smaller but much stupider secondary bug was also fixed, in which the
first connection could be dropped due simply to writing connections[0]
instead of connections[i] somewhere.
This commit adds support for temporary desync in netplay. When frontend
features that can't be truly synced, in particular rewind, are used,
netplay is momentarily disabled. As soon as the feature finished, e.g. a
rewind ending, netplay resumes with a state load. For rewind, netplay
peers won't actually experience the effect of rewind, but they will load
the rewound state.
This commit makes spectator mode and slave mode in netplay always stay
ahead of the input, thereby avoiding rewinds, which is sort of the
point. This also changes catch-up detection to be a bit less eager, so
that they hopefully don't flap between stalling for server input and
catching up with that input.
This patch transfers core_reset across netplay. Resets effectively
worked before thanks to check_frames, but this makes resets work even
without check_frames, and in particular should allow resets to force
sync in savestateless cores, bringing them one step closer to actually
being usable by non-experts.
