Initial revision

[taylor/freespace2.git] / src / network / multi_voice.cpp

/*
 * $Logfile: /Freespace2/code/Network/multi_voice.cpp $
 * $Revision$
 * $Date$
 * $Author$
 *
 * $Log$
 * Revision 1.1  2002/05/03 03:28:10  root
 * Initial revision
 * 
 * 
 * 9     3/10/99 6:50p Dave
 * Changed the way we buffer packets for all clients. Optimized turret
 * fired packets. Did some weapon firing optimizations.
 * 
 * 8     3/09/99 6:24p Dave
 * More work on object update revamping. Identified several sources of
 * unnecessary bandwidth.
 * 
 * 7     1/08/99 4:42p Anoop
 * Fixed memory overwrite problem.
 * 
 * 6     11/19/98 4:19p Dave
 * Put IPX sockets back in psnet. Consolidated all multiplayer config
 * files into one.
 * 
 * 5     11/19/98 8:03a Dave
 * Full support for D3-style reliable sockets. Revamped packet lag/loss
 * system, made it receiver side and at the lowest possible level.
 * 
 * 4     11/17/98 11:12a Dave
 * Removed player identification by address. Now assign explicit id #'s.
 * 
 * 3     11/05/98 5:55p Dave
 * Big pass at reducing #includes
 * 
 * 2     10/07/98 10:53a Dave
 * Initial checkin.
 * 
 * 1     10/07/98 10:50a Dave
 * 
 * 43    6/16/98 4:22p Allender
 * fix Net_player reference problem
 * 
 * 42    6/13/98 9:32p Mike
 * Kill last character in file which caused "Find in Files" to report the
 * file as "not a text file."
 * 
 * 41    6/13/98 6:01p Hoffoss
 * Externalized all new (or forgot to be added) strings to all the code.
 * 
 * 40    6/13/98 3:19p Hoffoss
 * NOX()ed out a bunch of strings that shouldn't be translated.
 * 
 * 39    6/12/98 2:49p Dave
 * Patch 1.02 changes.
 * 
 * 38    5/27/98 4:14p Dave
 * Spiffed up text display for voice status.
 * 
 * 37    5/27/98 1:15p Dave
 * Change pxo login failure popup text. Put in message display of player
 * who is currently doing voice.
 * 
 * 36    5/24/98 3:45a Dave
 * Minor object update fixes. Justify channel information on PXO. Add a
 * bunch of configuration stuff for the standalone.
 * 
 * 35    5/24/98 12:12a Frank
 * Make sure recording clients don't ever send/store/handle packets past
 * the limit of the accum buffer.
 * 
 * 34    5/22/98 9:35p Dave
 * Put in channel based support for PXO. Put in "shutdown" button for
 * standalone. UI tweaks for TvT
 * 
 * 33    5/20/98 2:24a Dave
 * Fixed server side voice muting. Tweaked multi debrief/endgame
 * sequencing a bit. Much friendlier for stats tossing/accepting now.
 * 
 * 32    4/25/98 2:02p Dave
 * Put in multiplayer context help screens. Reworked ingame join ship
 * select screen. Fixed places where network timestamps get hosed.
 * 
 * 31    4/21/98 4:44p Dave
 * Implement Vasudan ships in multiplayer. Added a debug function to bash
 * player rank. Fixed a few rtvoice buffer overrun problems. Fixed ui
 * problem in options screen. 
 * 
 * 30    4/17/98 5:27p Dave
 * More work on the multi options screen. Fixed many minor ui todo bugs.
 * 
 * 29    4/09/98 11:01p Dave
 * Put in new multi host options screen. Tweaked multiplayer options a
 * bit.
 * 
 * 28    4/08/98 10:11a Dave
 * Fixed a voice packet indexing problem.
 * 
 * 27    4/07/98 5:42p Dave
 * Put in support for ui display of voice system status (recording,
 * playing back, etc). Make sure main hall music is stopped before
 * entering a multiplayer game via ingame join.
 * 
 * 26    4/04/98 4:22p Dave
 * First rev of UDP reliable sockets is done. Seems to work well if not
 * overly burdened.
 * 
 * 25    4/03/98 4:22p Allender
 * Fixed a bug where voice recording was not cut off after MAX_TIME.
 * 
 * 24    4/03/98 1:03a Dave
 * First pass at unreliable guaranteed delivery packets.
 * 
 * 23    3/30/98 6:27p Dave
 * Put in a more official set of multiplayer options, including a system
 * for distributing netplayer and settings.
 * 
 * 22    3/27/98 9:46a Dave
 * Made default quality of sound 10.
 * 
 * 21    3/26/98 6:01p Dave
 * Put in file checksumming routine in cfile. Made pilot pic xferring more
 * robust. Cut header size of voice data packets in half. Put in
 * restricted game host query system.
 * 
 * 20    3/25/98 8:36p Dave
 * Removed potential bug spots from voice code.
 * 
 * 19    3/25/98 2:16p Dave
 * Select random default image for newly created pilots. Fixed several
 * multi-pause messaging bugs. Begin work on online help for multiplayer
 * keys.
 * 
 * 18    3/24/98 5:00p Dave
 * Fixed several ui bugs. Put in pre and post voice stream playback sound
 * fx. Put in error specific popups for clients getting dropped from games
 * through actions other than their own.
 * 
 * 17    3/23/98 7:18p Dave
 * Converted voice to use streaming of packets as they are recorded.
 * 
 * 16    3/22/98 7:13p Lawrance
 * Get streaming of recording voice working
 * 
 * 15    3/20/98 1:01p Dave
 * Made the token system for multi_voice more reliable. Fixed a bug where
 * standalones and other server who can't play sound don't route
 * multi_voice correctly.
 * 
 * 14    3/19/98 5:05p Dave
 * Put in support for targeted multiplayer text and voice messaging (all,
 * friendly, hostile, individual).
 * 
 * 13    3/18/98 5:52p Dave
 * Put in netgame password popup. Numerous ui changes. Laid groundwork for
 * streamed multi_voice data.
 * 
 * 12    3/18/98 3:32p Dave
 * Put in a hook for streamed rtvoice data from the rtvoice system.
 * 
 * 11    3/18/98 9:43a Dave
 * Removed an uninitialized data warning.
 * 
 * 10    3/17/98 12:30a Dave
 * Put in hud support for rtvoice. Several ui interface changes.
 * 
 * 9     3/16/98 2:35p Dave
 * Numerous bug fixes. Made the "cue sound" sound play before incoming
 * voice. 
 * 
 * 8     3/15/98 4:17p Dave
 * Fixed oberver hud problems. Put in handy netplayer macros. Reduced size
 * of network orientation matrices.
 * 
 * 7     3/04/98 4:36p Dave
 * Fixed a compiler warning.
 * 
 * 6     2/26/98 6:01p Dave
 * Made voice playback work correctly under NT.
 * 
 * 5     2/26/98 5:01p Dave
 * Fixed a multi_voice_reset() bug.
 * 
 * 4     2/26/98 4:21p Dave
 * More robust multiplayer voice.
 * 
 * 3     2/25/98 8:58p Dave
 * Got first full implementation of multi voice done.
 * 
 * 2     2/24/98 11:56p Lawrance
 * Change real-time voice code to provide the uncompressed size on decode.
 * 
 * 1     2/24/98 10:12p Dave
 * Initial pass at multiplayer voice streaming.
 *  
 * $NoKeywords: $
 */

#include "timer.h"
#include "key.h"
#include "freespace.h"
#include "gamesequence.h"
#include "multimsgs.h"
#include "multiutil.h"
#include "multi_voice.h"
#include "multi_pmsg.h"
#include "gamesnd.h"
#include "rtvoice.h"
#include "psnet.h"
#include "optionsmenumulti.h"
#include "stand_gui.h"
#include "multi.h"


// --------------------------------------------------------------------------------------------------
// MULTI VOICE DEFINES/VARS
//

// #define MULTI_VOICE_POST_DECOMPRESS                                                                  // when we're _not_ using streaming
#define MULTI_VOICE_PRE_DECOMPRESS                                                                              // when we _are_ using streaming

#define MULTI_VOICE_VERBOSE                                                                                             // keep this defined for verbose debug output

#define MULTI_VOICE_LOCAL_ECHO                                                                                  // keep this defined for local echo of recorded network voice

// flag indicating the status of the multi voice system
int Multi_voice_inited = 0;
int Multi_voice_can_record = 0;
int Multi_voice_can_play = 0;
int Multi_voice_send_mode = MULTI_MSG_NONE;                                                     // gotten from the multi_msg system when we start recording

// packet code defines
#define MV_CODE_GIVE_TOKEN                                                              0                                       // received player side - he now has the token to speak
#define MV_CODE_DENY_TOKEN                                                              1                                       // received player side - server has denied this request
#define MV_CODE_TAKE_TOKEN                                                              2                                       // received player side - the server is forcibly taking his token
#define MV_CODE_RELEASE_TOKEN                                                   3                                       // received server side - player is relinquishing token
#define MV_CODE_REQUEST_TOKEN                                                   4                                       // received server side - player is requesting token
#define MV_CODE_PLAYER_PREFS                                                    5                                       // received server side - player bitflags for who he'll receive from
#define MV_CODE_DATA                                                                            6                                       // sound data
#define MV_CODE_DATA_DUMMY                                                              7                                       // in place of a packet which has been deemed too large, so that receivers don't time out early

// default quality of sound
#define MV_DEFAULT_QOS                                                                  10                                      // default quality of sound
int Multi_voice_qos;                                                                                                                    // default quality of sound

// sounds added to the front and end of a playing voice stream (set to -1 if none are wanted)
#define MULTI_VOICE_PRE_SOUND                                                   SND_CUE_VOICE
#define MULTI_VOICE_POST_SOUND                                          SND_END_VOICE
int Multi_voice_pre_sound_size = 0;

// sound data

// NOTE : the following 2 defines should be used for reference only. they represent the worst case situation,
//                       sending voice to a specific target under IPX. you should use multi_voice_max_chunk_size(...) when 
//        determining if a given chunk will fit into an individual freespace packet
// max size of a data packet header (note, this changes as the code itself changes - should probably never use this except for reference)
#define MULTI_VOICE_MAX_HEADER_SIZE                                     22
// size of an individual chunk (CHUNK == block of data stuck into a packet), in the worst case of header size (see above)
#define MULTI_VOICE_MAX_CHUNK_SIZE                                      488

// total max size of an incoming or an outgoing uncompressed buffer (note this is probably too big, but we won't worry about that for now)
#define MULTI_VOICE_MAX_BUFFER_SIZE                                     ((1<<16)+(1<<14))               // 80k

// overall size of an total accum buffer for a stream
#define MULTI_VOICE_ACCUM_BUFFER_SIZE                           (1<<14)                                 // 16k

// how many accum buffers need to be in a total accum buffer
// NOTE : we reference MULTI_VOICE_MAX_CHUNK_SIZE here because it is worst case. ie, we'll always have enough
//        accum buffers in anything better than the worst case if we use MULTI_VOICE_MAX_CHUNK_SIZE
#define MULTI_VOICE_ACCUM_BUFFER_COUNT                          (MULTI_VOICE_ACCUM_BUFFER_SIZE / MULTI_VOICE_MAX_CHUNK_SIZE)

int Multi_voice_max_time;                                                                                                       // current maximum recording time
char *Multi_voice_record_buffer = NULL;                                                         // buffer for recording back voice
char *Multi_voice_playback_buffer = NULL;                                                               // buffer for processing the accum buffer and playing the result

// DEBUG CODE
#ifdef MULTI_VOICE_POST_DECOMPRESS
        char Multi_voice_unpack_buffer[MULTI_VOICE_MAX_BUFFER_SIZE];
#endif

// the max amount of tokens we want to be floating about (max sound streams)
#define MULTI_VOICE_MAX_STREAMS                                         1

// voice algorithm stuff
// it would probably be good to base the timeout time on some multiple of our average ping to the server
#define MV_ALG_TIMEOUT  500                                                                                             // if start get new data for a window then a pause this long, play the window
int Multi_voice_stamps[MULTI_VOICE_MAX_STREAMS];

// NOTE : this should be > then MULTI_VOICE_MAX_TIME + the time for the data to come over a network connection!!
#define MULTI_VOICE_TOKEN_TIMEOUT                                       7000                            // timeout - server will take the token back if he does not hear from the guy in this amount of time

#define MULTI_VOICE_TOKEN_RELEASE_WAIT                          (1.0f)                  // wait 1 second

// the token index of a voice stream is set to one of these values, or the index of the player who has the token
#define MULTI_VOICE_TOKEN_INDEX_FREE                            -1                                      // the token (and the stream are free)
#define MULTI_VOICE_TOKEN_INDEX_RELEASED                        0xDEADBEAD              // the token has been released but the stream is still active

typedef struct voice_stream {           
        int token_status;                                                                                                                       // status of the token (player index if a player has it) or one of the above defines
        int token_stamp;                                                                                                                        // timestamp for the MULTI_VOICE_TOKEN_TIMEOUT

        short stream_from;                                                                                                              // id of the player the stream is coming from

        ubyte *accum_buffer[MULTI_VOICE_ACCUM_BUFFER_COUNT];                    // accum buffer
        ubyte accum_buffer_flags[MULTI_VOICE_ACCUM_BUFFER_COUNT];       // flag indicating the existence of a given accum (sub)buffer
        ushort accum_buffer_usize[MULTI_VOICE_ACCUM_BUFFER_COUNT];      // uncompressed size of the corresponding (sub)buffer
        ushort accum_buffer_csize[MULTI_VOICE_ACCUM_BUFFER_COUNT];      // compressed size of the corresponding (sub)buffer
        double accum_buffer_gain[MULTI_VOICE_ACCUM_BUFFER_COUNT];       // gain of the corresponding (sub)buffer
                
        ubyte stream_id;                                                                                                                        // stream id #
        fix stream_last_heard;                                                                                                  // last time we heard from this stream  

        fix stream_start_time;                                                                                                  // time the stream started playing
        int stream_snd_handle;                                                                                                  // sound playing instance handle
        int stream_rtvoice_handle;                                                                                              // rtvoice buffer handle
} voice_stream;
voice_stream Multi_voice_stream[MULTI_VOICE_MAX_STREAMS];               // voice streams themselves

// player-side data
#define MULTI_VOICE_KEY                                                                 KEY_LAPOSTRO    // key used for realtime voice
int Multi_voice_keydown = 0;                                                                                            // is the record key currently being pressed
int Multi_voice_recording = 0;                                                                                  // flag indicating if we're currently recording or not
int Multi_voice_token = 0;                                                                                                      // if we currently have a token or not
int Multi_voice_recording_stamp = -1;                                                                   // how long we've been recording
ubyte Multi_voice_stream_id = 0;                                                                                        // stream id for the stream we're currently sending
int Multi_voice_current_stream_index = 0;                                                               // packet index of the currently recodring stream
int Multi_voice_current_stream_sent = -1;                                                               // index of packet we've sent up to

// server-side data
ubyte Multi_voice_next_stream_id = 0;                                                                   // kept on the server - given to the next valid token requester
int Multi_voice_player_prefs[MAX_PLAYERS];                                                      // player bitflag preferences

// voice status data - used for determing the result of multi_voice_status
#define MULTI_VOICE_DENIED_TIME                                         1000                            // how long to display the "denied" status
int Multi_voice_denied_stamp = -1;                                                                              // timestamp for when we got denied a token

// local muting preferences
int Multi_voice_local_prefs = 0xffffffff;


// --------------------------------------------------------------------------------------------------
// MULTI VOICE FORWARD DECLARATIONS
//

// process voice details as the server
void multi_voice_server_process();

// process voice details as a player (may also be the server)
void multi_voice_player_process();

// determine if the voice key is down this frame
int multi_voice_keydown();

// find the voice stream index by token player index
int multi_voice_find_token(int player_index);

// <server> gives the token to a given player
void multi_voice_give_token(int stream_index,int player_index);

// <server> takes the token from a given stream entry
void multi_voice_take_token(int stream_index);

// <server> tells the client he's been denied on this request
void multi_voice_deny_token(int player_index);

// <player> releases the token back to the server
void multi_voice_release_token();

// <player> requests the token from the server
void multi_voice_request_token();

// <server> process a request for the token
void multi_voice_process_token_request(int player_index);

// free up any memory which may have been malloced
void multi_voice_free_all();

// <player> send the currently recorded sound
void multi_voice_player_send_stream();

// process incoming sound data, return bytes processed
int multi_voice_process_data(ubyte *data, int player_index,int msg_mode,net_player *target);

// <server> increment the current stream id#
void multi_voice_inc_stream_id();

// flush any old sound stream data because we've started to receive data for a new stream
void multi_voice_flush_old_stream(int stream_index);

// route sound data through the server to all appropriate players
void multi_voice_route_data(ubyte *data, int packet_size,int player_index,int mode,net_player *target);

// find the stream to apply incoming sound data to, freeing up old ones as necessary
int multi_voice_get_stream(int stream_id);

// NOTE : these 4 functions can be arbitrarily written to perform in any way necessary. This way the algorithm is
//        completely seperate from the transport and token layers
// initialize the smart algorithm
void multi_voice_alg_init();

// process incoming sound data in whatever way necessary (this function should take care of playing data when necessary)
void multi_voice_alg_process_data(int player_index,int stream_index,ushort chunk_index,ushort chunk_size);              

// process existing streams
void multi_voice_alg_process_streams();

// we are going to flush the current stream because we have started to receive data for a new one. do something first
void multi_voice_alg_flush_old_stream(int stream_index);

// is the given sound stream playing (compares uncompressed sound size with current playback position)
int multi_voice_stream_playing(int stream_index);

// tack on a post voice sound (pass -1 for none)
// return final buffer size
int multi_voice_mix(int post_sound,char *data,int cur_size,int max_size);

// send a dummy packet in the place of a too-large data packet
void multi_voice_send_dummy_packet();

// process a dummy data packet
int multi_voice_process_data_dummy(ubyte *data);

// max size of a sound chunk which we can fit into a packet
int multi_voice_max_chunk_size(int msg_mode);

// process a player preferences packet, return bytes processed
int multi_voice_process_player_prefs(ubyte *data,int player_index);

// process and play the current window of sound stream data we have. reset the window for the next incoming data as well
void multi_voice_alg_play_window(int stream_index);

// send all pending voice packets
void multi_voice_client_send_pending();


// --------------------------------------------------------------------------------------------------
// MULTI VOICE FUNCTIONS
//

// initialize the multiplayer voice system
void multi_voice_init()
{
        int idx,s_idx,pre_size,pre_sound;

        // if the voice system is already initialized, just reset some stuff
        if(Multi_voice_inited){
                multi_voice_reset();
                return;
        }

        // set the default quality of sound
        Multi_voice_qos = MV_DEFAULT_QOS;

        // if we're the standalone server, we can't record _or_ playback, but we can still route data and manage tokens
        if(Game_mode & GM_STANDALONE_SERVER){
                Multi_voice_can_record = 0;
                Multi_voice_can_play = 0;
        } else {
                // initialize the realtime voice module
                if(rtvoice_init_recording(Multi_voice_qos)){
                        nprintf(("Network","MULTI VOICE : Error initializing rtvoice - recording will not be possible\n"));
                        Multi_voice_can_record = 0;
                } else {
                        Multi_voice_can_record = 1;
                }               

                if(rtvoice_init_playback()){            
                        nprintf(("Network","MULTI VOICE : Error initializing rtvoice - playback will not be possible\n"));
                        Multi_voice_can_play = 0;
                } else {
                        Multi_voice_can_play = 1;
                }

                // _always_ set the quality of server
                multi_voice_set_vars(MV_DEFAULT_QOS,MULTI_VOICE_MAX_TIME);
        }

        // initialize player-side data
        Multi_voice_token = 0;
        Multi_voice_keydown = 0;
        Multi_voice_recording = 0;      
        Multi_voice_stream_id = 0;
        Multi_voice_recording_stamp = -1;
        Multi_voice_current_stream_index = 0;
        Multi_voice_current_stream_sent = -1;

        // initialize server-side data
        memset(Multi_voice_player_prefs,0xff,sizeof(int)*MAX_PLAYERS);                          
        Multi_voice_next_stream_id = 0;

        Multi_voice_local_prefs = 0xffffffff;

        // initialize the sound buffers
        Multi_voice_record_buffer = NULL;       

        Multi_voice_playback_buffer = NULL;
        Multi_voice_pre_sound_size = 0;
        if(Multi_voice_can_play){
                // attempt to allocate the buffer
                Multi_voice_playback_buffer = (char*)malloc(MULTI_VOICE_MAX_BUFFER_SIZE);
                if(Multi_voice_playback_buffer == NULL){
                        nprintf(("Network","MULTI VOICE : Error allocating playback buffer - playback will not be possible\n"));                
                        Multi_voice_can_play = 0;               
                } 

                // attempt to copy in the "pre" voice sound
                pre_sound = snd_load(&Snds[MULTI_VOICE_PRE_SOUND]);
                if(pre_sound != -1){
                        // get the pre-sound size
                        if((snd_size(pre_sound,&pre_size) != -1) && (pre_size < MULTI_VOICE_MAX_BUFFER_SIZE)){
                                snd_get_data(pre_sound,Multi_voice_playback_buffer);
                                Multi_voice_pre_sound_size = pre_size;
                        } else {
                                Multi_voice_pre_sound_size = 0;
                        }
                } else {
                        Multi_voice_pre_sound_size = 0;
                }
        }

        // initialize the streams       
        memset(Multi_voice_stream,0,sizeof(voice_stream) * MULTI_VOICE_MAX_STREAMS);    
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                Multi_voice_stream[idx].token_status = MULTI_VOICE_TOKEN_INDEX_FREE;
                Multi_voice_stream[idx].token_stamp = -1;               
                Multi_voice_stream[idx].stream_snd_handle = -1;

                // get a playback buffer handle
                if(Multi_voice_can_play){
                        Multi_voice_stream[idx].stream_rtvoice_handle = -1;
                        Multi_voice_stream[idx].stream_rtvoice_handle = rtvoice_create_playback_buffer();
                        if(Multi_voice_stream[idx].stream_rtvoice_handle == -1){
                                nprintf(("Network","MULTI VOICE : Error getting rtvoice buffer handle - playback will not be possible!\n"));
                                multi_voice_free_all(); 

                                Multi_voice_can_play = 0;
                        }
                                        
                        // allocate the accum buffer
                        for(s_idx=0;s_idx<MULTI_VOICE_ACCUM_BUFFER_COUNT;s_idx++){
                                Multi_voice_stream[idx].accum_buffer[s_idx] = NULL;
                                Multi_voice_stream[idx].accum_buffer[s_idx] = (ubyte*)malloc(MULTI_VOICE_ACCUM_BUFFER_SIZE);
                                if(Multi_voice_stream[idx].accum_buffer[s_idx] == NULL){
                                        nprintf(("Network","MULTI VOICE : Error allocating accum buffer - playback will not be possible\n"));
                                        multi_voice_free_all();
                                        
                                        Multi_voice_can_play = 0;
                                }
                        }
                }
        }       

        // initialize the default max time
        Multi_voice_max_time = MULTI_VOICE_MAX_TIME;

        // initialize voice status data
        Multi_voice_denied_stamp = -1;
        
        // initialize the smart algorithm
        multi_voice_alg_init(); 
        
        Multi_voice_inited = 1;
}

// shutdown the multiplayer voice system
void multi_voice_close()
{
        int idx;
        
        // if the voice system isn't already initialized, don't do anything
        if(!Multi_voice_inited){
                return;
        }

        // free up buffers
        multi_voice_free_all();

        // release all the rtvoice buffers
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(Multi_voice_stream[idx].stream_rtvoice_handle != -1){
                        rtvoice_free_playback_buffer(Multi_voice_stream[idx].stream_rtvoice_handle);
                        Multi_voice_stream[idx].stream_rtvoice_handle = -1;
                        Multi_voice_stream[idx].stream_snd_handle = -1;
                }
        }

        // close the realtime voice module
        rtvoice_close_recording();
        rtvoice_close_playback();

        Multi_voice_inited = 0;
}

// reset between levels
void multi_voice_reset()
{
        int idx;

#ifdef MULTI_VOICE_VERBOSE
        nprintf(("Network","MULTI VOICE : Resetting\n"));
#endif

        Assert(Multi_voice_inited);     

        // if we're the standalone server, we can't record _or_ playback, but we can still route data and manage tokens
        if(Game_mode & GM_STANDALONE_SERVER){
                Multi_voice_can_record = 0;
                Multi_voice_can_play = 0;
        } 

        // initialize player-side data
        Multi_voice_token = 0;
        Multi_voice_keydown = 0;
        Multi_voice_recording = 0;      
        Multi_voice_stream_id = 0;
        Multi_voice_recording_stamp = -1;

        // initialize server-side data
        memset(Multi_voice_player_prefs,0xff,sizeof(int)*MAX_PLAYERS);                          
        Multi_voice_local_prefs = 0xffffffff;
        Multi_voice_next_stream_id = 0;

        // initialize the sound buffers
        Multi_voice_record_buffer = NULL;       
        
        // initialize the streams               
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                Multi_voice_stream[idx].token_status = MULTI_VOICE_TOKEN_INDEX_FREE;
                Multi_voice_stream[idx].token_stamp = -1;                                                       
        }       
        
        // initialize the smart algorithm
        multi_voice_alg_init(); 
}

// process all voice details
void multi_voice_process()
{
        int idx;
        
        // don't do anything if the voice module is not initialized
        if((!Multi_voice_inited) || !(Net_player->flags & NETINFO_FLAG_CONNECTED)){
                return;
        }               

        // send all pending voice packets
        multi_voice_client_send_pending();

        // find any playing sound streams which have finished and unmark them
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if((Multi_voice_stream[idx].stream_snd_handle != -1) && !multi_voice_stream_playing(idx)){
                        Multi_voice_stream[idx].stream_snd_handle = -1;
                }
        }

        // process seperately as player or server
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                multi_voice_server_process();
        }

        // all "players" do this, except the standalone who isn't a real player by definition
        if(!(Game_mode & GM_STANDALONE_SERVER)){
                multi_voice_player_process();   
        }

        // everyont calls the general algorithm process function
        multi_voice_alg_process_streams();
}

// voice settings debug console function
void multi_voice_dcf()
{
        dc_get_arg(ARG_STRING);

        // set the quality of sound
        if (strcmp(Dc_arg, NOX("qos")) == 0) {
                dc_get_arg(ARG_INT);
                if(Dc_arg_type & ARG_INT){
                        if((Dc_arg_int >= 1) && (Dc_arg_int <= 10) && (Net_player->flags & NETINFO_FLAG_AM_MASTER)){
                                multi_voice_set_vars(Dc_arg_int,-1);
                                dc_printf("Quality of sound : %d\n",Dc_arg_int);
                        }
                }
        }
}

// the status of the voice system - use this to determine what bitmaps to display, etc see above MULTI_VOICE_STATUS_* defines
int multi_voice_status()
{
        int idx;
        int earliest;
        fix earliest_time;
        
        // if the "denied" timestamp is set, return that as the status
        if(Multi_voice_denied_stamp != -1){
                return MULTI_VOICE_STATUS_DENIED;
        }

        // if we're currently recording (has precedence over playing back a sound from somebody)
        if(Multi_voice_recording){
                return MULTI_VOICE_STATUS_RECORDING;
        }
        
        // find the stream which started playing the farthest back (if any)
        earliest = -1;
        earliest_time = -1;
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                // if we found a playing stream
                if(Multi_voice_stream[idx].stream_snd_handle != -1){
                        if((earliest == -1) || (Multi_voice_stream[idx].stream_start_time < earliest_time)){
                                earliest = idx;
                                earliest_time = Multi_voice_stream[idx].stream_start_time;
                        }
                }
        }
        // if we found a stream
        if(earliest != -1){
                return MULTI_VOICE_STATUS_PLAYING;
        }

        // system is idle
        return MULTI_VOICE_STATUS_IDLE;
}

// update the qos if the current setting is different from the passed in value
void multi_voice_maybe_update_vars(int new_qos,int new_duration)
{
        // if the current qos is different from the passed qos, set it
        if((new_qos != Multi_voice_qos) || (new_duration != Multi_voice_max_time)){
                multi_voice_set_vars(new_qos,new_duration);
        }
}


// --------------------------------------------------------------------------------------------------
// MULTI VOICE FORWARD DECLARATIONS
//

// process voice details as the server
void multi_voice_server_process()
{
        int idx;

        // process all the tokens for all the available streams
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                switch(Multi_voice_stream[idx].token_status){
                // if the token is free, so is the stream - don't do anything
                case MULTI_VOICE_TOKEN_INDEX_FREE:
                        break;

                // if the token has been released - check to see if the stream is "done" (ie, can be marked as FREE once again)
                case MULTI_VOICE_TOKEN_INDEX_RELEASED:
                        // if the stream_last_heard var is -1, it means we never got sound from this guy so free the token up immediately
                        if(Multi_voice_stream[idx].stream_last_heard == -1){
                                Multi_voice_stream[idx].token_status = MULTI_VOICE_TOKEN_INDEX_FREE;                            

#ifdef MULTI_VOICE_VERBOSE
                                nprintf(("Network","MULTI VOICE : freeing released token (no packets)\n"));
#endif
                        } 
                        // if a sufficiently long amount of time has elapsed since he released the token, free it up
                        else {
                                float t1,t2;
                                t1 = f2fl(Multi_voice_stream[idx].stream_last_heard);
                                t2 = f2fl(timer_get_fixed_seconds());
                                if((t2 - t1) >= MULTI_VOICE_TOKEN_RELEASE_WAIT){
                                        Multi_voice_stream[idx].token_status = MULTI_VOICE_TOKEN_INDEX_FREE;

#ifdef MULTI_VOICE_VERBOSE
                                        nprintf(("Network","MULTI VOICE : freeing released token (time elapsed)\n"));
#endif
                                }
                        }
                        break;

                // if the token is still being held by a player
                default :
                        // if the token timestamp has elapsed, take the token back
                        if((Multi_voice_stream[idx].token_stamp != -1) && timestamp_elapsed(Multi_voice_stream[idx].token_stamp)){
                                Assert(Multi_voice_stream[idx].token_status != MULTI_VOICE_TOKEN_INDEX_FREE);
                                multi_voice_take_token(idx);
                        }                               
                        break;
                }
        }

        // for each netplayer, if his token wait timestamp is running, see if it has popped yet
        for(idx=0;idx<MAX_PLAYERS;idx++){
                if(MULTI_CONNECTED(Net_players[idx]) && (Net_players[idx].s_info.voice_token_timestamp != -1) && timestamp_elapsed(Net_players[idx].s_info.voice_token_timestamp)){
                        // unset it so that he can have the token again
                        Net_players[idx].s_info.voice_token_timestamp = -1;
                }
        }
}

// process voice details as a player (may also be the server)
void multi_voice_player_process()
{
        // if the voice key is down for the first time this frame, send a request for the token
        if(!Multi_voice_keydown && multi_voice_keydown() && Multi_voice_can_record && !(Netgame.options.flags & MSO_FLAG_NO_VOICE)){
                // mark the key as being down
                Multi_voice_keydown = 1;

                // send a request for a token
                multi_voice_request_token();

#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : Request\n"));
#endif
        }       
        
        // if the key is still being pressed
        if(Multi_voice_keydown && multi_voice_keydown() && Multi_voice_can_record){
                // if we have the token
                if(Multi_voice_token){                  
                        // if we're not already recording, start recording
                        if(!Multi_voice_recording){
#ifdef MULTI_VOICE_VERBOSE
                                nprintf(("Network","MULTI VOICE : RECORD %d\n",(int)Multi_voice_stream_id));
#endif  
                                // flush the old stream
                                multi_voice_flush_old_stream(0);

                                // start the recording process with the appropriate callback function
                                if(rtvoice_start_recording(multi_voice_process_next_chunk)){
                                        nprintf(("Network","MULTI VOICE : Error initializing recording!\n"));                                   
                                        return;
                                }
                                
                                // set myself to be recording
                                Multi_voice_recording = 1;

                                // set the time when I started recording
                                Multi_voice_recording_stamp = timestamp(Multi_voice_max_time);

                                // set the current packet/chunk index to 0
                                Multi_voice_current_stream_index = 0;
                                Multi_voice_current_stream_sent = 0;
                                
                                // get the proper messaging mode
                                if(Game_mode & GM_IN_MISSION){
                                        // in mission, paused
                                        if(gameseq_get_state() == GS_STATE_MULTI_PAUSED){
                                                Multi_voice_send_mode = MULTI_MSG_ALL;
                                        } 
                                        // in mission, unpaused
                                        else {
                                                Multi_voice_send_mode = multi_msg_mode();
                                        }
                                } else {
                                        Multi_voice_send_mode = MULTI_MSG_ALL;
                                }
                        }

                        // if we've recorded the max time allowed, send the data
                        if((Multi_voice_recording_stamp != -1) && timestamp_elapsed(Multi_voice_recording_stamp)){
#ifdef MULTI_VOICE_VERBOSE
                                nprintf(("Network","MULTI VOICE : timestamp popped"));
#endif
                                // mark me as no longer recording
                                Multi_voice_recording = 0;                      
                                Multi_voice_current_stream_sent = -1;

                                // stop the recording process
                                rtvoice_stop_recording();                               
                                
#ifdef MULTI_VOICE_POST_DECOMPRESS
                                multi_voice_player_send_stream();
#endif

                                // play my sound locally as well
#ifdef MULTI_VOICE_LOCAL_ECHO   
                                multi_voice_alg_play_window(0);
#endif
                                // release the token back to the server
                                multi_voice_release_token();
                        }
                }
        }
        // if the key has been released
        else if(Multi_voice_keydown && !multi_voice_keydown() && Multi_voice_can_record){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : Release\n"));
#endif

                // mark the kay as not being down
                Multi_voice_keydown = 0;
        
                // if we were recording, send the data
                if(Multi_voice_recording){              
                        // mark me as no longer recording
                        Multi_voice_recording = 0;

                        Multi_voice_current_stream_sent = -1;

                        // stop the recording process
                        rtvoice_stop_recording();                       
                
#ifdef MULTI_VOICE_POST_DECOMPRESS
                        multi_voice_player_send_stream();                       
#endif

                        // play my sound locally as well
#ifdef MULTI_VOICE_LOCAL_ECHO   
                        multi_voice_alg_play_window(0);
#endif

                        // release the token back to the server
                        multi_voice_release_token();
                }               
        }       

        // if the "denied" timestamp is set, but has elapsed or the user has let up on the key, set it to -1
        if((Multi_voice_denied_stamp != -1) && (timestamp_elapsed(Multi_voice_denied_stamp) || !multi_voice_keydown())){
                Multi_voice_denied_stamp = -1;
        }
}

// determine if the voice key is down this frame
int multi_voice_keydown()
{
        // if we're in the options screen, we should never allow the button to be pressed
        if(gameseq_get_state() == GS_STATE_OPTIONS_MENU){
                return 0;
        }

        // if we're pre-game, we should just be checking the keyboard bitflags
        if(!(Game_mode & GM_IN_MISSION)){       
                return (keyd_pressed[MULTI_VOICE_KEY] && !(keyd_pressed[KEY_LSHIFT] || keyd_pressed[KEY_RSHIFT])) ? 1 : 0;
        } 

        // in-mission, paused - treat just like any other "chattable" screen.
        if(gameseq_get_state() == GS_STATE_MULTI_PAUSED){
                return (keyd_pressed[MULTI_VOICE_KEY] && !(keyd_pressed[KEY_LSHIFT] || keyd_pressed[KEY_RSHIFT])) ? 1 : 0;
        }

        // ingame, unpaused, rely on the multi-messaging system (ingame)
        return multi_msg_voice_record();
}

// find the voice stream index by token player index
int multi_voice_find_token(int player_index)
{
        int idx;

        // look through all the existing streams
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(Multi_voice_stream[idx].token_status == player_index){
                        return idx;
                }
        }

        // couldn't find it
        return -1;
}

// <server> gives the token to a given player
void multi_voice_give_token(int stream_index,int player_index)
{
        ubyte data[10],code;
        int packet_size = 0;
        
        // only the server should ever be here
        Assert(Net_player->flags & NETINFO_FLAG_AM_MASTER);

        // set this player as having the token  
        Multi_voice_stream[stream_index].token_status = player_index;
        
        // set the token timeout
        Multi_voice_stream[stream_index].token_stamp = timestamp(MULTI_VOICE_TOKEN_TIMEOUT);

        // set the stream id and increment the count
        Multi_voice_stream[stream_index].stream_id = Multi_voice_next_stream_id;
        multi_voice_inc_stream_id();

        // set the last heard from time to -1 to indicate we've heard no sound from this guy
        Multi_voice_stream[stream_index].stream_last_heard = -1;

#ifdef MULTI_VOICE_VERBOSE
        nprintf(("Network","MULTI VOICE : GIVE TOKEN %d\n",(int)Multi_voice_next_stream_id));   
#endif

        // if we're giving to ourself, don't send any data
        if(Net_player == &Net_players[player_index]){
                Multi_voice_token = 1;

                Multi_voice_stream_id = Multi_voice_stream[stream_index].stream_id;
        } else {
                // send the "give" packet to the guy
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_GIVE_TOKEN;
                ADD_DATA(code);

                // add the current stream id#
                ADD_DATA(Multi_voice_stream[stream_index].stream_id);

                // send reliably                
                multi_io_send_reliable(&Net_players[player_index], data, packet_size);
        }       
}

// <server> takes the token from a given player
void multi_voice_take_token(int stream_index)
{
        ubyte data[10],code;
        int packet_size = 0;

        // only the server should ever be here
        Assert(Net_player->flags & NETINFO_FLAG_AM_MASTER);     

        // if the index is -1, the token has probably been released to us "officially" already
        if((Multi_voice_stream[stream_index].token_status == MULTI_VOICE_TOKEN_INDEX_FREE) || (Multi_voice_stream[stream_index].token_status == MULTI_VOICE_TOKEN_INDEX_RELEASED)){
                Multi_voice_stream[stream_index].token_stamp = -1;
                return;
        }

        // if i'm taking from myself, don't send any data
        if(Net_player == &Net_players[Multi_voice_stream[stream_index].token_status]){
                Multi_voice_token = 0;

                // timestamp this guy so that he can't get the token back immediately
                Net_player->s_info.voice_token_timestamp = timestamp(Netgame.options.voice_token_wait);
        } else {
                // send the "take" packet to the guy
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_TAKE_TOKEN;
                ADD_DATA(code);

                // send reliably                
                multi_io_send_reliable(&Net_players[Multi_voice_stream[stream_index].token_status], data, packet_size);

                // timestamp this guy so that he can't get the token back immediately
                Net_players[Multi_voice_stream[stream_index].token_status].s_info.voice_token_timestamp = timestamp(Netgame.options.voice_token_wait);
        }

        // take the token back from the dude
        Multi_voice_stream[stream_index].token_status = MULTI_VOICE_TOKEN_INDEX_RELEASED;       
        Multi_voice_stream[stream_index].token_stamp = -1;
}

// <server> tells the client he's been denied on this request
void multi_voice_deny_token(int player_index)
{
        ubyte data[10],code;
        int packet_size = 0;

        // only the server should ever be here
        Assert(Net_player->flags & NETINFO_FLAG_AM_MASTER);     
        

        // if i'm denying myself, set the denied timestamp
        if(Net_player == &Net_players[player_index]){   
                Multi_voice_denied_stamp = timestamp(MULTI_VOICE_DENIED_TIME);          
        } else {
                // send the "deny" packet to the guy
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_DENY_TOKEN;
                ADD_DATA(code);

                // send reliably                
                multi_io_send_reliable(&Net_players[player_index], data, packet_size);
        }
}

// <player> releases the token back to the server
void multi_voice_release_token()
{
        ubyte data[10],code;
        int packet_size = 0;

        // I don't have the token anymore
        Multi_voice_token = 0;

        // if i'm the server, don't send any data
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                // mark the token as being released
                int stream_index = multi_voice_find_token(MY_NET_PLAYER_NUM);
                Multi_voice_stream[stream_index].token_status = MULTI_VOICE_TOKEN_INDEX_RELEASED;
                                
                // timestamp this guy so that he can't get the token back immediately
                Net_player->s_info.voice_token_timestamp = timestamp(Netgame.options.voice_token_wait);
        } else {
                // send the "release" packet to the server
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_RELEASE_TOKEN;
                ADD_DATA(code);

                // send reliably                
                multi_io_send_reliable(Net_player, data, packet_size);
        }
}

// <player> requests the token from the server
void multi_voice_request_token()
{
        ubyte data[10],code;
        int packet_size = 0;

        // if i'm the server, process the request right now
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                multi_voice_process_token_request(MY_NET_PLAYER_NUM);
        } else {
                // send the "request" packet to the server
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_REQUEST_TOKEN;
                ADD_DATA(code);

                // send reliably                
                multi_io_send_reliable(Net_player, data, packet_size);
        }
}

// <player> sends hit bitflag settings (who he'll receive sound from, etc)
void multi_voice_set_prefs(int pref_flags)
{
        ubyte data[MAX_PACKET_SIZE],code;
        int idx;
        int packet_size = 0;

        // set the local flags
        Multi_voice_local_prefs = pref_flags;

        // if i'm the server, set the sound prefs right now
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                Multi_voice_player_prefs[MY_NET_PLAYER_NUM] = pref_flags;
        } else {
                // send the prefs to the server
                BUILD_HEADER(VOICE_PACKET);
                code = MV_CODE_PLAYER_PREFS;
                ADD_DATA(code);

                // add the address of all players being ignored
                for(idx=0;idx<MAX_PLAYERS;idx++){
                        if(!(pref_flags & (1<<idx))){
                                code = 0x0;
                                ADD_DATA(code);

                                // add the player's id
                                ADD_DATA(Net_players[idx].player_id);
                        }
                }
                // add final stop byte
                code = 0xff;
                ADD_DATA(code);

                // send reliably                
                multi_io_send_reliable(Net_player, data, packet_size);
        }
}

// set the default voice quality and duration (if server passes -1, he just broadcasts the qos to all clients)
void multi_voice_set_vars(int qos,int duration)
{                                       
        int need_update = 0;
        
        // make sure its in the right range
        if((qos > 0) && (qos <= 10)){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : SETTING QOS %d\n",qos));
#endif 

                // set the default value
                Multi_voice_qos = qos;          

                // set the value in the rtvoice module
                rtvoice_set_qos(Multi_voice_qos);

                // update the netgame settings
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){                                         
                        Netgame.options.voice_qos = (ubyte)Multi_voice_qos;
                        need_update = 1;                        
                }
        }

        // set the maximum duration
        if((duration > 0) && (duration <= MULTI_VOICE_MAX_TIME)){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : SETTING MAX RECORD TIME %d\n",duration));
#endif
                // set the default value
                Multi_voice_max_time = duration;

                // update the netgame settings
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                        Netgame.options.voice_record_time = duration;
                        need_update = 1;
                }
        }       

        // send an options update if necessary
        if(need_update && !(Game_mode & GM_STANDALONE_SERVER)){
                multi_options_update_netgame();
        }
}

// <server> process a request for the token
void multi_voice_process_token_request(int player_index)
{
        int stream_index,idx;
        
        // if we're not doing voice on this server, return now
        if(Netgame.options.flags & MSO_FLAG_NO_VOICE){
                return;
        }

        // if the player's token timestamp is not -1, can't give him the token
        if(Net_players[player_index].s_info.voice_token_timestamp != -1){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : Not giving token because player %s's timestamp hasn't elapsed yet!\n",Net_players[player_index].player->callsign));
                nprintf(("Network","MULTI VOICE : token status %d\n",Multi_voice_stream[0].token_status));
#endif
                // deny the guy
                multi_voice_deny_token(player_index);
                return;
        }

        // attempt to find a free token token
        stream_index = -1;
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(Multi_voice_stream[idx].token_status == MULTI_VOICE_TOKEN_INDEX_FREE){
                        multi_voice_give_token(idx,player_index);
                        return;
                }
        }       
}

// free up any memory which may have been malloced
void multi_voice_free_all()
{
        int idx,s_idx;

        // free up the playback buffer
        if(Multi_voice_playback_buffer != NULL){
                free(Multi_voice_playback_buffer);
                Multi_voice_playback_buffer = NULL;
        }

        // free up the accum buffers
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                for(s_idx=0;s_idx<MULTI_VOICE_ACCUM_BUFFER_COUNT;s_idx++){
                        if(Multi_voice_stream[idx].accum_buffer[s_idx] != NULL){
                                free(Multi_voice_stream[idx].accum_buffer[s_idx]);
                                Multi_voice_stream[idx].accum_buffer[s_idx] = NULL;
                        }
                }
        }       
}

// <player> send the currently recorded sound
void multi_voice_player_send_stream()
{
        ubyte data[MAX_PACKET_SIZE],code,*rbuf,msg_mode,chunk_index;
        ushort chunk_size,uc_size;
        int packet_size = 0;
        int sound_size,size_sent,uncompressed_size,target_index,max_chunk_size;
        float gain;
        double d_gain;

        // we'd better not ever get here as we can't record voice
        Assert(Multi_voice_can_record);

        // get the data 
        rtvoice_get_data((unsigned char**)&Multi_voice_record_buffer,&sound_size,&uncompressed_size,&d_gain);           
        gain = (float)d_gain;

        msg_mode = (ubyte)Multi_voice_send_mode;
        // get the specific target if we're in MSG_TARGET mode
        target_index = -1;
        if(msg_mode == MULTI_MSG_TARGET){
                if(Player_ai->target_objnum != -1){
                        target_index = multi_find_player_by_object(&Objects[Player_ai->target_objnum]);
                        if(target_index == -1){
                                return;
                        }
                } else {
                        return;
                }
        }

        // get the max chunk size
        max_chunk_size = multi_voice_max_chunk_size(Multi_voice_send_mode);

        // go through the data and send all of it
        code = MV_CODE_DATA;
        chunk_index = 0;
        size_sent = 0;
        rbuf = (unsigned char*)Multi_voice_record_buffer;       
        while(size_sent < sound_size){
                // build the header and add the opcode
                BUILD_HEADER(VOICE_PACKET);

                // add the packet code type
                ADD_DATA(code);

                // add the routing data and any necessary targeting information
                ADD_DATA(msg_mode);
                if(msg_mode == MULTI_MSG_TARGET){
                        ADD_DATA(Objects[Net_players[target_index].player->objnum].net_signature);
                }

                // add my id#
                ADD_DATA(Net_player->player_id);

                // add the current stream id#
                ADD_DATA(Multi_voice_stream_id);

                Assert(uncompressed_size < MULTI_VOICE_MAX_BUFFER_SIZE);
                uc_size = (ushort)uncompressed_size;
                ADD_DATA(uc_size);

                // add the chunk index
                ADD_DATA(chunk_index);

                // determine how much we are going to send in this packet
                if((sound_size - size_sent) >= max_chunk_size){
                        chunk_size = (ushort)max_chunk_size;
                } else {
                        chunk_size = (ushort)(sound_size - size_sent);
                }
                ADD_DATA(chunk_size);

                // add the gain
                ADD_DATA(gain);

                // add the chunk of data                
                memcpy(data+packet_size, rbuf,chunk_size);              
                packet_size += chunk_size;

                // send to the server or rebroadcast if I _am_ the server
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                        multi_voice_route_data(data,packet_size,MY_NET_PLAYER_NUM,(int)msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);
                } else {                        
                        multi_io_send(Net_player, data, packet_size);
                }

                // increment the chunk_index
                chunk_index++;

                // increment bytes sent and the buffer
                size_sent += (int)chunk_size;
                rbuf += chunk_size;
        }       
}

// process incoming sound data, return bytes processed
int multi_voice_process_data(ubyte *data, int player_index,int msg_mode,net_player *target)
{
        ubyte stream_id,chunk_index;
        ushort chunk_size,uc_size;      
        short who_from;
        int stream_index;
        float gain;
        int offset = 0;

        // read in all packet data except for the sound chunk itself
        GET_DATA(who_from);
        GET_DATA(stream_id);
        GET_DATA(uc_size);
        GET_DATA(chunk_index);
        GET_DATA(chunk_size);
        GET_DATA(gain);                         

        // if our netgame options are currently set for no voice, ignore the packet
        if((Netgame.options.flags & MSO_FLAG_NO_VOICE) || !Multi_options_g.std_voice){
                offset += chunk_size;
                return offset;
        }

        // get a handle to a valid stream to be using, freeing old streams as necessary
        stream_index = multi_voice_get_stream((int)stream_id);  

        // if this index is too high, flush the stream
        if(chunk_index >= MULTI_VOICE_ACCUM_BUFFER_COUNT){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : flushing stream because packet index is too high!!\n"));
#endif
                
                // flush the stream
                multi_voice_flush_old_stream(stream_index);

                // return bytes processed
                offset += chunk_size;
                return offset;
        }

        // if we found a stream to work with
        if(stream_index != -1){
                // set the id of where it came from
                Multi_voice_stream[stream_index].stream_from = who_from;                

                // set the stream id#
                Multi_voice_stream[stream_index].stream_id = stream_id;

                // set the gain
                Multi_voice_stream[stream_index].accum_buffer_gain[chunk_index] = (double)gain;                 

                // set the stream uncompressed size size
                Multi_voice_stream[stream_index].accum_buffer_usize[chunk_index] = uc_size;                     

                // set the token timestamp
                Multi_voice_stream[stream_index].token_stamp = timestamp(MULTI_VOICE_TOKEN_TIMEOUT);

                // set the last heard time
                Multi_voice_stream[stream_index].stream_last_heard = timer_get_fixed_seconds();
        
                // copy the data and setup any other accum buffer data necessary
                // ignore data if we can't play sounds
                if(Multi_voice_can_play){
                        memcpy(Multi_voice_stream[stream_index].accum_buffer[chunk_index],data+offset,(int)chunk_size);
                }

                Multi_voice_stream[stream_index].accum_buffer_flags[chunk_index] = 1;
                Multi_voice_stream[stream_index].accum_buffer_csize[chunk_index] = chunk_size;                  
                        
                // pass the data into the smart voice algorithm
                if(player_index != -1){
                        multi_voice_alg_process_data(player_index,stream_index,chunk_index,chunk_size);         
                }
        }

        // increment the offset
        offset += (int)chunk_size;                                      

        return offset;
}

// <server> increment the current stream id#
void multi_voice_inc_stream_id()
{
        Assert(Net_player->flags & NETINFO_FLAG_AM_MASTER);
        
        if(Multi_voice_next_stream_id == 0xff){
                Multi_voice_next_stream_id = 0;
        } else {
                Multi_voice_next_stream_id++;
        }
}

// flush any old sound stream data because we've started to receive data for a new stream
void multi_voice_flush_old_stream(int stream_index)
{               
#ifdef MULTI_VOICE_VERBOSE
        nprintf(("Network","MULTI VOICE : old stream flush\n"));                
#endif

        // call the smart algorithm for flushing streams
        multi_voice_alg_flush_old_stream(stream_index);
        
        // clear all the accum buffer flags
        memset(Multi_voice_stream[stream_index].accum_buffer_flags,0,MULTI_VOICE_ACCUM_BUFFER_COUNT);

        // clear the token 
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                multi_voice_take_token(stream_index);
        }

        Multi_voice_stream[stream_index].token_stamp = -1;
        Multi_voice_stream[stream_index].token_status = MULTI_VOICE_TOKEN_INDEX_FREE;

        // timestamp the player
        Net_player->s_info.voice_token_timestamp = timestamp(Netgame.options.voice_token_wait);
}

// route sound data through the server to all appropriate players
void multi_voice_route_data(ubyte *data, int packet_size,int player_index,int mode,net_player *target)
{
        int idx;

        // route the data to all other players
        switch(mode){
        case MULTI_MSG_ALL:
                for(idx=0;idx<MAX_PLAYERS;idx++){
                        if(MULTI_CONNECTED( Net_players[idx] ) &&                                                                                                       // player is connected
                          ( &Net_players[idx] != &Net_players[player_index] ) &&                                                                // not the sending player
                          ( Net_player != &Net_players[idx] ) &&                                                                                                        // not me
                          ( Multi_voice_player_prefs[idx] & (1 << player_index) ) &&                                            // is accepting sound from this player
                          !( Net_players[idx].p_info.options.flags & MLO_FLAG_NO_VOICE ) ){                             // is accepting sound periods
                                                        
                                multi_io_send(&Net_players[idx], data, packet_size);
                        }
                }
                break;
        
        case MULTI_MSG_FRIENDLY:
                for(idx=0;idx<MAX_PLAYERS;idx++){
                        if(MULTI_CONNECTED( Net_players[idx] ) &&                                                                                                       // player is connected
                          ( &Net_players[idx] != &Net_players[player_index] ) &&                                                                // not the sending player
                          ( Net_player != &Net_players[idx] ) &&                                                                                                        // not me
                          ( Net_players[idx].p_info.team == Net_players[player_index].p_info.team ) &&// on the same team
                          ( Multi_voice_player_prefs[idx] & (1 << player_index) ) &&                                            // is accepting sound from the sender
                          !( Net_players[idx].p_info.options.flags & MLO_FLAG_NO_VOICE) ){                              // is accepting sound periods
                                                
                                multi_io_send(&Net_players[idx], data, packet_size);
                        }
                }
                break;
        case MULTI_MSG_HOSTILE:
                for(idx=0;idx<MAX_PLAYERS;idx++){
                        if(MULTI_CONNECTED( Net_players[idx] ) &&                                                                                                       // player is connected
                          ( &Net_players[idx] != &Net_players[player_index] ) &&                                                                // not the sending player       
                          ( Net_player != &Net_players[idx] ) &&                                                                                                        // not me
                          ( Net_players[idx].p_info.team != Net_players[player_index].p_info.team ) &&// on the opposite team
                          ( Multi_voice_player_prefs[idx] & (1 << player_index) ) &&                                            // is accepting sound from the sender
                          !( Net_players[idx].p_info.options.flags & MLO_FLAG_NO_VOICE ) ){                             // is accepting sound periods
                                                        
                                multi_io_send(&Net_players[idx], data, packet_size);
                        }
                }
                break;
        
        case MULTI_MSG_TARGET:
                Assert(target != NULL);
                if(!(target->p_info.options.flags & MLO_FLAG_NO_VOICE)){                                        
                        multi_io_send(target, data, packet_size);
                }
                break;
        }
}

// find the stream to apply incoming sound data to, freeing up old ones as necessary
int multi_voice_get_stream(int stream_id)
{
        int idx,max_diff_index;
        fix cur_time,max_diff;

        // first check to see if this stream exists
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(Multi_voice_stream[idx].stream_id == (ubyte)stream_id){
                        return idx;
                }
        }

        // if we got to this point, we didn't find the matching stream, so we should try and find an empty stream
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(Multi_voice_stream[idx].token_stamp == -1){
                        return idx;
                }
        }

#ifdef MULTI_VOICE_VERBOSE
        nprintf(("Network","MULTI VOICE : going to blast old voice stream while looking for a free one - beware!!\n"));
#endif

        // if we got to this point, we should free up the oldest stream we have
        cur_time = timer_get_fixed_seconds();
        max_diff_index = -1;
        max_diff = -1;
        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                if(((max_diff_index == -1) || ((cur_time - Multi_voice_stream[idx].stream_last_heard) > max_diff)) && (Multi_voice_stream[idx].token_stamp != -1)){
                        max_diff_index = idx;
                        max_diff = cur_time - Multi_voice_stream[idx].stream_last_heard;                        
                }
        }

        // if we found the oldest 
        if(max_diff_index != -1){
                // flush the old stream
                multi_voice_flush_old_stream(max_diff_index);           
                
                return max_diff_index;
        }

        // some other fail condition
        return -1;
}

// is the given sound stream playing (compares uncompressed sound size with current playback position)
int multi_voice_stream_playing(int stream_index)
{
        // if the handle is invalid, it can't be playing
        /*
        if(Multi_voice_stream[stream_index].stream_snd_handle < 0){
                return 0;
        }

        // if the sound is playing and the buffer is past the uncompressed size, its effectively done   
        if(ds_get_play_position(ds_get_channel(Multi_voice_stream[stream_index].stream_snd_handle)) >= (DWORD)Multi_voice_stream[stream_index].stream_uc_size){
                return 1;
        }
        */

        // not done yet
        return 0;
}

// tack on pre and post sounds to a sound stream (pass -1 for either if no sound is wanted)
// return final buffer size
int multi_voice_mix(int post_sound,char *data,int cur_size,int max_size)
{
        int post_size;
        
        // if the user passed -1 for both pre and post sounds, don't do a thing
        if(post_sound == -1){
                return cur_size;
        }

        // get the sizes of the additional sounds
        
        // post sound
        if(post_sound >= 0){
                post_sound = snd_load(&Snds[post_sound]);
                if(post_sound >= 0){
                        if(snd_size(post_sound,&post_size) == -1){
                                post_size = 0;
                        }
                } else {
                        post_size = 0;
                }
        } else {
                post_size = 0;
        }
                        
        // if we have a "post" sound to add
        if(post_size > 0){
                if((max_size - cur_size) > post_size){
                        // copy in the sound
                        snd_get_data(post_sound,data + cur_size);

                        // increment the cur_size
                        cur_size += post_size;
                }
        }

        // return the size of the new buffer
        return cur_size;
}

// max size of a sound chunk which we can fit into a packet
int multi_voice_max_chunk_size(int msg_mode)
{
        int header_size;

        // all headers contain the following data
        header_size =   1 +                                                                     // messaging mode
                                                1 +                                                                     // stream id #
                                                2 +                                                                     // packet uncompressed size
                                                2 +                                                                     // compressed size
                                                4;                                                                              // gain 

        // if we're targeting a specific player
        if(msg_mode == MULTI_MSG_TARGET){
                header_size += 2;                                                                       // targeted player's object net_signature
        }
        
        // if we're in IPX mode
        if(Psnet_my_addr.type == NET_IPX){
                header_size += 10;                                                              // my address (10 bytes in IPX)         
        }
        // if we're in TCP mode
        else {
                header_size +=  4;                                                                      // my address (4 bytes in TCP)
        }

        // calculate max chunk size
        return (MAX_PACKET_SIZE -                                                       // max freespace packet size
                          1                                     -                                                       // packet type 
                          1                                     -                                                       // voice packet code subtype
                          header_size);                                                         // calculated header size
}

// --------------------------------------------------------------------------------------------------
// MULTI VOICE / RTVOICE INTERFACE
//

// process the "next" chunk of standalone valid sound data from the rtvoice system
void multi_voice_process_next_chunk()
{                       
        int sound_size,uncompressed_size;       
        float gain;
        double d_gain;
        voice_stream *str;

        // we'd better not ever get here is we can't record voice
        Assert(Multi_voice_can_record);

        // get the data 
        rtvoice_get_data((unsigned char**)&Multi_voice_record_buffer,&sound_size,&uncompressed_size,&d_gain);           
        gain = (float)d_gain;

        // if we've reached the max # of packets for this stream, bail
        if(Multi_voice_current_stream_index >= (MULTI_VOICE_ACCUM_BUFFER_COUNT - 1)){
                nprintf(("Network","MULTI VOICE : Forcing stream to stop on the record size!!!\n"));

                // mark me as no longer recording
                Multi_voice_recording = 0;                      

                Multi_voice_current_stream_sent = -1;
                                
                // stop the recording process
                rtvoice_stop_recording();                               
                                
#ifdef MULTI_VOICE_POST_DECOMPRESS
                multi_voice_player_send_stream();
#endif

                // play my sound locally as well
#ifdef MULTI_VOICE_LOCAL_ECHO   
                multi_voice_alg_play_window(0);
#endif
                // release the token back to the server
                multi_voice_release_token();

                // unset the timestamp so we don't still think we're still recording
                Multi_voice_recording_stamp = -1;

                return;
        }

        // pack the data locally as well (so I can hear myself)
        str = &Multi_voice_stream[0];
        memcpy(str->accum_buffer[Multi_voice_current_stream_index],Multi_voice_record_buffer,sound_size);
        str->stream_from = Net_player->player_id;       
        str->accum_buffer_flags[Multi_voice_current_stream_index] = 1;
        str->accum_buffer_usize[Multi_voice_current_stream_index] = (ushort)uncompressed_size;
        str->accum_buffer_csize[Multi_voice_current_stream_index] = (ushort)sound_size;
        str->accum_buffer_gain[Multi_voice_current_stream_index] = d_gain;      

        // increment the stream index
        Multi_voice_current_stream_index++;
}


// --------------------------------------------------------------------------------------------------
// MULTI VOICE PACKET HANDLERS
//

// send a dummy packet in the place of a too-large data packet
void multi_voice_send_dummy_packet()
{
        ubyte data[10],code,msg_mode;
        int packet_size,target_index;   

        // build the header and add the opcode
        BUILD_HEADER(VOICE_PACKET);
        code = (ubyte)MV_CODE_DATA_DUMMY;
        ADD_DATA(code);

        msg_mode = (ubyte)Multi_voice_send_mode;
        // get the specific target if we're in MSG_TARGET mode
        target_index = -1;
        if(msg_mode == MULTI_MSG_TARGET){
                if(Player_ai->target_objnum != -1){
                        target_index = multi_find_player_by_object(&Objects[Player_ai->target_objnum]);
                        if(target_index == -1){
                                return;
                        }
                } else {
                        return;
                }
        }
        ADD_DATA(msg_mode);
        if(msg_mode == MULTI_MSG_TARGET){
                ADD_DATA(Objects[Net_players[target_index].player->objnum].net_signature);
        }

        // add the voice stream id
        ADD_DATA(Multi_voice_stream_id);

        // send to the server or rebroadcast if I _am_ the server
        if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                multi_voice_route_data(data,packet_size,MY_NET_PLAYER_NUM,(int)msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);
        } else {                
                multi_io_send(Net_player, data, packet_size);
        }       
}

// process a dummy data packet
int multi_voice_process_data_dummy(ubyte *data)
{
        int offset = 0;
        int stream_index;
        ubyte stream_id;

        // get the stream id
        GET_DATA(stream_id);

        // get the proper stream index
        stream_index = multi_voice_get_stream((int)stream_id);

        // set the token timestamp
        Multi_voice_stream[stream_index].token_stamp = timestamp(MULTI_VOICE_TOKEN_TIMEOUT);

        // set the last heard time
        Multi_voice_stream[stream_index].stream_last_heard = timer_get_fixed_seconds();

        // set the timeout timestamp
        Multi_voice_stamps[stream_index] = timestamp(MV_ALG_TIMEOUT);   

        // return bytes processed
        return offset;
}

// process a player preferences packet, return bytes processed
int multi_voice_process_player_prefs(ubyte *data,int player_index)
{
        ubyte val;
        int mute_index;
        short mute_id;
        int offset = 0;

        // set all channels active
        Multi_voice_player_prefs[player_index] = 0xffffffff;

        // get all muted players
        GET_DATA(val);
        while(val != 0xff){
                GET_DATA(mute_id);

                // get the player to mute
                mute_index = find_player_id(mute_id);
                if(mute_index != -1){
#ifdef MULTI_VOICE_VERBOSE
                        nprintf(("Network","Player %s muting player %s\n",Net_players[player_index].player->callsign,Net_players[mute_index].player->callsign));
#endif
                        // mute the guy
                        Multi_voice_player_prefs[player_index] &= ~(1<<mute_index);
                }

                // get the next stop value
                GET_DATA(val);
        }

        // return bytes processed
        return offset;
}

// process an incoming voice packet of some kind or another
void multi_voice_process_packet(ubyte *data, header *hinfo)
{
        ubyte code,msg_mode;
        ushort target_sig;      
        int player_index,stream_index,target_index;     
        int offset = HEADER_LENGTH;     

        // find out who is sending this data    
        player_index = find_player_id(hinfo->id);               

        // get the opcode
        GET_DATA(code);

        // process the packet
        switch(code){
        // I don't have the token anymore
        case MV_CODE_TAKE_TOKEN:
                // we should never have the token if we cannot record
                if(!Multi_voice_can_record){
                        Int3();
                }

                Multi_voice_token = 0;
                break;

        // I have been denied the token
        case MV_CODE_DENY_TOKEN:
                // set the "denied" timestamp
                Multi_voice_denied_stamp = timestamp(MULTI_VOICE_DENIED_TIME);
                break;
        
        // I now have the token
        case MV_CODE_GIVE_TOKEN:                
                GET_DATA(Multi_voice_stream_id);

                // we should never get the token if we cannot record
                if(!Multi_voice_can_record){
                        Int3();
                }

                // if we no longer have the keydown, automatically release the token
                if(!Multi_voice_keydown){
                        multi_voice_release_token();
                } else {
                        Multi_voice_token = 1;
                }
                break;

        // a request for the token from a player
        case MV_CODE_REQUEST_TOKEN:
                if(player_index >= 0){
                        multi_voice_process_token_request(player_index);
                }
                break;
        
        // a player gave up the token
        case MV_CODE_RELEASE_TOKEN:             
                if(player_index >= 0){
                        stream_index = multi_voice_find_token(player_index);
                } else {
                        break;
                }
                
                if(stream_index >= 0){
                        // set the token as having been released                
                        Multi_voice_stream[stream_index].token_status = MULTI_VOICE_TOKEN_INDEX_RELEASED;               

                        // timestamp this guy so that he can't get the token back immediately
                        Net_players[player_index].s_info.voice_token_timestamp = timestamp(Netgame.options.voice_token_wait);
                } 
                break;

        // a player has set prefs for himself
        case MV_CODE_PLAYER_PREFS:
                Assert(player_index != -1);             
                offset += multi_voice_process_player_prefs(data+offset,player_index);
                break;

        // a data packet
        case MV_CODE_DATA:
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","VOICE : PROC DATA\n"));
#endif
                // get routing information
                target_index = -1;
                GET_DATA(msg_mode);
                if(msg_mode == MULTI_MSG_TARGET){
                        GET_DATA(target_sig);
                        target_index = multi_find_player_by_net_signature(target_sig);
                        Assert(target_index != -1);
                }

                offset += multi_voice_process_data(data+offset,player_index,msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);

                // if we're the server of the game, we should also route this data to all other players
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                        multi_voice_route_data(data,offset,player_index,msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);
                }
                break;  

        // a data dummy packet
        case MV_CODE_DATA_DUMMY:
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","VOICE : PROC DATA DUMMY\n"));
#endif
                // get routing information
                target_index = -1;
                GET_DATA(msg_mode);
                if(msg_mode == MULTI_MSG_TARGET){
                        GET_DATA(target_sig);
                        target_index = multi_find_player_by_net_signature(target_sig);
                        Assert(target_index != -1);
                }

                offset += multi_voice_process_data_dummy(data+offset);

                // if we're the server of the game, we should also route this data to all other players
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){
                        multi_voice_route_data(data,offset,player_index,msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);
                }
                break;
        }
        PACKET_SET_SIZE();
}

// send all pending voice packets
void multi_voice_client_send_pending()
{
        ubyte data[MAX_PACKET_SIZE],code;
        ubyte msg_mode,chunk_index;
        ushort uc_size,chunk_size;
        int max_chunk_size,sent,target_index;
        int packet_size;
        float gain;
        voice_stream *str;
        
        // if we're not recording
        if(!Multi_voice_recording || (Multi_voice_current_stream_sent < 0) || (Multi_voice_current_stream_sent > Multi_voice_current_stream_index)){
                return;
        }

        // stream all buffered up packets
        str = &Multi_voice_stream[0];
        while(Multi_voice_current_stream_sent < Multi_voice_current_stream_index){              
                sent = Multi_voice_current_stream_sent++;

                // get the current messaging mode
                msg_mode = (ubyte)Multi_voice_send_mode;                

                // if the size of this voice chunk will fit in the packet
                max_chunk_size = multi_voice_max_chunk_size(Multi_voice_send_mode);
                if(str->accum_buffer_csize[sent] > max_chunk_size){
#ifdef MULTI_VOICE_VERBOSE
                        nprintf(("Network","MULTI VOICE : streamed packet size too large!!\n"));
#endif

                        Multi_voice_current_stream_sent++;

                        // send a dummy data packet instead
                        multi_voice_send_dummy_packet();
                        
                        continue;
                }

#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : PACKET %d %d\n",(int)str->accum_buffer_csize[sent],(int)str->accum_buffer_usize[sent]));
#endif
        
                // get the specific target if we're in MSG_TARGET mode
                target_index = -1;
                if(msg_mode == MULTI_MSG_TARGET){
                        if(Player_ai->target_objnum != -1){
                                target_index = multi_find_player_by_object(&Objects[Player_ai->target_objnum]);
                                if(target_index == -1){
                                        return;
                                }
                        } else {
                                return;
                        }
                }

                // go through the data and send all of it
                code = MV_CODE_DATA;
                chunk_index = 0;                
        
                // if this packet is small enough to fit within a psnet data packet             
                BUILD_HEADER(VOICE_PACKET);

                // add the packet code type
                ADD_DATA(code);

                // add the routing data and any necessary targeting information
                ADD_DATA(msg_mode);
                if(msg_mode == MULTI_MSG_TARGET){
                        Assert(Game_mode & GM_IN_MISSION);
                        ADD_DATA(Objects[Net_players[target_index].player->objnum].net_signature);
                }

                // add my address 
                ADD_DATA(Net_player->player_id);

                // add the current stream id#
                ADD_DATA(Multi_voice_stream_id);

                Assert(str->accum_buffer_usize[sent] < MULTI_VOICE_MAX_BUFFER_SIZE);
                uc_size = (ushort)str->accum_buffer_usize[sent];
                ADD_DATA(uc_size);

                // add the chunk index
                chunk_index = (ubyte)sent;
                ADD_DATA(chunk_index);

                // size of the sound data
                chunk_size = (ushort)str->accum_buffer_csize[sent];             
                ADD_DATA(chunk_size);

                // add the gain
                gain = (float)str->accum_buffer_gain[sent];
                ADD_DATA(gain);

                // add the chunk of data                
                memcpy(data+packet_size, str->accum_buffer[sent],chunk_size);           
                packet_size += chunk_size;

                // send to the server or rebroadcast if I _am_ the server
                if(Net_player->flags & NETINFO_FLAG_AM_MASTER){         
                        multi_voice_route_data(data,packet_size,MY_NET_PLAYER_NUM,(int)msg_mode,(target_index == -1) ? NULL : &Net_players[target_index]);
                } else {                        
                        multi_io_send(Net_player, data, packet_size);
                }       
        }
}


// --------------------------------------------------------------------------------------------------
// MULTI VOICE ALGORITHM stuff
//

// process and play the current window of sound stream data we have. reset the window for the next incoming data as well
void multi_voice_alg_play_window(int stream_index)
{
        int idx,buffer_offset;  
        voice_stream *st;

#ifdef MULTI_VOICE_VERBOSE
        nprintf(("Network","MULTI VOICE : PLAYING STREAM %d\n",stream_index));
#endif

        // get a pointer to the stream
        st = &Multi_voice_stream[stream_index];

        // don't play anything back if we can't hear sound
        if(Multi_voice_can_play){       
                // first, pack all the accum buffers into the playback buffer
#ifdef MULTI_VOICE_PRE_DECOMPRESS
                buffer_offset = Multi_voice_pre_sound_size;
                nprintf(("Network","VOICE : pre sound size %d\n",Multi_voice_pre_sound_size));
                for(idx=0;idx<MULTI_VOICE_ACCUM_BUFFER_COUNT;idx++){
                        // if the flag is set, uncompress the data into the playback buffer
                        if(st->accum_buffer_flags[idx]){
                                // first, uncompress the data
                                rtvoice_uncompress(st->accum_buffer[idx],(int)st->accum_buffer_csize[idx],st->accum_buffer_gain[idx],(ubyte*)Multi_voice_playback_buffer+buffer_offset,st->accum_buffer_usize[idx]);
                        
                                // increment the buffer offset
                                buffer_offset += st->accum_buffer_usize[idx];
                        }
                }                               
#endif
#ifdef MULTI_VOICE_POST_DECOMPRESS
                buffer_offset = 0;
                for(idx=0;idx<MULTI_VOICE_ACCUM_BUFFER_COUNT;idx++){
                        // if the flag is set, copy the data
                        if(st->accum_buffer_flags[idx]){
                                memcpy(Multi_voice_unpack_buffer+buffer_offset,st->accum_buffer[idx],st->accum_buffer_csize[idx]);
                                buffer_offset += st->accum_buffer_csize[idx];
                        }
                }       

                // decompress the whole shebang
                rtvoice_uncompress((ubyte*)Multi_voice_unpack_buffer,buffer_offset,st->accum_buffer_gain[0],(ubyte*)Multi_voice_playback_buffer,st->accum_buffer_usize[0]);
                buffer_offset = st->accum_buffer_usize[0];
#endif          

                // mix in the SND_CUE_VOICE and the SND_END_VOICE game sounds
                buffer_offset = multi_voice_mix(MULTI_VOICE_POST_SOUND,Multi_voice_playback_buffer,buffer_offset,MULTI_VOICE_MAX_BUFFER_SIZE);
                        
                Assert(Multi_voice_stream[stream_index].stream_rtvoice_handle != -1);

                // kill any previously playing sounds
                rtvoice_stop_playback(Multi_voice_stream[stream_index].stream_rtvoice_handle);  
                Multi_voice_stream[stream_index].stream_snd_handle = -1;

                // if we can play sound and we know who this is from, display it
                if(Multi_voice_can_play){
                        char voice_msg[256];
                        int player_index = find_player_id(Multi_voice_stream[stream_index].stream_from);

                        if(player_index != -1){
                                memset(voice_msg,0,256);
                                sprintf(voice_msg,XSTR("<%s is speaking>",712),Net_players[player_index].player->callsign);

                                // display a chat message (write to the correct spot - hud, standalone gui, chatbox, etc)
                                multi_display_chat_msg(voice_msg,player_index,0);
                        }
                }
        
                // now call the rtvoice playback functions              
                Multi_voice_stream[stream_index].stream_snd_handle = rtvoice_play_uncompressed(Multi_voice_stream[stream_index].stream_rtvoice_handle,(unsigned char*)Multi_voice_playback_buffer,buffer_offset);       
                Multi_voice_stream[stream_index].stream_start_time = timer_get_fixed_seconds();
        }
        
        // unset the stamp so that its not "free"
        Multi_voice_stamps[stream_index] = -1;

        // flush the stream (will also grab the token back, if the server)
        multi_voice_flush_old_stream(stream_index);
}

// decision function which decides if we should play the current block of sound we have
int multi_voice_alg_should_play(int stream_index)
{
        // if the timestamp has expired, play the sound
        if((Multi_voice_stamps[stream_index] != -1) && timestamp_elapsed(Multi_voice_stamps[stream_index])){
#ifdef MULTI_VOICE_VERBOSE
                nprintf(("Network","MULTI VOICE : DECIDE, TIMEOUT\n"));         
#endif
                return 1;
        }
                        
        return 0;
}

// process incoming sound data in whatever way necessary (this function should take care of playing data when necessary)
void multi_voice_alg_process_data(int player_index,int stream_index,ushort chunk_index,ushort chunk_size)
{
        // do this so we don't get compiler warnings
        chunk_index = 0;
        chunk_size = 0;
        player_index = 0;

        // update the timestamp for this window
        Multi_voice_stamps[stream_index] = timestamp(MV_ALG_TIMEOUT);   
}

// process existing streams
void multi_voice_alg_process_streams()
{
        int idx;
        int player_index;

        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){                   
                // determine if we should play this window of data
                if((Multi_voice_stamps[idx] != -1) && multi_voice_alg_should_play(idx)){
                        // determine who this stream came from
                        player_index = find_player_id(Multi_voice_stream[idx].stream_from);                     

                        // server should check his own settings here
                        if((Net_player->flags & NETINFO_FLAG_AM_MASTER) && ((Net_player->p_info.options.flags & MLO_FLAG_NO_VOICE) || (player_index == -1) || !(Multi_voice_player_prefs[MY_NET_PLAYER_NUM] & (1<<player_index))) ){
                                // unset the stamp so that its not "free"
                                Multi_voice_stamps[idx] = -1;

                                // flush the stream (will also grab the token back, if the server)
                                multi_voice_flush_old_stream(idx);

                                nprintf(("Network","Server not playing sound because of set options!\n"));
                        }
                        // play the current sound
                        else {                          
                                multi_voice_alg_play_window(idx);                       
                        }
                }
        }
}

// we are going to flush the current stream because we have started to receive data for a new one. do something first
void multi_voice_alg_flush_old_stream(int stream_index)
{
        // just unset the heard from timestamp for now
        Multi_voice_stamps[stream_index] = -1;
}

// initialize the smart algorithm
void multi_voice_alg_init()
{
        int idx;

        for(idx=0;idx<MULTI_VOICE_MAX_STREAMS;idx++){
                Multi_voice_stamps[idx] = -1;
        }
}


// --------------------------------------------------------------------------------------------------
// MULTI VOICE TESTING FUNCTIONS
//

#define MV_TEST_RECORD_TIME                             3000                                    // recording time in ms for testing voice
int Multi_voice_test_record_stamp = -1;
int Multi_voice_test_packet_tossed = 0;

// process the next chunk of voice data
void multi_voice_test_process_next_chunk()
{
        unsigned char *outbuf,*outbuf_raw;
        int compressed_size,uncompressed_size,raw_size;
        double gain;
        
        // if the test recording stamp is -1, we should stop
        if(Multi_voice_test_record_stamp == -1){
                rtvoice_stop_recording();
                return;
        }

        // if the recording timestamp has elapsed, stop the whole thing
        if(timestamp_elapsed(Multi_voice_test_record_stamp)){
                nprintf(("Network","Stopping voice test recording\n"));

                rtvoice_stop_recording();

                Multi_voice_test_record_stamp = -1;
                Multi_voice_test_packet_tossed = 0;
                return;
        }

        // otherwise get the compressed and uncompressed data and do something interesting with it
        rtvoice_get_data(&outbuf,&compressed_size,&uncompressed_size,&gain,&outbuf_raw,&raw_size);      

        // determine whether the packet would have been dropped
        if(compressed_size > multi_voice_max_chunk_size(MULTI_MSG_ALL)){
                Multi_voice_test_packet_tossed = 1;
        } else {
                Multi_voice_test_packet_tossed = 0;
        }

        // send the raw output buffer to the voice options screen
        options_multi_set_voice_data(outbuf_raw,raw_size,outbuf,compressed_size,uncompressed_size,gain);
}

// start recording voice locally for playback testing
void multi_voice_test_record_start()
{
        // if there is test recording going on already, don't do anything
        if(Multi_voice_test_record_stamp != -1){
                return;
        }

        // stop any playback which may be occuring
        rtvoice_stop_playback_all();

        // stop any recording which may be occuring
        rtvoice_stop_recording();

        // set the timestamp
        Multi_voice_test_record_stamp = timestamp(MV_TEST_RECORD_TIME);

        // start the recording of voice
        rtvoice_start_recording(multi_voice_test_process_next_chunk);
}

// force stop any recording voice test
void multi_voice_test_record_stop()
{
        Multi_voice_test_record_stamp = -1;
        Multi_voice_test_packet_tossed = 0;
        rtvoice_stop_recording();
}

// return if the test recording is going on
int multi_voice_test_recording()
{
        return (Multi_voice_test_record_stamp == -1) ? 0 : 1;
}

// call this function if multi_voice_test_recording() is true to process various odds and ends of the test recording
void multi_voice_test_process()
{
        // if we're not recording, do nothing
        if(Multi_voice_test_record_stamp == -1){
                return;
        }

        // check to see if the timestamp has elapsed
        if(timestamp_elapsed(Multi_voice_test_record_stamp)){
                Multi_voice_test_record_stamp = -1;
                Multi_voice_test_packet_tossed = 0;
        }
}

// get a playback buffer handle (return -1 if none exist - bad)
int multi_voice_test_get_playback_buffer()
{
        // return voice stream 0
        Assert(Multi_voice_stream[0].stream_snd_handle == -1);
        Assert(Multi_voice_stream[0].stream_rtvoice_handle != -1);

        return Multi_voice_stream[0].stream_rtvoice_handle;
}

// return whether the last sampled chunk would have been too large to fit in a packet
int multi_voice_test_packet_tossed()
{
        return Multi_voice_test_packet_tossed;
}