live555学习笔记8

八 RTSPClient分析

有RTSPServer，当然就要有RTSPClient。
如果按照Server端的架构，想一下Client端各部分的组成可能是这样：
因为要连接RTSP server，所以RTSPClient要有TCP socket。当获取到server端的DESCRIBE后，应建立一个对应于ServerMediaSession的ClientMediaSession。对应每个Track，ClientMediaSession中应建立ClientMediaSubsession。当建立RTP Session时，应分别为所拥有的Track发送SETUP请求连接，在获取回应后，分别为所有的track建立RTP socket，然后请求PLAY，然后开始传输数据。事实是这样吗？只能分析代码了。

testProgs中的OpenRTSP是典型的RTSPClient示例，所以分析它吧。
main()函数在playCommon.cpp文件中。main()的流程比较简单，跟服务端差别不大：建立任务计划对象－－建立环境对象－－处理用户输入的参数(RTSP地址)－－创建RTSPClient实例－－发出第一个RTSP请求（可能是OPTIONS也可能是DESCRIBE）－－进入Loop。

RTSP的tcp连接是在发送第一个RTSP请求时才建立的，在RTSPClient的那几个发请求的函数sendXXXXXXCommand()中最终都调用sendRequest()，sendRequest()中会跟据情况建立起TCP连接。在建立连接时马上向任务计划中加入处理从这个TCP接收数据的socket handler：RTSPClient::incomingDataHandler()。
下面就是发送RTSP请求，OPTIONS就不必看了，从请求DESCRIBE开始：

void getSDPDescription(RTSPClient::responseHandler* afterFunc) { ourRTSPClient->sendDescribeCommand(afterFunc, ourAuthenticator); } unsigned RTSPClient::sendDescribeCommand(responseHandler* responseHandler, Authenticator* authenticator) { if (authenticator != NULL) fCurrentAuthenticator = *authenticator; return sendRequest(new RequestRecord(++fCSeq, "DESCRIBE", responseHandler)); }参数responseHandler是调用者提供的回调函数，用于在处理完请求的回应后再调用之。并且在这个回调函数中会发出下一个请求－－所有的请求都是这样依次发出的。使用回调函数的原因主要是因为socket的发送与接收不是同步进行的。类RequestRecord就代表一个请求,它不但保存了RTSP请求相关的信息，而且保存了请求完成后的回调函数－－就是responseHandler。有些请求发出时还没建立tcp连接，不能立即发送，则加入fRequestsAwaitingConnection队列；有些发出后要等待Server端的回应，就加入fRequestsAwaitingResponse队列，当收到回应后再从队列中把它取出。
由于RTSPClient::sendRequest()太复杂，就不列其代码了，其无非是建立起RTSP请求字符串然后用TCP socket发送之。

现在看一下收到DESCRIBE的回应后如何处理它。理论上是跟据媒体信息建立起MediaSession了，看看是不是这样：

void continueAfterDESCRIBE(RTSPClient*, int resultCode, char* resultString) { char* sdpDescription = resultString; //跟据SDP创建MediaSession。 // Create a media session object from this SDP description: session = MediaSession::createNew(*env, sdpDescription); delete[] sdpDescription; // Then, setup the "RTPSource"s for the session: MediaSubsessionIterator iter(*session); MediaSubsession *subsession; Boolean madeProgress = False; char const* singleMediumToTest = singleMedium; //循环所有的MediaSubsession，为每个设置其RTPSource的参数 while ((subsession = iter.next()) != NULL) { //初始化subsession，在其中会建立RTP/RTCP socket以及RTPSource。 if (subsession->initiate(simpleRTPoffsetArg)) { madeProgress = True; if (subsession->rtpSource() != NULL) { // Because we're saving the incoming data, rather than playing // it in real time, allow an especially large time threshold // (1 second) for reordering misordered incoming packets: unsigned const thresh = 1000000; // 1 second subsession->rtpSource()->setPacketReorderingThresholdTime(thresh); // Set the RTP source's OS socket buffer size as appropriate - either if we were explicitly asked (using -B), // or if the desired FileSink buffer size happens to be larger than the current OS socket buffer size. // (The latter case is a heuristic, on the assumption that if the user asked for a large FileSink buffer size, // then the input data rate may be large enough to justify increasing the OS socket buffer size also.) int socketNum = subsession->rtpSource()->RTPgs()->socketNum(); unsigned curBufferSize = getReceiveBufferSize(*env,socketNum); if (socketInputBufferSize > 0　|| fileSinkBufferSize > curBufferSize) { unsigned newBufferSize =　socketInputBufferSize > 0 ?　 socketInputBufferSize : fileSinkBufferSize; newBufferSize = setReceiveBufferTo(*env, socketNum, newBufferSize); if (socketInputBufferSize > 0) { // The user explicitly asked for the new socket buffer size; announce it: *env << "Changed socket receive buffer size for the \"" << subsession->mediumName() << "/" << subsession->codecName() << "\" subsession from " << curBufferSize << " to " << newBufferSize << " bytes\n"; } } } } } if (!madeProgress) shutdown(); // Perform additional 'setup' on each subsession, before playing them: //下一步就是发送SETUP请求了。需要为每个Track分别发送一次。 setupStreams(); }此函数被删掉很多枝叶，所以发现与原版不同请不要惊掉大牙。
的确在DESCRIBE回应后建立起了MediaSession，而且我们发现Client端的MediaSession不叫ClientMediaSesson，SubSession亦不是。我现在很想看看MediaSession与MediaSubsession的建立过程:

MediaSession* MediaSession::createNew(UsageEnvironment& env,char const* sdpDescription) { MediaSession* newSession = new MediaSession(env); if (newSession != NULL) { if (!newSession->initializeWithSDP(sdpDescription)) { delete newSession; return NULL; } } return newSession; }
我可以告诉你，MediaSession的构造函数没什么可看的，那么就来看initializeWithSDP()：
内容太多，不必看了，我大体说说吧：就是处理SDP，跟据每一行来初始化一些变量。当遇到"m="行时，就建立一个MediaSubsession，然后再处理这一行之下，下一个"m="行之上的行们，用这些参数初始化MediaSubsession的变量。循环往复，直到尽头。然而这其中并没有建立RTP socket。我们发现在continueAfterDESCRIBE()中，创建MediaSession之后又调用了subsession->initiate(simpleRTPoffsetArg)，那么socket是不是在它里面创建的呢？look:

Boolean MediaSubsession::initiate(int useSpecialRTPoffset) { if (fReadSource != NULL) return True; // has already been initiated do { if (fCodecName == NULL) { env().setResultMsg("Codec is unspecified"); break; } //创建RTP/RTCP sockets // Create RTP and RTCP 'Groupsocks' on which to receive incoming data. // (Groupsocks will work even for unicast addresses) struct in_addr tempAddr; tempAddr.s_addr = connectionEndpointAddress(); // This could get changed later, as a result of a RTSP "SETUP" if (fClientPortNum != 0) { //当server端指定了建议的client端口 // The sockets' port numbers were specified for us. Use these: fClientPortNum = fClientPortNum & ~1; // even if (isSSM()) { fRTPSocket = new Groupsock(env(), tempAddr, fSourceFilterAddr, fClientPortNum); } else { fRTPSocket = new Groupsock(env(), tempAddr, fClientPortNum, 255); } if (fRTPSocket == NULL) { env().setResultMsg("Failed to create RTP socket"); break; } // Set our RTCP port to be the RTP port +1 portNumBits const rtcpPortNum = fClientPortNum | 1; if (isSSM()) { fRTCPSocket = new Groupsock(env(), tempAddr, fSourceFilterAddr, rtcpPortNum); } else { fRTCPSocket = new Groupsock(env(), tempAddr, rtcpPortNum, 255); } if (fRTCPSocket == NULL) { char tmpBuf[100]; sprintf(tmpBuf, "Failed to create RTCP socket (port %d)", rtcpPortNum); env().setResultMsg(tmpBuf); break; } } else { //Server端没有指定client端口，我们自己找一个。之所以做的这样复杂，是为了能找到连续的两个端口 //RTP/RTCP的端口号不是要连续吗？还记得不？ // Port numbers were not specified in advance, so we use ephemeral port numbers. // Create sockets until we get a port-number pair (even: RTP; even+1: RTCP). // We need to make sure that we don't keep trying to use the same bad port numbers over and over again. // so we store bad sockets in a table, and delete them all when we're done. HashTable* socketHashTable = HashTable::create(ONE_WORD_HASH_KEYS); if (socketHashTable == NULL) break; Boolean success = False; NoReuse dummy; // ensures that our new ephemeral port number won't be one that's already in use while (1) { // Create a new socket: if (isSSM()) { fRTPSocket = new Groupsock(env(), tempAddr, fSourceFilterAddr, 0); } else { fRTPSocket = new Groupsock(env(), tempAddr, 0, 255); } if (fRTPSocket == NULL) { env().setResultMsg( "MediaSession::initiate(): unable to create RTP and RTCP sockets"); break; } // Get the client port number, and check whether it's even (for RTP): Port clientPort(0); if (!getSourcePort(env(), fRTPSocket->socketNum(), clientPort)) { break; } fClientPortNum = ntohs(clientPort.num()); if ((fClientPortNum & 1) != 0) { // it's odd // Record this socket in our table, and keep trying: unsigned key = (unsigned) fClientPortNum; Groupsock* existing = (Groupsock*) socketHashTable->Add( (char const*) key, fRTPSocket); delete existing; // in case it wasn't NULL continue; } // Make sure we can use the next (i.e., odd) port number, for RTCP: portNumBits rtcpPortNum = fClientPortNum | 1; if (isSSM()) { fRTCPSocket = new Groupsock(env(), tempAddr, fSourceFilterAddr, rtcpPortNum); } else { fRTCPSocket = new Groupsock(env(), tempAddr, rtcpPortNum, 255); } if (fRTCPSocket != NULL && fRTCPSocket->socketNum() >= 0) { // Success! Use these two sockets. success = True; break; } else { // We couldn't create the RTCP socket (perhaps that port number's already in use elsewhere?). delete fRTCPSocket; // Record the first socket in our table, and keep trying: unsigned key = (unsigned) fClientPortNum; Groupsock* existing = (Groupsock*) socketHashTable->Add( (char const*) key, fRTPSocket); delete existing; // in case it wasn't NULL continue; } } // Clean up the socket hash table (and contents): Groupsock* oldGS; while ((oldGS = (Groupsock*) socketHashTable->RemoveNext()) != NULL) { delete oldGS; } delete socketHashTable; if (!success) break; // a fatal error occurred trying to create the RTP and RTCP sockets; we can't continue } // Try to use a big receive buffer for RTP - at least 0.1 second of // specified bandwidth and at least 50 KB unsigned rtpBufSize = fBandwidth * 25 / 2; // 1 kbps * 0.1 s = 12.5 bytes if (rtpBufSize < 50 * 1024) rtpBufSize = 50 * 1024; increaseReceiveBufferTo(env(), fRTPSocket->socketNum(), rtpBufSize); // ASSERT: fRTPSocket != NULL && fRTCPSocket != NULL if (isSSM()) { // Special case for RTCP SSM: Send RTCP packets back to the source via unicast: fRTCPSocket->changeDestinationParameters(fSourceFilterAddr, 0, ~0); } //创建RTPSource的地方 // Create "fRTPSource" and "fReadSource": if (!createSourceObjects(useSpecialRTPoffset)) break; if (fReadSource == NULL) { env().setResultMsg("Failed to create read source"); break; } // Finally, create our RTCP instance. (It starts running automatically) if (fRTPSource != NULL) { // If bandwidth is specified, use it and add 5% for RTCP overhead. // Otherwise make a guess at 500 kbps. unsigned totSessionBandwidth = fBandwidth ? fBandwidth + fBandwidth / 20 : 500; fRTCPInstance = RTCPInstance::createNew(env(), fRTCPSocket, totSessionBandwidth, (unsigned char const*) fParent.CNAME(), NULL /* we're a client */, fRTPSource); if (fRTCPInstance == NULL) { env().setResultMsg("Failed to create RTCP instance"); break; } } return True; } while (0); //失败时执行到这里 delete fRTPSocket; fRTPSocket = NULL; delete fRTCPSocket; fRTCPSocket = NULL; Medium::close(fRTCPInstance); fRTCPInstance = NULL; Medium::close(fReadSource); fReadSource = fRTPSource = NULL; fClientPortNum = 0; return False; }是的，在其中创建了RTP/RTCP socket并创建了RTPSource，创建RTPSource在函数createSourceObjects()中，看一下：

Boolean MediaSubsession::createSourceObjects(int useSpecialRTPoffset) { do { // First, check "fProtocolName" if (strcmp(fProtocolName, "UDP") == 0) { // A UDP-packetized stream (*not* a RTP stream) fReadSource = BasicUDPSource::createNew(env(), fRTPSocket); fRTPSource = NULL; // Note! if (strcmp(fCodecName, "MP2T") == 0) { // MPEG-2 Transport Stream fReadSource = MPEG2TransportStreamFramer::createNew(env(), fReadSource); // this sets "durationInMicroseconds" correctly, based on the PCR values } } else { // Check "fCodecName" against the set of codecs that we support, // and create our RTP source accordingly // (Later make this code more efficient, as this set grows #####) // (Also, add more fmts that can be implemented by SimpleRTPSource#####) Boolean createSimpleRTPSource = False; // by default; can be changed below Boolean doNormalMBitRule = False; // default behavior if "createSimpleRTPSource" is True if (strcmp(fCodecName, "QCELP") == 0) { // QCELP audio fReadSource = QCELPAudioRTPSource::createNew(env(), fRTPSocket, fRTPSource, fRTPPayloadFormat, fRTPTimestampFrequency); // Note that fReadSource will differ from fRTPSource in this case } else if (strcmp(fCodecName, "AMR") == 0) { // AMR audio (narrowband) fReadSource = AMRAudioRTPSource::createNew(env(), fRTPSocket, fRTPSource, fRTPPayloadFormat, 0 /*isWideband*/, fNumChannels, fOctetalign, fInterleaving, fRobustsorting, fCRC); // Note that fReadSource will differ from fRTPSource in this case } else if (strcmp(fCodecName, "AMR-WB") == 0) { // AMR audio (wideband) fReadSource = AMRAudioRTPSource::createNew(env(), fRTPSocket, fRTPSource, fRTPPayloadFormat, 1 /*isWideband*/, fNumChannels, fOctetalign, fInterleaving, fRobustsorting, fCRC); // Note that fReadSource will differ from fRTPSource in this case } else if (strcmp(fCodecName, "MPA") == 0) { // MPEG-1 or 2 audio fReadSource = fRTPSource = MPEG1or2AudioRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "MPA-ROBUST") == 0) { // robust MP3 audio fRTPSource = MP3ADURTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); if (fRTPSource == NULL) break; // Add a filter that deinterleaves the ADUs after depacketizing them: MP3ADUdeinterleaver* deinterleaver = MP3ADUdeinterleaver::createNew( env(), fRTPSource); if (deinterleaver == NULL) break; // Add another filter that converts these ADUs to MP3 frames: fReadSource = MP3FromADUSource::createNew(env(), deinterleaver); } else if (strcmp(fCodecName, "X-MP3-DRAFT-00") == 0) { // a non-standard variant of "MPA-ROBUST" used by RealNetworks // (one 'ADU'ized MP3 frame per packet; no headers) fRTPSource = SimpleRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, "audio/MPA-ROBUST" /*hack*/); if (fRTPSource == NULL) break; // Add a filter that converts these ADUs to MP3 frames: fReadSource = MP3FromADUSource::createNew(env(), fRTPSource, False /*no ADU header*/); } else if (strcmp(fCodecName, "MP4A-LATM") == 0) { // MPEG-4 LATM audio fReadSource = fRTPSource = MPEG4LATMAudioRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "AC3") == 0 || strcmp(fCodecName, "EAC3") == 0) { // AC3 audio fReadSource = fRTPSource = AC3AudioRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "MP4V-ES") == 0) { // MPEG-4 Elem Str vid fReadSource = fRTPSource = MPEG4ESVideoRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "MPEG4-GENERIC") == 0) { fReadSource = fRTPSource = MPEG4GenericRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, fMediumName, fMode, fSizelength, fIndexlength, fIndexdeltalength); } else if (strcmp(fCodecName, "MPV") == 0) { // MPEG-1 or 2 video fReadSource = fRTPSource = MPEG1or2VideoRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "MP2T") == 0) { // MPEG-2 Transport Stream fRTPSource = SimpleRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, "video/MP2T", 0, False); fReadSource = MPEG2TransportStreamFramer::createNew(env(), fRTPSource); // this sets "durationInMicroseconds" correctly, based on the PCR values } else if (strcmp(fCodecName, "H261") == 0) { // H.261 fReadSource = fRTPSource = H261VideoRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "H263-1998") == 0 || strcmp(fCodecName, "H263-2000") == 0) { // H.263+ fReadSource = fRTPSource = H263plusVideoRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "H264") == 0) { fReadSource = fRTPSource = H264VideoRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "DV") == 0) { fReadSource = fRTPSource = DVVideoRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency); } else if (strcmp(fCodecName, "JPEG") == 0) { // motion JPEG fReadSource = fRTPSource = JPEGVideoRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, videoWidth(), videoHeight()); } else if (strcmp(fCodecName, "X-QT") == 0 || strcmp(fCodecName, "X-QUICKTIME") == 0) { // Generic QuickTime streams, as defined in // char* mimeType = new char[strlen(mediumName()) + strlen(codecName()) + 2]; sprintf(mimeType, "%s/%s", mediumName(), codecName()); fReadSource = fRTPSource = QuickTimeGenericRTPSource::createNew( env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, mimeType); delete[] mimeType; } else if (strcmp(fCodecName, "PCMU") == 0 // PCM u-law audio || strcmp(fCodecName, "GSM") == 0 // GSM audio || strcmp(fCodecName, "DVI4") == 0 // DVI4 (IMA ADPCM) audio || strcmp(fCodecName, "PCMA") == 0 // PCM a-law audio || strcmp(fCodecName, "MP1S") == 0 // MPEG-1 System Stream || strcmp(fCodecName, "MP2P") == 0 // MPEG-2 Program Stream || strcmp(fCodecName, "L8") == 0 // 8-bit linear audio || strcmp(fCodecName, "L16") == 0 // 16-bit linear audio || strcmp(fCodecName, "L20") == 0 // 20-bit linear audio (RFC 3190) || strcmp(fCodecName, "L24") == 0 // 24-bit linear audio (RFC 3190) || strcmp(fCodecName, "G726-16") == 0 // G.726, 16 kbps || strcmp(fCodecName, "G726-24") == 0 // G.726, 24 kbps || strcmp(fCodecName, "G726-32") == 0 // G.726, 32 kbps || strcmp(fCodecName, "G726-40") == 0 // G.726, 40 kbps || strcmp(fCodecName, "SPEEX") == 0 // SPEEX audio || strcmp(fCodecName, "T140") == 0 // T.140 text (RFC 4103) || strcmp(fCodecName, "DAT12") == 0 // 12-bit nonlinear audio (RFC 3190) ) { createSimpleRTPSource = True; useSpecialRTPoffset = 0; } else if (useSpecialRTPoffset >= 0) { // We don't know this RTP payload format, but try to receive // it using a 'SimpleRTPSource' with the specified header offset: createSimpleRTPSource = True; } else { env().setResultMsg( "RTP payload format unknown or not supported"); break; } if (createSimpleRTPSource) { char* mimeType = new char[strlen(mediumName()) + strlen(codecName()) + 2]; sprintf(mimeType, "%s/%s", mediumName(), codecName()); fReadSource = fRTPSource = SimpleRTPSource::createNew(env(), fRTPSocket, fRTPPayloadFormat, fRTPTimestampFrequency, mimeType, (unsigned) useSpecialRTPoffset, doNormalMBitRule); delete[] mimeType; } } return True; } while (0); return False; // an error occurred }可以看到，这个函数里主要是跟据前面分析出的媒体和传输信息建立合适的Source。

socket建立了，Source也创建了，下一步应该是连接Sink，形成一个流。到此为止还未看到Sink的影子，应该是在下一步SETUP中建立，我们看到在continueAfterDESCRIBE()的最后调用了setupStreams（），那么就来探索一下setupStreams():

void setupStreams() { static MediaSubsessionIterator* setupIter = NULL; if (setupIter == NULL) setupIter = new MediaSubsessionIterator(*session); //每次调用此函数只为一个Subsession发出SETUP请求。 while ((subsession = setupIter->next()) != NULL) { // We have another subsession left to set up: if (subsession->clientPortNum() == 0) continue; // port # was not set //为一个Subsession发送SETUP请求。请求处理完成时调用continueAfterSETUP()， //continueAfterSETUP()又调用了setupStreams()，在此函数中为下一个SubSession发送SETUP请求。 //直到处理完所有的SubSession setupSubsession(subsession, streamUsingTCP, continueAfterSETUP); return; } //执行到这里时，已循环完所有的SubSession了 // We're done setting up subsessions. delete setupIter; if (!madeProgress) shutdown(); //创建输出文件，看来是在这里创建Sink了。创建sink后，就开始播放它。这个播放应该只是把socket的handler加入到 //计划任务中，而没有数据的接收或发送。只有等到发出PLAY请求后才有数据的收发。 // Create output files: if (createReceivers) { if (outputQuickTimeFile) { // Create a "QuickTimeFileSink", to write to 'stdout': qtOut = QuickTimeFileSink::createNew(*env, *session, "stdout", fileSinkBufferSize, movieWidth, movieHeight, movieFPS, packetLossCompensate, syncStreams, generateHintTracks, generateMP4Format); if (qtOut == NULL) { *env << "Failed to create QuickTime file sink for stdout: " << env->getResultMsg(); shutdown(); } qtOut->startPlaying(sessionAfterPlaying, NULL); } else if (outputAVIFile) { // Create an "AVIFileSink", to write to 'stdout': aviOut = AVIFileSink::createNew(*env, *session, "stdout", fileSinkBufferSize, movieWidth, movieHeight, movieFPS, packetLossCompensate); if (aviOut == NULL) { *env << "Failed to create AVI file sink for stdout: " << env->getResultMsg(); shutdown(); } aviOut->startPlaying(sessionAfterPlaying, NULL); } else { // Create and start "FileSink"s for each subsession: madeProgress = False; MediaSubsessionIterator iter(*session); while ((subsession = iter.next()) != NULL) { if (subsession->readSource() == NULL) continue; // was not initiated // Create an output file for each desired stream: char outFileName[1000]; if (singleMedium == NULL) { // Output file name is // "--" static unsigned streamCounter = 0; snprintf(outFileName, sizeof outFileName, "%s%s-%s-%d", fileNamePrefix, subsession->mediumName(), subsession->codecName(), ++streamCounter); } else { sprintf(outFileName, "stdout"); } FileSink* fileSink; if (strcmp(subsession->mediumName(), "audio") == 0 && (strcmp(subsession->codecName(), "AMR") == 0 || strcmp(subsession->codecName(), "AMR-WB") == 0)) { // For AMR audio streams, we use a special sink that inserts AMR frame hdrs: fileSink = AMRAudioFileSink::createNew(*env, outFileName, fileSinkBufferSize, oneFilePerFrame); } else if (strcmp(subsession->mediumName(), "video") == 0 && (strcmp(subsession->codecName(), "H264") == 0)) { // For H.264 video stream, we use a special sink that insert start_codes: fileSink = H264VideoFileSink::createNew(*env, outFileName, subsession->fmtp_spropparametersets(), fileSinkBufferSize, oneFilePerFrame); } else { // Normal case: fileSink = FileSink::createNew(*env, outFileName, fileSinkBufferSize, oneFilePerFrame); } subsession->sink = fileSink; if (subsession->sink == NULL) { *env << "Failed to create FileSink for \"" << outFileName << "\": " << env->getResultMsg() << "\n"; } else { if (singleMedium == NULL) { *env << "Created output file: \"" << outFileName << "\"\n"; } else { *env << "Outputting data from the \"" << subsession->mediumName() << "/" << subsession->codecName() << "\" subsession to 'stdout'\n"; } if (strcmp(subsession->mediumName(), "video") == 0 && strcmp(subsession->codecName(), "MP4V-ES") == 0 && subsession->fmtp_config() != NULL) { // For MPEG-4 video RTP streams, the 'config' information // from the SDP description contains useful VOL etc. headers. // Insert this data at the front of the output file: unsigned configLen; unsigned char* configData = parseGeneralConfigStr(subsession->fmtp_config(), configLen); struct timeval timeNow; gettimeofday(&timeNow, NULL); fileSink->addData(configData, configLen, timeNow); delete[] configData; } //开始传输 subsession->sink->startPlaying(*(subsession->readSource()), subsessionAfterPlaying, subsession); // Also set a handler to be called if a RTCP "BYE" arrives // for this subsession: if (subsession->rtcpInstance() != NULL) { subsession->rtcpInstance()->setByeHandler( subsessionByeHandler, subsession); } madeProgress = True; } } if (!madeProgress) shutdown(); } } // Finally, start playing each subsession, to start the data flow: if (duration == 0) { if (scale > 0) duration = session->playEndTime() - initialSeekTime; // use SDP end time else if (scale < 0) duration = initialSeekTime; } if (duration < 0) duration = 0.0; endTime = initialSeekTime; if (scale > 0) { if (duration <= 0) endTime = -1.0f; else endTime = initialSeekTime + duration; } else { endTime = initialSeekTime - duration; if (endTime < 0) endTime = 0.0f; } //发送PLAY请求，之后才能从Server端接收数据 startPlayingSession(session, initialSeekTime, endTime, scale, continueAfterPLAY); }仔细看看注释，应很容易了解此函数。