根据上文确定的方案，我们会在每一个推流的会话建立时，增加一个自定义的Output对象，用来实现吧接收到的RTP包写到文件中。
首先，我们在RTSPReflectorOutput.cpp中，参考RTPSessionOutput类定义一个RTPSessionSaveOutput类：
class RTPSessionSaveOutput: public ReflectorOutput { public: RTPSessionSaveOutput(ReflectorSession* inReflectorSession, char* theFileName = NULL); virtual ~RTPSessionSaveOutput(); //关键是重写这个方法，完成RTP包的存储。 QTSS_Error WritePacket(StrPtrLen* inPacket, void* inStreamCookie, UInt32 inFlags, SInt64 packetLatenessInMSec, SInt64* timeToSendThisPacketAgain, UInt64* packetIDPtr, SInt64* arrivalTimeMSec, Bool16 firstPacket ); ReflectorSession* GetReflectorSession();//{ return fReflectorSession; } void TearDown() {} Bool16 IsUDP() { return true;} Bool16 IsPlaying() { return true;} bool MkFileDirandFile(); void Write_Header(); private: ReflectorSession* fReflectorSession; char fFilePath[MaxPathSize]; char fFileName[255]; SInt32 ffd; // file handle Bool16 bLastBlockFlg;// SInt64 lastArrivalTimeMSec; };
并完成实现：
#define DEFAULTPATH "/usr/local/movies/pushstream/" RTPSessionSaveOutput::RTPSessionSaveOutput(ReflectorSession* inReflectorSession, char* theFileName) :fReflectorSession(inReflectorSession), ffd(-1) { printf("------- XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX ,theFileName =%s \n", theFileName); ::memset(fFileName, 0, sizeof(fFileName)); ::memset(fFilePath, 0 , sizeof(fFilePath)); bLastBlockFlg = false; lastArrivalTimeMSec = -1; strcpy(fFilePath, DEFAULTPATH); char *firstPos = ::strstr(theFileName, ".sdp"); if (firstPos != NULL){ int length = strlen(theFileName); int i = length - 1; int end = i; while(i > 0){ if (theFileName[i] == firstPos[0]){ end = i; } if (theFileName[i] == '/'){ printf("----setup path:%d, i:%d-------\n", end, i); break; } i--; } if (i > 0 && end > i + 1){ memcpy((char *)fFileName, ((char*)theFileName + i + 1), end - i - 1); fFileName[end - i - 1] = '\0'; sprintf(fFilePath, "%s%s", DEFAULTPATH, fFileName); printf("----fFileName:%s, path:%s-------\n", fFileName, fFilePath); } } MkFileDirandFile(); printf("------- RTPSessionSaveAsMP4, GetNumStreams=%d \n", inReflectorSession->GetNumStreams()); this->InititializeBookmarks( inReflectorSession->GetNumStreams()); } RTPSessionSaveOutput::~RTPSessionSaveOutput() { printf("------- YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY \n"); if(ffd != -1) { close(ffd); ffd = -1; } ::memset(fFileName, 0, sizeof(fFileName)); ::memset(fFilePath, 0 , sizeof(fFilePath)); } bool RTPSessionSaveOutput::MkFileDirandFile() { if(access(fFilePath, R_OK) != 0) // dir exists { int isCreate = mkdir(fFilePath , S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH); if( !isCreate ) qtss_printf("create dir ok, %s\n", fFileName); else { qtss_printf("create %s failed! error code : %d\n", fFilePath, isCreate); return false; } } qtss_printf("---------- dir is ok ---------------%s\n", fFileName); time_t tmt = OS::UnixTime_Secs(); // Seconds since 1970 char szBuf[128] = {0}; strftime(szBuf, sizeof(szBuf), "%Y-%m-%d_%H_%M_%S", localtime(&tmt)); printf(" start0 = %s\n", szBuf); if(-1 == ffd) { sprintf(fFilePath, "%s/%s.m4a", fFilePath, szBuf); printf("-------open file , fFilePath=%s \n", fFilePath); ffd = open(fFilePath, O_CREAT | O_WRONLY | O_APPEND, 0666); if (ffd != -1){ //Write_Header(); } } return true; } ReflectorSession* RTPSessionSaveOutput::GetReflectorSession() { qtss_printf("################################## RTPSessionSaveAsMP4::GetReflectorSession \n"); return fReflectorSession; } void RTPSessionSaveOutput::Write_Header() { if (1){ unsigned char header[4]; // H264 header header[0] = 0; header[1] = 0; header[2] = 0; header[3] = 1; write(ffd, header, 4); }else{ unsigned char header[6] = "\0"; // AMR header header[0] = 0x23; header[1] = 0x21; header[2] = 0x41; header[3] = 0x4D; header[4] = 0x52; header[5] = 0x0A; write(ffd, header, sizeof(header)); } } //这个方法主要实现了AAC RTP音频流的存储，将存储的文件拷贝出来后，能正常播放 QTSS_Error RTPSessionSaveOutput::WritePacket(StrPtrLen* inPacket, void* inStreamCookie, UInt32 inFlags, SInt64 packetLatenessInMSec, SInt64* timeToSendThisPacketAgain, UInt64* packetIDPtr, SInt64* arrivalTimeMSecPtr, Bool16 firstPacket) { if (inPacket == NULL || inPacket->Len < 14){ qtss_printf("inPacket == NULL || inPacket->Len < 14"); return QTSS_NoErr; } SInt64 currentTime = OS::Milliseconds(); if(-1 != ffd) { UInt8* payload = (UInt8*)inPacket->Ptr; int start = 12; //rtp header length //just save aac audio. if (inPacket->Len < 16){ return QTSS_NoErr; } if (payload[start] != 0x00 || payload[start + 1] != 0x10){ return QTSS_NoErr; } //check end. int rtptype = payload[start] & 0x1f; //int payload_type = payload[1]& 0x7f; bool bStartFlg = ((payload[start+1] & 0x80) == 0x80) ? true : false; //qtss_printf("################# rtptype = %d \n", rtptype); #if 0//这里用来做H264文件的存储 if (rtptype== 28) { // FU-A //printf("################# rtptype== 28, arrivalTimeMSecPtr=%ld, len =%d\n", *arrivalTimeMSecPtr, inPacket->Len - start -2); if(bStartFlg) { Write_Header(); } int ret = write(ffd, (void*)(payload + start + 2), (inPacket->Len - start -2)); if(ret == -1) { qtss_printf("error=%d, strerror=%s", errno, strerror(errno)); } } else if (rtptype <= 24){ // single nal unit // printf("####################### rtptype rtptype =%d, arrivalTimeMSecPtr=%ld, len =%d\n", rtptype ,*arrivalTimeMSecPtr, (inPacket->Len - start)); //Write_Header(); int ret = write(ffd, (void*)(payload+start), (inPacket->Len - start)); // 00 00 00 01 if(ret == -1) { qtss_printf("error=%d, strerror=%s", errno, strerror(errno)); } } #else if (lastArrivalTimeMSec >= *arrivalTimeMSecPtr){ return QTSS_NoErr; } qtss_printf("RTPSessionSaveAsMP4::WritePacket lastArrivalTimeMSec=%d\n", lastArrivalTimeMSec); lastArrivalTimeMSec = *arrivalTimeMSecPtr; #if 1 //aac header unsigned char adtsHeader[7] = {0}; adtsHeader[0] = 0xFF; adtsHeader[1] = 0xF1; int profile = 2; int freqIdx = 11; int chanCfg = 1; int packetLen = inPacket->Len - start + 7 - 4; adtsHeader[2] = ((profile -1 )<<6) + (freqIdx << 2) + (chanCfg >> 2); adtsHeader[3] = ((chanCfg & 3) << 6) + (packetLen >> 11); adtsHeader[4] = (packetLen & 0x7ff) >> 3;//(packetLen >> 3) & 0xff; adtsHeader[5] = ((packetLen & 0x7) << 5)|0x1f; adtsHeader[6] = 0xFC; int ret = write(ffd, adtsHeader, sizeof(adtsHeader)); #endif// //amr header //aaclatm 4header ret = write(ffd, (void*)(payload + start + 4), (inPacket->Len - start - 4)); if(ret == -1) { qtss_printf("error=%d, strerror=%s", errno, strerror(errno)); } fLastIntervalMilliSec = currentTime - fLastPacketTransmitTime; if (fLastIntervalMilliSec > 100) { //reset interval maybe first packet or it has been blocked for awhile fLastIntervalMilliSec = 5; } fLastPacketTransmitTime = currentTime; if (!bLastBlockFlg){ bLastBlockFlg = true; }else{ bLastBlockFlg = false; } #endif //flush(ffd); } return QTSS_NoErr; }在完成文件输出的RTPSessionSaveOutput类后，我们修改QTSSReflectorModule中的，在推流的逻辑中，增加一个OutPut对象到会话中： // If this is an incoming data session, skip everything having to do with setting up a new // RTP Stream. if (isPush){ //... //这里暂时只存储音频流 if(theTrackID == 0) // qtssVideoPayloadType { RTPSessionSaveOutput* theSaveAsMP4 = NEW RTPSessionSaveOutput(theSession, theFullPath.Ptr); theSession->AddOutput(theSaveOutput, false); } }
这样就完成了推送音频流的存储，视频流的存储也能这样来实现，如果考虑音视频同时存储，可能要考虑其他方案来实现，比方需要通过ffmpeg来完成声音和视频融合成一个文件。
