quicktimefilesink.cpp

H.264 RTSP 串流(live 555)視窗版本

/**********
This library is free software; you can redistribute it and/or modify it under
the terms of the GNU Lesser General Public License as published by the
Free Software Foundation; either version 2.1 of the License, or (at your
option) any later version. (See <http://www.gnu.org/copyleft/lesser.html>.)

This library is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public License for
more details.

You should have received a copy of the GNU Lesser General Public License
along with this library; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301  USA
**********/
// "liveMedia"
// Copyright (c) 1996-2010 Live Networks, Inc.  All rights reserved.
// A sink that generates a QuickTime file from a composite media session
// Implementation

#include "QuickTimeFileSink.hh"
#include "QuickTimeGenericRTPSource.hh"
#include "GroupsockHelper.hh"
#include "InputFile.hh"
#include "OutputFile.hh"
#include "H263plusVideoRTPSource.hh" // for the special header
#include "MPEG4GenericRTPSource.hh" //for "samplingFrequencyFromAudioSpecificConfig()"
#include "MPEG4LATMAudioRTPSource.hh" // for "parseGeneralConfigStr()"
#include "Base64.hh"

#include <ctype.h>

#define fourChar(x,y,z,w) ( ((x)<<24)|((y)<<16)|((z)<<8)|(w) )

#define H264_IDR_FRAME 0x65  //bit 8 == 0, bits 7-6 (ref) == 3, bits 5-0 (type) == 5

////////// SubsessionIOState, ChunkDescriptor ///////////
// A structure used to represent the I/O state of each input 'subsession':

class ChunkDescriptor {
public:
  ChunkDescriptor(int64_t offsetInFile, unsigned size,
		  unsigned frameSize, unsigned frameDuration,
		  struct timeval presentationTime);
  virtual ~ChunkDescriptor();

  ChunkDescriptor* extendChunk(int64_t newOffsetInFile, unsigned newSize,
			       unsigned newFrameSize,
			       unsigned newFrameDuration,
			       struct timeval newPresentationTime);
      // this may end up allocating a new chunk instead
public:
  ChunkDescriptor* fNextChunk;
  int64_t fOffsetInFile;
  unsigned fNumFrames;
  unsigned fFrameSize;
  unsigned fFrameDuration;
  struct timeval fPresentationTime; // of the start of the data
};

class SubsessionBuffer {
public:
  SubsessionBuffer(unsigned bufferSize)
    : fBufferSize(bufferSize) {
    reset();
    fData = new unsigned char[bufferSize];
  }
  virtual ~SubsessionBuffer() { delete[] fData; }
  void reset() { fBytesInUse = 0; }
  void addBytes(unsigned numBytes) { fBytesInUse += numBytes; }

  unsigned char* dataStart() { return &fData[0]; }
  unsigned char* dataEnd() { return &fData[fBytesInUse]; }
  unsigned bytesInUse() const { return fBytesInUse; }
  unsigned bytesAvailable() const { return fBufferSize - fBytesInUse; }

  void setPresentationTime(struct timeval const& presentationTime) {
    fPresentationTime = presentationTime;
  }
  struct timeval const& presentationTime() const {return fPresentationTime;}

private:
  unsigned fBufferSize;
  struct timeval fPresentationTime;
  unsigned char* fData;
  unsigned fBytesInUse;
};

class SyncFrame {
public:
  SyncFrame(unsigned frameNum);
  virtual ~SyncFrame();

public:
  class SyncFrame *nextSyncFrame;
  unsigned sfFrameNum;  
};

// A 64-bit counter, used below:
class Count64 {
public:
  Count64()
    : hi(0), lo(0) {
  }

  void operator+=(unsigned arg);

  u_int32_t hi, lo;
};

class SubsessionIOState {
public:
  SubsessionIOState(QuickTimeFileSink& sink, MediaSubsession& subsession);
  virtual ~SubsessionIOState();

  Boolean setQTstate();
  void setFinalQTstate();

  void afterGettingFrame(unsigned packetDataSize,
			 struct timeval presentationTime);
  void onSourceClosure();

  Boolean syncOK(struct timeval presentationTime);
      // returns true iff data is usable despite a sync check

  static void setHintTrack(SubsessionIOState* hintedTrack,
			   SubsessionIOState* hintTrack);
  Boolean isHintTrack() const { return fTrackHintedByUs != NULL; }
  Boolean hasHintTrack() const { return fHintTrackForUs != NULL; }

  UsageEnvironment& envir() const { return fOurSink.envir(); }

public:
  static unsigned fCurrentTrackNumber;
  unsigned fTrackID;
  SubsessionIOState* fHintTrackForUs; SubsessionIOState* fTrackHintedByUs;

  SubsessionBuffer *fBuffer, *fPrevBuffer;
  QuickTimeFileSink& fOurSink;
  MediaSubsession& fOurSubsession;

  unsigned short fLastPacketRTPSeqNum;
  Boolean fOurSourceIsActive;

  Boolean fHaveBeenSynced; // used in synchronizing with other streams
  struct timeval fSyncTime;

  Boolean fQTEnableTrack;
  unsigned fQTcomponentSubtype;
  char const* fQTcomponentName;
  typedef unsigned (QuickTimeFileSink::*atomCreationFunc)();
  atomCreationFunc fQTMediaInformationAtomCreator;
  atomCreationFunc fQTMediaDataAtomCreator;
  char const* fQTAudioDataType;
  unsigned short fQTSoundSampleVersion;
  unsigned fQTTimeScale;
  unsigned fQTTimeUnitsPerSample;
  unsigned fQTBytesPerFrame;
  unsigned fQTSamplesPerFrame;
  // These next fields are derived from the ones above,
  // plus the information from each chunk:
  unsigned fQTTotNumSamples;
  unsigned fQTDurationM; // in media time units
  unsigned fQTDurationT; // in track time units
  int64_t fTKHD_durationPosn;
      // position of the duration in the output 'tkhd' atom
  unsigned fQTInitialOffsetDuration;
      // if there's a pause at the beginning

  ChunkDescriptor *fHeadChunk, *fTailChunk;
  unsigned fNumChunks;
  SyncFrame *fHeadSyncFrame, *fTailSyncFrame;

  // Counters to be used in the hint track's 'udta'/'hinf' atom;
  struct hinf {
    Count64 trpy;
    Count64 nump;
    Count64 tpyl;
    // Is 'maxr' needed? Computing this would be a PITA. #####
    Count64 dmed;
    Count64 dimm;
    // 'drep' is always 0
    // 'tmin' and 'tmax' are always 0
    unsigned pmax;
    unsigned dmax;
  } fHINF;

private:
  void useFrame(SubsessionBuffer& buffer);
  void useFrameForHinting(unsigned frameSize,
			  struct timeval presentationTime,
			  unsigned startSampleNumber);

  // used by the above two routines:
  unsigned useFrame1(unsigned sourceDataSize,
		     struct timeval presentationTime,
		     unsigned frameDuration, int64_t destFileOffset);
      // returns the number of samples in this data

private:
  // A structure used for temporarily storing frame state:
  struct {
    unsigned frameSize;
    struct timeval presentationTime;
    int64_t destFileOffset; // used for non-hint tracks only

    // The remaining fields are used for hint tracks only:
    unsigned startSampleNumber;
    unsigned short seqNum;
    unsigned rtpHeader;
    unsigned char numSpecialHeaders; // used when our RTP source has special headers
    unsigned specialHeaderBytesLength; // ditto
    unsigned char specialHeaderBytes[SPECIAL_HEADER_BUFFER_SIZE]; // ditto
    unsigned packetSizes[256];
  } fPrevFrameState;
};


////////// QuickTimeFileSink implementation //////////

QuickTimeFileSink::QuickTimeFileSink(UsageEnvironment& env,
				     MediaSession& inputSession,
				     char const* outputFileName,
				     unsigned bufferSize,
				     unsigned short movieWidth,
				     unsigned short movieHeight,
				     unsigned movieFPS,
				     Boolean packetLossCompensate,
				     Boolean syncStreams,
				     Boolean generateHintTracks,
				     Boolean generateMP4Format)
  : Medium(env), fInputSession(inputSession),
    fBufferSize(bufferSize), fPacketLossCompensate(packetLossCompensate),
    fSyncStreams(syncStreams), fGenerateMP4Format(generateMP4Format),
    fAreCurrentlyBeingPlayed(False),
    fLargestRTPtimestampFrequency(0),
    fNumSubsessions(0), fNumSyncedSubsessions(0),
    fHaveCompletedOutputFile(False),
    fMovieWidth(movieWidth), fMovieHeight(movieHeight),
    fMovieFPS(movieFPS), fMaxTrackDurationM(0) {
  fOutFid = OpenOutputFile(env, outputFileName);
  if (fOutFid == NULL) return;

  fNewestSyncTime.tv_sec = fNewestSyncTime.tv_usec = 0;
  fFirstDataTime.tv_sec = fFirstDataTime.tv_usec = (unsigned)(~0);

  // Set up I/O state for each input subsession:
  MediaSubsessionIterator iter(fInputSession);
  MediaSubsession* subsession;
  while ((subsession = iter.next()) != NULL) {
    // Ignore subsessions without a data source:
    FramedSource* subsessionSource = subsession->readSource();
    if (subsessionSource == NULL) continue;

    // If "subsession's" SDP description specified screen dimension
    // or frame rate parameters, then use these.  (Note that this must
    // be done before the call to "setQTState()" below.)
    if (subsession->videoWidth() != 0) {
      fMovieWidth = subsession->videoWidth();
    }
    if (subsession->videoHeight() != 0) {
      fMovieHeight = subsession->videoHeight();
    }
    if (subsession->videoFPS() != 0) {
      fMovieFPS = subsession->videoFPS();
    }

    SubsessionIOState* ioState
      = new SubsessionIOState(*this, *subsession);
    if (ioState == NULL || !ioState->setQTstate()) {
      // We're not able to output a QuickTime track for this subsession
      delete ioState; ioState = NULL;
      continue;
    }
    subsession->miscPtr = (void*)ioState;

    if (generateHintTracks) {
      // Also create a hint track for this track:
      SubsessionIOState* hintTrack
	= new SubsessionIOState(*this, *subsession);
      SubsessionIOState::setHintTrack(ioState, hintTrack);
      if (!hintTrack->setQTstate()) {
	delete hintTrack;
	SubsessionIOState::setHintTrack(ioState, NULL);
      }
    }

    // Also set a 'BYE' handler for this subsession's RTCP instance:
    if (subsession->rtcpInstance() != NULL) {
      subsession->rtcpInstance()->setByeHandler(onRTCPBye, ioState);
    }

    unsigned rtpTimestampFrequency = subsession->rtpTimestampFrequency();
    if (rtpTimestampFrequency > fLargestRTPtimestampFrequency) {
      fLargestRTPtimestampFrequency = rtpTimestampFrequency;
    }

    ++fNumSubsessions;
  }

  // Use the current time as the file's creation and modification
  // time.  Use Apple's time format: seconds since January 1, 1904

  gettimeofday(&fStartTime, NULL);
  fAppleCreationTime = fStartTime.tv_sec - 0x83dac000;

  // Begin by writing a "mdat" atom at the start of the file.
  // (Later, when we've finished copying data to the file, we'll come
  // back and fill in its size.)
  fMDATposition = TellFile64(fOutFid);
  addAtomHeader64("mdat");
  // add 64Bit offset
  fMDATposition += 8;
}

QuickTimeFileSink::~QuickTimeFileSink() {
  completeOutputFile();

  // Then, delete each active "SubsessionIOState":
  MediaSubsessionIterator iter(fInputSession);
  MediaSubsession* subsession;
  while ((subsession = iter.next()) != NULL) {
    SubsessionIOState* ioState
      = (SubsessionIOState*)(subsession->miscPtr);
    if (ioState == NULL) continue;

    delete ioState->fHintTrackForUs; // if any
    delete ioState;
  }

  // Finally, close our output file:
  CloseOutputFile(fOutFid);
}

QuickTimeFileSink*
QuickTimeFileSink::createNew(UsageEnvironment& env,
			     MediaSession& inputSession,
			     char const* outputFileName,
			     unsigned bufferSize,
			     unsigned short movieWidth,
			     unsigned short movieHeight,
			     unsigned movieFPS,
			     Boolean packetLossCompensate,
			     Boolean syncStreams,
			     Boolean generateHintTracks,
			     Boolean generateMP4Format) {
  QuickTimeFileSink* newSink = 
    new QuickTimeFileSink(env, inputSession, outputFileName, bufferSize, movieWidth, movieHeight, movieFPS,
			  packetLossCompensate, syncStreams, generateHintTracks, generateMP4Format);
  if (newSink == NULL || newSink->fOutFid == NULL) {
    Medium::close(newSink);
    return NULL;
  }

  return newSink;
}

Boolean QuickTimeFileSink::startPlaying(afterPlayingFunc* afterFunc,
					void* afterClientData) {
  // Make sure we're not already being played:
  if (fAreCurrentlyBeingPlayed) {
    envir().setResultMsg("This sink has already been played");
    return False;
  }

  fAreCurrentlyBeingPlayed = True;
  fAfterFunc = afterFunc;
  fAfterClientData = afterClientData;