quicktimefilesink.cpp

Live media server used to transfer mpeg or other video/audio media

    // We can extend this existing chunk, provided that the frame size
    // and frame duration have not changed:
    if (newFrameSize == fFrameSize && newFrameDuration == fFrameDuration) {
      fNumFrames += newSize/fFrameSize;
      return this;
    }
  }

  // We'll allocate a new ChunkDescriptor, and link it to the end of us:
  ChunkDescriptor* newDescriptor
    = new ChunkDescriptor(newOffsetInFile, newSize,
			  newFrameSize, newFrameDuration,
			  newPresentationTime);

  fNextChunk = newDescriptor;

  return newDescriptor;
}


////////// QuickTime-specific implementation //////////

unsigned QuickTimeFileSink::addWord(unsigned word) {
  addByte(word>>24); addByte(word>>16);
  addByte(word>>8); addByte(word);

  return 4;
}

unsigned QuickTimeFileSink::addHalfWord(unsigned short halfWord) {
  addByte((unsigned char)(halfWord>>8)); addByte((unsigned char)halfWord);

  return 2;
}

unsigned QuickTimeFileSink::addZeroWords(unsigned numWords) {
  for (unsigned i = 0; i < numWords; ++i) {
    addWord(0);
  }

  return numWords*4;
}

unsigned QuickTimeFileSink::add4ByteString(char const* str) {
  addByte(str[0]); addByte(str[1]); addByte(str[2]); addByte(str[3]);

  return 4;
}

unsigned QuickTimeFileSink::addArbitraryString(char const* str,
					       Boolean oneByteLength) {
  unsigned size = 0;
  if (oneByteLength) {
    // Begin with a byte containing the string length:
    unsigned strLength = strlen(str);
    if (strLength >= 256) {
      envir() << "QuickTimeFileSink::addArbitraryString(\""
	      << str << "\") saw string longer than we know how to handle ("
	      << strLength << "\n";
    }
    size += addByte((unsigned char)strLength);
  }

  while (*str != '\0') {
    size += addByte(*str++);
  }

  return size;
}

unsigned QuickTimeFileSink::addAtomHeader(char const* atomName) {
  // Output a placeholder for the 4-byte size:
  addWord(0);

  // Output the 4-byte atom name:
  add4ByteString(atomName);

  return 8;
}

void QuickTimeFileSink::setWord(unsigned filePosn, unsigned size) {
  do {
    if (fseek(fOutFid, filePosn, SEEK_SET) < 0) break;
    addWord(size);
    if (fseek(fOutFid, 0, SEEK_END) < 0) break; // go back to where we were

    return;
  } while (0);

  // One of the fseek()s failed, probable because we're not a seekable file
  envir() << "QuickTimeFileSink::setWord(): fseek failed (err "
	  << envir().getErrno() << ")\n";
}

// Methods for writing particular atoms.  Note the following macros:

#define addAtom(name) \
    unsigned QuickTimeFileSink::addAtom_##name() { \
    unsigned initFilePosn = ftell(fOutFid); \
    unsigned size = addAtomHeader("" #name "")

#define addAtomEnd \
  setWord(initFilePosn, size); \
  return size; \
}

addAtom(ftyp);
  size += add4ByteString("mp42");
  size += addWord(0x00000000);
  size += add4ByteString("mp42");
  size += add4ByteString("isom");
addAtomEnd;

addAtom(moov);
  size += addAtom_mvhd();

  if (fGenerateMP4Format) {
    size += addAtom_iods();
  }

  // Add a 'trak' atom for each subsession:
  // (For some unknown reason, QuickTime Player (5.0 at least)
  //  doesn't display the movie correctly unless the audio track
  //  (if present) appears before the video track.  So ensure this here.)
  MediaSubsessionIterator iter(fInputSession);
  MediaSubsession* subsession;
  while ((subsession = iter.next()) != NULL) {
    fCurrentIOState = (SubsessionIOState*)(subsession->miscPtr); 
    if (fCurrentIOState == NULL) continue;
    if (strcmp(subsession->mediumName(), "audio") != 0) continue;

    size += addAtom_trak();

    if (fCurrentIOState->hasHintTrack()) {
      // This track has a hint track; output it also:
      fCurrentIOState = fCurrentIOState->fHintTrackForUs;
      size += addAtom_trak();
    }
  }
  iter.reset();
  while ((subsession = iter.next()) != NULL) {
    fCurrentIOState = (SubsessionIOState*)(subsession->miscPtr); 
    if (fCurrentIOState == NULL) continue;
    if (strcmp(subsession->mediumName(), "audio") == 0) continue;

    size += addAtom_trak();

    if (fCurrentIOState->hasHintTrack()) {
      // This track has a hint track; output it also:
      fCurrentIOState = fCurrentIOState->fHintTrackForUs;
      size += addAtom_trak();
    }
  }
addAtomEnd;

addAtom(mvhd);
  size += addWord(0x00000000); // Version + Flags
  size += addWord(fAppleCreationTime); // Creation time
  size += addWord(fAppleCreationTime); // Modification time

  // For the "Time scale" field, use the largest RTP timestamp frequency
  // that we saw in any of the subsessions.
  size += addWord(movieTimeScale()); // Time scale

  unsigned const duration = fMaxTrackDurationM;
  fMVHD_durationPosn = ftell(fOutFid);
  size += addWord(duration); // Duration

  size += addWord(0x00010000); // Preferred rate
  size += addWord(0x01000000); // Preferred volume + Reserved[0]
  size += addZeroWords(2); // Reserved[1-2]
  size += addWord(0x00010000); // matrix top left corner
  size += addZeroWords(3); // matrix
  size += addWord(0x00010000); // matrix center
  size += addZeroWords(3); // matrix
  size += addWord(0x40000000); // matrix bottom right corner
  size += addZeroWords(6); // various time fields
  size += addWord(SubsessionIOState::fCurrentTrackNumber+1);// Next track ID
addAtomEnd;

addAtom(iods);
  size += addWord(0x00000000); // Version + Flags
  size += addWord(0x10808080);
  size += addWord(0x07004FFF);
  size += addWord(0xFF0FFFFF);
addAtomEnd;

addAtom(trak);
  size += addAtom_tkhd();

  // If we're synchronizing the media streams (or are a hint track),
  // add an edit list that helps do this:
  if (fCurrentIOState->fHeadChunk != NULL
      && (fSyncStreams || fCurrentIOState->isHintTrack())) {
    size += addAtom_edts();
  }

  // If we're generating a hint track, add a 'tref' atom:
  if (fCurrentIOState->isHintTrack()) size += addAtom_tref();

  size += addAtom_mdia();

  // If we're generating a hint track, add a 'udta' atom:
  if (fCurrentIOState->isHintTrack()) size += addAtom_udta();
addAtomEnd;

addAtom(tkhd);
  if (fCurrentIOState->fQTEnableTrack) {
    size += addWord(0x0000000F); // Version +  Flags
  } else {
    // Disable this track in the movie:
    size += addWord(0x00000000); // Version +  Flags
  }
  size += addWord(fAppleCreationTime); // Creation time
  size += addWord(fAppleCreationTime); // Modification time
  size += addWord(fCurrentIOState->fTrackID); // Track ID
  size += addWord(0x00000000); // Reserved

  unsigned const duration = fCurrentIOState->fQTDurationM; // movie units
  fCurrentIOState->fTKHD_durationPosn = ftell(fOutFid);
  size += addWord(duration); // Duration
  size += addZeroWords(3); // Reserved+Layer+Alternate grp
  size += addWord(0x01000000); // Volume + Reserved
  size += addWord(0x00010000); // matrix top left corner
  size += addZeroWords(3); // matrix
  size += addWord(0x00010000); // matrix center
  size += addZeroWords(3); // matrix
  size += addWord(0x40000000); // matrix bottom right corner
  if (strcmp(fCurrentIOState->fOurSubsession.mediumName(), "video") == 0) {
    size += addWord(fMovieWidth<<16); // Track width
    size += addWord(fMovieHeight<<16); // Track height
  } else {
    size += addZeroWords(2); // not video: leave width and height fields zero
  }
addAtomEnd;

addAtom(edts);
  size += addAtom_elst();
addAtomEnd;

#define addEdit1(duration,trackPosition) do { \
      unsigned trackDuration \
        = (unsigned) ((2*(duration)*movieTimeScale()+1)/2); \
            /* in movie time units */ \
      size += addWord(trackDuration); /* Track duration */ \
      totalDurationOfEdits += trackDuration; \
      size += addWord(trackPosition); /* Media time */ \
      size += addWord(0x00010000); /* Media rate (1x) */ \
      ++numEdits; \
} while (0)
#define addEdit(duration) addEdit1((duration),editTrackPosition)
#define addEmptyEdit(duration) addEdit1((duration),(~0))

addAtom(elst);
  size += addWord(0x00000000); // Version + Flags

  // Add a dummy "Number of entries" field
  // (and remember its position).  We'll fill this field in later:
  unsigned numEntriesPosition = ftell(fOutFid);
  size += addWord(0); // dummy for "Number of entries"
  unsigned numEdits = 0;
  unsigned totalDurationOfEdits = 0; // in movie time units

  // Run through our chunks, looking at their presentation times.
  // From these, figure out the edits that need to be made to keep
  // the track media data in sync with the presentation times.
  
  double const syncThreshold = 0.1; // 100 ms
    // don't allow the track to get out of sync by more than this

  struct timeval editStartTime = fFirstDataTime;
  unsigned editTrackPosition = 0;
  unsigned currentTrackPosition = 0;
  double trackDurationOfEdit = 0.0;
  unsigned chunkDuration = 0;

  ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk;
  while (chunk != NULL) {
    struct timeval const& chunkStartTime = chunk->fPresentationTime;
    double movieDurationOfEdit
      = (chunkStartTime.tv_sec - editStartTime.tv_sec)
      + (chunkStartTime.tv_usec - editStartTime.tv_usec)/1000000.0;
    trackDurationOfEdit = (currentTrackPosition-editTrackPosition)
      / (double)(fCurrentIOState->fQTTimeScale);
      
    double outOfSync = movieDurationOfEdit - trackDurationOfEdit;

    if (outOfSync > syncThreshold) {
      // The track's data is too short, so end this edit, add a new
      // 'empty' edit after it, and start a new edit
      // (at the current track posn.):
      if (trackDurationOfEdit > 0.0) addEdit(trackDurationOfEdit);
      addEmptyEdit(outOfSync);

      editStartTime = chunkStartTime;
      editTrackPosition = currentTrackPosition;
    } else if (outOfSync < -syncThreshold) {
      // The track's data is too long, so end this edit, and start
      // a new edit (pointing at the current track posn.):
      if (movieDurationOfEdit > 0.0) addEdit(movieDurationOfEdit);

      editStartTime = chunkStartTime;
      editTrackPosition = currentTrackPosition;
    }

    // Note the duration of this chunk:
    unsigned numChannels = fCurrentIOState->fOurSubsession.numChannels();
    chunkDuration = chunk->fNumFrames*chunk->fFrameDuration/numChannels;
    currentTrackPosition += chunkDuration;

    chunk = chunk->fNextChunk;
  }

  // Write out the final edit
  trackDurationOfEdit
      += (double)chunkDuration/fCurrentIOState->fQTTimeScale;
  if (trackDurationOfEdit > 0.0) addEdit(trackDurationOfEdit);

  // Now go back and fill in the "Number of entries" field:
  setWord(numEntriesPosition, numEdits);

  // Also, if the sum of all of the edit durations exceeds the
  // track duration that we already computed (from sample durations),
  // then reset the track duration to this new value:
  if (totalDurationOfEdits > fCurrentIOState->fQTDurationM) {
    fCurrentIOState->fQTDurationM = totalDurationOfEdits;
    setWord(fCurrentIOState->fTKHD_durationPosn, totalDurationOfEdits);

    // Also, check whether the overall movie duration needs to change:
    if (totalDurationOfEdits > fMaxTrackDurationM) {
      fMaxTrackDurationM = totalDurationOfEdits;
      setWord(fMVHD_durationPosn, totalDurationOfEdits);
    }

    // Also, convert to track time scale:
    double scaleFactor
      = fCurrentIOState->fQTTimeScale/(double)movieTimeScale();
    fCurrentIOState->fQTDurationT
      = (unsigned)(totalDurationOfEdits*scaleFactor);
  }
addAtomEnd;

addAtom(tref);
  size += addAtom_hint();
addAtomEnd;

addAtom(hint);
  SubsessionIOState* hintedTrack = fCurrentIOState->fTrackHintedByUs;
    // Assert: hintedTrack != NULL
  size += addWord(hintedTrack->fTrackID);
addAtomEnd;

addAtom(mdia);
  size += addAtom_mdhd();
  size += addAtom_hdlr();
  size += addAtom_minf();
addAtomEnd;

addAtom(mdhd);
  size += addWord(0x00000000); // Version + Flags
  size += addWord(fAppleCreationTime); // Creation time
  size += addWord(fAppleCreationTime); // Modification time

  unsigned const timeScale = fCurrentIOState->fQTTimeScale;
  size += addWord(timeScale); // Time scale

  unsigned const duration = fCurrentIOState->fQTDurationT; // track units