Live555学习之(七)---------- Live555实现h264视频的点播

　　上一篇介绍了Live555如何实现录像功能，我录的是H264编码的视频文件。在《Live555的基本介绍》这一篇中，我介绍说把mp3文件放到live/mediaServer目录下，然后使用Live555流化，就可以通过vlc去点播该文件。那么我们录好的h264文件能否被Live555流化然后使用VLC点播呢？经试验，发现是不行的。

　　然后我就对比VLC去请求Live555流化mp3文件和h264文件的过程，发现了原因：在请求h264文件时返回的SDP信息中，总出现"a=range:npt=0-"，而请求mp3文件时返回的SDP信息中，npt后面是0到一个具体的数字，即指定了该ServerMediaSubsession的时间长度。在ServerMediaSession::generateSDPDescription函数中，我们找到与此相关的内容：

 // Unless subsessions have differing durations, we also have a "a=range:" line:
    float dur = duration();
    if (dur == 0.0) {
      rangeLine = strDup("a=range:npt=0-
");
    } else if (dur > 0.0) {
      char buf[100];
      sprintf(buf, "a=range:npt=0-%.3f
", dur);
      rangeLine = strDup(buf);
    } else { // subsessions have differing durations, so "a=range:" lines go there
      rangeLine = strDup("");
    }

　　也就是说，duration函数返回值为0，才会出现"a=range:npt=0-"，表示此ServerMediaSession的持续时间未知（不是表示此ServerMediaSession的持续时间为0）。

float ServerMediaSession::duration() const {
  float minSubsessionDuration = 0.0;
  float maxSubsessionDuration = 0.0;
  for (ServerMediaSubsession* subsession = fSubsessionsHead; subsession != NULL;
       subsession = subsession->fNext) {
    // Hack: If any subsession supports seeking by 'absolute' time, then return a negative value, to indicate that only subsessions
    // will have a "a=range:" attribute:
    char* absStartTime = NULL; char* absEndTime = NULL;
    subsession->getAbsoluteTimeRange(absStartTime, absEndTime);
    if (absStartTime != NULL) return -1.0f;

    float ssduration = subsession->duration();
    if (subsession == fSubsessionsHead) { // this is the first subsession
      minSubsessionDuration = maxSubsessionDuration = ssduration;
    } else if (ssduration < minSubsessionDuration) {
    minSubsessionDuration = ssduration;
    } else if (ssduration > maxSubsessionDuration) {
    maxSubsessionDuration = ssduration;
    }
  }

  if (maxSubsessionDuration != minSubsessionDuration) {
    return -maxSubsessionDuration; // because subsession durations differ
  } else {
    return maxSubsessionDuration; // all subsession durations are the same
  }
}

　　看一下ServerMediaSession::duration函数，发现ServerMediaSession的duration取决于各个ServerMediaSubsession的duration。那我们再看一下ServerMediaSubsession的duration函数:

float ServerMediaSubsession::duration() const {
  // default implementation: assume an unbounded session:
  return 0.0;
}

　　可以看到默认的实现是返回0，而对于H264视频文件对应的ServerMediaSubsession具体类是H264VideoFileServerMediaSubsession，在该类中没有找到覆盖duration的实现，因此对于H264视频文件duration函数返回0。而对于mp3文件，我们看一下MP3AudioFileServerMediaSubsession类中关于duration的实现:

 float MP3AudioFileServerMediaSubsession::duration() const {
   return fFileDuration;　　　　　　　//返回的是fFileDuration，这个值由MP3FileSource得到
 }

　　后来发现对于mkv文件，Live555也是支持点播的，那么我又去看了一下MatroskaFileServerMediaSubsession类关于duration的实现:

float MatroskaFileServerMediaSubsession::duration() const { return fOurDemux.fileDuration(); }

　　总之，对于mp3文件和mkv文件，druation函数都有具体的实现，而对于h264文件，使用的是默认的实现（返回0）。后来在官网找到这么一段说明:

　　这段话列出了Live555支持几种播放动作（包括暂停、点播、快进、倒放）的文件类型，Seeking即点播，从中我们可以看到支持的文件类型没有h264。但是后面又提示说了如果要使这些播放动作可以用于MPEG Transport Stream file（即ts流文件，后缀名是.ts），则必须有一个index file，还提示说可以使用MPEG2TransportStreamIndexer来产生index file。

　　然后我在testOnDemandRTSPServer.cpp中找到了对于.ts文件的处理

 // A MPEG-2 Transport Stream:
  {
    char const* streamName = "mpeg2TransportStreamTest";
    char const* inputFileName = "test.ts";
    char const* indexFileName = "test.tsx";
    ServerMediaSession* sms
      = ServerMediaSession::createNew(*env, streamName, streamName,
                      descriptionString);
    sms->addSubsession(MPEG2TransportFileServerMediaSubsession
               ::createNew(*env, inputFileName, indexFileName, reuseFirstSource));
    rtspServer->addServerMediaSession(sms);

    announceStream(rtspServer, sms, streamName, inputFileName);
  }

　　从中可以看到，对于.ts文件的流化，Live555还需要一个.tsx文件，这个就是对应于.ts文件的index file了。然后我在live/testProgs目录下找到了MPEG2TransportStreamIndexer.cpp

int main(int argc, char const** argv) {
  // Begin by setting up our usage environment:
  TaskScheduler* scheduler = BasicTaskScheduler::createNew();
  env = BasicUsageEnvironment::createNew(*scheduler);


  // Parse the command line:
  programName = argv[0];
  //if (argc != 2) usage();

  char const* inputFileName = "test.ts";
  // Check whether the input file name ends with ".ts":
  int len = strlen(inputFileName);
  if (len < 4 || strcmp(&inputFileName[len-3], ".ts") != 0) {
    *env << "ERROR: input file name "" << inputFileName
     << "" does not end with ".ts"
";
    usage();
  }

  // Open the input file (as a 'byte stream file source'):
  FramedSource* input
    = ByteStreamFileSource::createNew(*env, inputFileName, TRANSPORT_PACKET_SIZE);
  if (input == NULL) {
    *env << "Failed to open input file "" << inputFileName << "" (does it exist?)
";
    exit(1);
  }

  // Create a filter that indexes the input Transport Stream data:
  FramedSource* indexer
    = MPEG2IFrameIndexFromTransportStream::createNew(*env, input);

  // The output file name is the same as the input file name, except with suffix ".tsx":
  char* outputFileName = new char[len+2]; // allow for trailing x
  sprintf(outputFileName, "%sx", inputFileName);

  // Open the output file (for writing), as a 'file sink':
  MediaSink* output = FileSink::createNew(*env, outputFileName);
  if (output == NULL) {
    *env << "Failed to open output file "" << outputFileName << ""
";
    exit(1);
  }



  // Start playing, to generate the output index file:
  *env << "Writing index file "" << outputFileName << ""...";
  output->startPlaying(*indexer, afterPlaying, NULL);

  env->taskScheduler().doEventLoop(); // does not return

  return 0; // only to prevent compiler warning
}

void afterPlaying(void* /*clientData*/) {
  *env << "...done
";
  exit(0);
}

　　这个程序演示了如何根据一个.ts文件产生对应的.tsx文件，只要同时具有.ts文件和.tsx文件就可以实现点播了。那现在如何把h264文件转成.ts文件呢？然后我又在live/testProgs目录下惊喜发现了testH264VideoToTransportStream.cpp文件，来看看

int main(int argc, char** argv) {
  // Begin by setting up our usage environment:
  TaskScheduler* scheduler = BasicTaskScheduler::createNew();
  env = BasicUsageEnvironment::createNew(*scheduler);

  // Open the input file as a 'byte-stream file source':
  FramedSource* inputSource = ByteStreamFileSource::createNew(*env, inputFileName);
  if (inputSource == NULL) {
    *env << "Unable to open file "" << inputFileName
     << "" as a byte-stream file source
";
    exit(1);
  }

  // Create a 'framer' filter for this file source, to generate presentation times for each NAL unit:
  H264VideoStreamFramer* framer = H264VideoStreamFramer::createNew(*env, inputSource, True/*includeStartCodeInOutput*/);

  // Then create a filter that packs the H.264 video data into a Transport Stream:
  MPEG2TransportStreamFromESSource* tsFrames = MPEG2TransportStreamFromESSource::createNew(*env);
  tsFrames->addNewVideoSource(framer, 5/*mpegVersion: H.264*/);
  
  // Open the output file as a 'file sink':
  MediaSink* outputSink = FileSink::createNew(*env, outputFileName);
  if (outputSink == NULL) {
    *env << "Unable to open file "" << outputFileName << "" as a file sink
";
    exit(1);
  }

  // Finally, start playing:
  *env << "Beginning to read...
";
  outputSink->startPlaying(*tsFrames, afterPlaying, NULL);

  env->taskScheduler().doEventLoop(); // does not return

  return 0; // only to prevent compiler warning
}

void afterPlaying(void* /*clientData*/) {
  *env << "Done reading.
";
  *env << "Wrote output file: "" << outputFileName << ""
";
  exit(0);
}

　　这个程序又演示了如何将一个h264视频文件转成.ts格式的文件，这样把上面的两个例子程序结合起来，就可以实现H264文件的点播了。

　　在此，顺便说一下自己对Live555多线程编程的理解，Live555是基于事件驱动的单线程模式，每个TaskScheduler就对应一个Live555线程，那么在我们的程序中可以创建多个TaskScheduler来实现多线程。我们自己程序的其他线程和Live555线程的交互可以通过全局的flag变量或者调用triggerEvent函数。我们来看看Live555官方对于此问题的解答：