Basics of FFmpeg examiner works.

7 files changed, 137 insertions, 43 deletions

diff --git a/src/lib/ffmpeg.cc b/src/lib/ffmpeg.cc
index 0d897abfa..a39de391a 100644
--- a/src/lib/ffmpeg.cc
+++ b/src/lib/ffmpeg.cc
@@ -40,10 +40,6 @@ FFmpeg::FFmpeg (boost::shared_ptr<const FFmpegContent> c)
 	, _format_context (0)
 	, _frame (0)
 	, _video_stream (-1)
-	, _video_codec_context (0)
-	, _video_codec (0)
-	, _audio_codec_context (0)
-	, _audio_codec (0)
 {
 	setup_general ();
 	setup_video ();
@@ -53,13 +49,12 @@ FFmpeg::FFmpeg (boost::shared_ptr<const FFmpegContent> c)
 FFmpeg::~FFmpeg ()
 {
 	boost::mutex::scoped_lock lm (_mutex);
-	
-	if (_audio_codec_context) {
-		avcodec_close (_audio_codec_context);
-	}
 
-	if (_video_codec_context) {
-		avcodec_close (_video_codec_context);
+	for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
+		AVCodecContext* context = _format_context->streams[i]->codec;
+		if (context->codec_type == AVMEDIA_TYPE_VIDEO || context->codec_type == AVMEDIA_TYPE_AUDIO) {
+			avcodec_close (context);
+		}
 	}
 
 	av_free (_frame);
@@ -104,14 +99,14 @@ FFmpeg::setup_video ()
 {
 	boost::mutex::scoped_lock lm (_mutex);
 	
-	_video_codec_context = _format_context->streams[_video_stream]->codec;
-	_video_codec = avcodec_find_decoder (_video_codec_context->codec_id);
+	AVCodecContext* context = _format_context->streams[_video_stream]->codec;
+	AVCodec* codec = avcodec_find_decoder (context->codec_id);
 
-	if (_video_codec == 0) {
+	if (codec == 0) {
 		throw DecodeError (_("could not find video decoder"));
 	}
 
-	if (avcodec_open2 (_video_codec_context, _video_codec, 0) < 0) {
+	if (avcodec_open2 (context, codec, 0) < 0) {
 		throw DecodeError (N_("could not open video decoder"));
 	}
 }
@@ -120,19 +115,33 @@ void
 FFmpeg::setup_audio ()
 {
 	boost::mutex::scoped_lock lm (_mutex);
-	
-	if (!_ffmpeg_content->audio_stream ()) {
-		return;
+
+	for (uint32_t i = 0; i < _format_context->nb_streams; ++i) {
+		AVCodecContext* context = _format_context->streams[i]->codec;
+		if (context->codec_type != AVMEDIA_TYPE_AUDIO) {
+			continue;
+		}
+		
+		AVCodec* codec = avcodec_find_decoder (context->codec_id);
+		if (codec == 0) {
+			throw DecodeError (_("could not find audio decoder"));
+		}
+		
+		if (avcodec_open2 (context, codec, 0) < 0) {
+			throw DecodeError (N_("could not open audio decoder"));
+		}
 	}
+}
 
-	_audio_codec_context = _format_context->streams[_ffmpeg_content->audio_stream()->id]->codec;
-	_audio_codec = avcodec_find_decoder (_audio_codec_context->codec_id);
 
-	if (_audio_codec == 0) {
-		throw DecodeError (_("could not find audio decoder"));
-	}
+AVCodecContext *
+FFmpeg::video_codec_context () const
+{
+	return _format_context->streams[_video_stream]->codec;
+}
 
-	if (avcodec_open2 (_audio_codec_context, _audio_codec, 0) < 0) {
-		throw DecodeError (N_("could not open audio decoder"));
-	}
+AVCodecContext *
+FFmpeg::audio_codec_context () const
+{
+	return _format_context->streams[_ffmpeg_content->audio_stream()->id]->codec;
 }
diff --git a/src/lib/ffmpeg.h b/src/lib/ffmpeg.h
index dcafe17f7..4d1a45da3 100644
--- a/src/lib/ffmpeg.h
+++ b/src/lib/ffmpeg.h
@@ -17,6 +17,9 @@
 
 */
 
+#ifndef DCPOMATIC_FFMPEG_H
+#define DCPOMATIC_FFMPEG_H
+
 #include <vector>
 #include <boost/shared_ptr.hpp>
 #include <boost/thread/mutex.hpp>
@@ -46,16 +49,16 @@ public:
 	}
 
 protected:
+	AVCodecContext* video_codec_context () const;
+	AVCodecContext* audio_codec_context () const;
+	
 	boost::shared_ptr<const FFmpegContent> _ffmpeg_content;
+
 	AVFormatContext* _format_context;
 	AVPacket _packet;
 	AVFrame* _frame;
-	int _video_stream;
 
-	AVCodecContext* _video_codec_context;
-	AVCodec* _video_codec;
-	AVCodecContext* _audio_codec_context;    ///< may be 0 if there is no audio
-	AVCodec* _audio_codec;                   ///< may be 0 if there is no audio
+	int _video_stream;
 
 	/* It would appear (though not completely verified) that one must have
 	   a mutex around calls to avcodec_open* and avcodec_close... and here
@@ -68,3 +71,5 @@ private:
 	void setup_video ();
 	void setup_audio ();
 };
+
+#endif
diff --git a/src/lib/ffmpeg_content.cc b/src/lib/ffmpeg_content.cc
index 43e88c428..68132c5ab 100644
--- a/src/lib/ffmpeg_content.cc
+++ b/src/lib/ffmpeg_content.cc
@@ -76,6 +76,8 @@ FFmpegContent::FFmpegContent (shared_ptr<const Film> f, shared_ptr<const cxml::N
 	for (list<shared_ptr<cxml::Node> >::iterator i = c.begin(); i != c.end(); ++i) {
 		_filters.push_back (Filter::from_id ((*i)->content ()));
 	}
+
+	_first_video = node->optional_number_child<Time> ("FirstVideo");
 }
 
 FFmpegContent::FFmpegContent (FFmpegContent const & o)
@@ -119,6 +121,10 @@ FFmpegContent::as_xml (xmlpp::Node* node) const
 	for (vector<Filter const *>::const_iterator i = _filters.begin(); i != _filters.end(); ++i) {
 		node->add_child("Filter")->add_child_text ((*i)->id ());
 	}
+
+	if (_first_video) {
+		node->add_child("FirstVideo")->add_child_text (lexical_cast<string> (_first_video.get ()));
+	}
 }
 
 void
@@ -135,7 +141,7 @@ FFmpegContent::examine (shared_ptr<Job> job)
 
 	ContentVideoFrame video_length = 0;
 	video_length = examiner->video_length ();
-	film->log()->log (String::compose ("Video length obtained from header as %1 frames", examiner->video_length ()));
+	film->log()->log (String::compose ("Video length obtained from header as %1 frames", video_length));
 
         {
                 boost::mutex::scoped_lock lm (_mutex);
@@ -151,6 +157,8 @@ FFmpegContent::examine (shared_ptr<Job> job)
                 if (!_audio_streams.empty ()) {
                         _audio_stream = _audio_streams.front ();
                 }
+
+		_first_video = examiner->first_video ();
         }
 
         take_from_video_examiner (examiner);
@@ -288,6 +296,7 @@ FFmpegAudioStream::FFmpegAudioStream (shared_ptr<const cxml::Node> node)
 	frame_rate = node->number_child<int> ("FrameRate");
 	channels = node->number_child<int64_t> ("Channels");
 	mapping = AudioMapping (node->node_child ("Mapping"));
+	start = node->optional_number_child<Time> ("Start");
 }
 
 void
@@ -297,6 +306,9 @@ FFmpegAudioStream::as_xml (xmlpp::Node* root) const
 	root->add_child("Id")->add_child_text (lexical_cast<string> (id));
 	root->add_child("FrameRate")->add_child_text (lexical_cast<string> (frame_rate));
 	root->add_child("Channels")->add_child_text (lexical_cast<string> (channels));
+	if (start) {
+		root->add_child("Start")->add_child_text (lexical_cast<string> (start));
+	}
 	mapping.as_xml (root->add_child("Mapping"));
 }
 
diff --git a/src/lib/ffmpeg_content.h b/src/lib/ffmpeg_content.h
index 98078ef3e..36c24c2b3 100644
--- a/src/lib/ffmpeg_content.h
+++ b/src/lib/ffmpeg_content.h
@@ -46,6 +46,7 @@ public:
         int frame_rate;
 	int channels;
 	AudioMapping mapping;
+	boost::optional<Time> start;
 };
 
 extern bool operator== (FFmpegAudioStream const & a, FFmpegAudioStream const & b);
@@ -139,6 +140,7 @@ private:
 	boost::shared_ptr<FFmpegSubtitleStream> _subtitle_stream;
 	std::vector<boost::shared_ptr<FFmpegAudioStream> > _audio_streams;
 	boost::shared_ptr<FFmpegAudioStream> _audio_stream;
+	boost::optional<Time> _first_video;
 	/** Video filters that should be used when generating DCPs */
 	std::vector<Filter const *> _filters;
 };
diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc
index b1f8aa3d0..f1d984ee1 100644
--- a/src/lib/ffmpeg_decoder.cc
+++ b/src/lib/ffmpeg_decoder.cc
@@ -242,11 +242,11 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size)
 AVSampleFormat
 FFmpegDecoder::audio_sample_format () const
 {
-	if (_audio_codec_context == 0) {
+	if (!_ffmpeg_content->audio_stream()) {
 		return (AVSampleFormat) 0;
 	}
 	
-	return _audio_codec_context->sample_fmt;
+	return audio_codec_context()->sample_fmt;
 }
 
 int
@@ -290,7 +290,7 @@ FFmpegDecoder::do_seek (Time t, bool backwards, bool accurate)
 	int64_t const vt = t / (av_q2d (_format_context->streams[_video_stream]->time_base) * TIME_HZ);
 	av_seek_frame (_format_context, _video_stream, vt, backwards ? AVSEEK_FLAG_BACKWARD : 0);
 
-	avcodec_flush_buffers (_video_codec_context);
+	avcodec_flush_buffers (video_codec_context());
 	if (_subtitle_codec_context) {
 		avcodec_flush_buffers (_subtitle_codec_context);
 	}
@@ -306,7 +306,7 @@ FFmpegDecoder::do_seek (Time t, bool backwards, bool accurate)
 			
 			if (_packet.stream_index == _video_stream) {
 				int finished = 0;
-				int const r = avcodec_decode_video2 (_video_codec_context, _frame, &finished, &_packet);
+				int const r = avcodec_decode_video2 (video_codec_context(), _frame, &finished, &_packet);
 				if (r >= 0 && finished) {
 					int64_t const bet = av_frame_get_best_effort_timestamp (_frame);
 					if (bet > vt) {
@@ -334,7 +334,7 @@ FFmpegDecoder::decode_audio_packet ()
 	while (copy_packet.size > 0) {
 
 		int frame_finished;
-		int const decode_result = avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &copy_packet);
+		int const decode_result = avcodec_decode_audio4 (audio_codec_context(), _frame, &frame_finished, &copy_packet);
 		if (decode_result >= 0) {
 			if (frame_finished) {
 			
@@ -343,10 +343,10 @@ FFmpegDecoder::decode_audio_packet ()
 					* av_frame_get_best_effort_timestamp(_frame);
 				
 				int const data_size = av_samples_get_buffer_size (
-					0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
+					0, audio_codec_context()->channels, _frame->nb_samples, audio_sample_format (), 1
 					);
 				
-				assert (_audio_codec_context->channels == _ffmpeg_content->audio_channels());
+				assert (audio_codec_context()->channels == _ffmpeg_content->audio_channels());
 				audio (deinterleave_audio (_frame->data, data_size), source_pts_seconds * TIME_HZ);
 			}
 			
@@ -360,7 +360,7 @@ bool
 FFmpegDecoder::decode_video_packet ()
 {
 	int frame_finished;
-	if (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) < 0 || !frame_finished) {
+	if (avcodec_decode_video2 (video_codec_context(), _frame, &frame_finished, &_packet) < 0 || !frame_finished) {
 		return false;
 	}
 		
@@ -416,11 +416,11 @@ FFmpegDecoder::decode_video_packet ()
 Time
 FFmpegDecoder::position () const
 {
-	if (_decode_video && _decode_audio && _audio_codec_context) {
+	if (_decode_video && _decode_audio && _ffmpeg_content->audio_stream()) {
 		return min (_next_video, _next_audio);
 	}
 
-	if (_decode_audio && _audio_codec_context) {
+	if (_decode_audio && _ffmpeg_content->audio_stream()) {
 		return _next_audio;
 	}
 
@@ -430,7 +430,9 @@ FFmpegDecoder::position () const
 bool
 FFmpegDecoder::done () const
 {
-	return (!_decode_audio || !_audio_codec_context || audio_done()) && (!_decode_video || video_done());
+	bool const ad = !_decode_audio || !_ffmpeg_content->audio_stream() || audio_done();
+	bool const vd = !_decode_video || video_done();
+	return ad && vd;
 }
 	
 void
diff --git a/src/lib/ffmpeg_examiner.cc b/src/lib/ffmpeg_examiner.cc
index e5d356a27..c09395e76 100644
--- a/src/lib/ffmpeg_examiner.cc
+++ b/src/lib/ffmpeg_examiner.cc
@@ -25,8 +25,10 @@ extern "C" {
 #include "ffmpeg_content.h"
 
 using std::string;
+using std::cout;
 using std::stringstream;
 using boost::shared_ptr;
+using boost::optional;
 
 FFmpegExaminer::FFmpegExaminer (shared_ptr<const FFmpegContent> c)
 	: FFmpeg (c)
@@ -56,6 +58,61 @@ FFmpegExaminer::FFmpegExaminer (shared_ptr<const FFmpegContent> c)
 		}
 	}
 
+	/* Run through until we find the first audio (for each stream) and video */
+
+	while (1) {
+		int r = av_read_frame (_format_context, &_packet);
+		if (r < 0) {
+			break;
+		}
+
+		int frame_finished;
+		avcodec_get_frame_defaults (_frame);
+
+		cout << "got packet " << _packet.stream_index << "\n";
+
+		AVCodecContext* context = _format_context->streams[_packet.stream_index]->codec;
+
+		if (_packet.stream_index == _video_stream && !_first_video) {
+			if (avcodec_decode_video2 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
+				_first_video = frame_time (_video_stream);
+			}
+		} else {
+			for (size_t i = 0; i < _audio_streams.size(); ++i) {
+				if (_packet.stream_index == _audio_streams[i]->id && !_audio_streams[i]->start) {
+					if (avcodec_decode_audio4 (context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
+						_audio_streams[i]->start = frame_time (_audio_streams[i]->id);
+					}
+				}
+			}
+		}
+
+		bool have_all_audio = true;
+		size_t i = 0;
+		while (i < _audio_streams.size() && have_all_audio) {
+			have_all_audio = _audio_streams[i]->start;
+			++i;
+		}
+
+		if (_first_video && have_all_audio) {
+			break;
+		}
+
+		av_free_packet (&_packet);
+	}
+}
+
+optional<Time>
+FFmpegExaminer::frame_time (int stream) const
+{
+	optional<Time> t;
+	
+	int64_t const bet = av_frame_get_best_effort_timestamp (_frame);
+	if (bet != AV_NOPTS_VALUE) {
+		t = bet * av_q2d (_format_context->streams[stream]->time_base) * TIME_HZ;
+	}
+
+	return t;
 }
 
 float
@@ -73,7 +130,7 @@ FFmpegExaminer::video_frame_rate () const
 libdcp::Size
 FFmpegExaminer::video_size () const
 {
-	return libdcp::Size (_video_codec_context->width, _video_codec_context->height);
+	return libdcp::Size (video_codec_context()->width, video_codec_context()->height);
 }
 
 /** @return Length (in video frames) according to our content's header */
diff --git a/src/lib/ffmpeg_examiner.h b/src/lib/ffmpeg_examiner.h
index 875451507..57b7775d4 100644
--- a/src/lib/ffmpeg_examiner.h
+++ b/src/lib/ffmpeg_examiner.h
@@ -17,6 +17,7 @@
 
 */
 
+#include <boost/optional.hpp>
 #include "ffmpeg.h"
 #include "video_examiner.h"
 
@@ -40,9 +41,15 @@ public:
 		return _audio_streams;
 	}
 
+	boost::optional<Time> first_video () const {
+		return _first_video;
+	}
+	
 private:
 	std::string stream_name (AVStream* s) const;
+	boost::optional<Time> frame_time (int) const;
 	
         std::vector<boost::shared_ptr<FFmpegSubtitleStream> > _subtitle_streams;
         std::vector<boost::shared_ptr<FFmpegAudioStream> > _audio_streams;
+	boost::optional<Time> _first_video;
 };