Merge branch 'master' into plot-audio
[dcpomatic.git] / src / lib / ffmpeg_decoder.cc
index a19f26ad79bdfde7dc66e3d795d87254e5796d43..148764162c20fe70bfdd9d4be2e7ed96faaf8d57 100644 (file)
@@ -58,11 +58,12 @@ using std::list;
 using boost::shared_ptr;
 using boost::optional;
 using boost::dynamic_pointer_cast;
+using libdcp::Size;
 
-FFmpegDecoder::FFmpegDecoder (shared_ptr<Film> f, shared_ptr<const DecodeOptions> o, Job* j)
-       : Decoder (f, o, j)
-       , VideoDecoder (f, o, j)
-       , AudioDecoder (f, o, j)
+FFmpegDecoder::FFmpegDecoder (shared_ptr<Film> f, DecodeOptions o)
+       : Decoder (f, o)
+       , VideoDecoder (f, o)
+       , AudioDecoder (f, o)
        , _format_context (0)
        , _video_stream (-1)
        , _frame (0)
@@ -78,7 +79,7 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<Film> f, shared_ptr<const DecodeOptions
        setup_audio ();
        setup_subtitle ();
 
-       if (!o->video_sync) {
+       if (!o.video_sync) {
                _first_video = 0;
        }
 }
@@ -126,7 +127,7 @@ FFmpegDecoder::setup_general ()
                        /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up,
                           so bodge it here.  No idea why we should have to do this.
                        */
-                       
+
                        if (s->codec->channel_layout == 0) {
                                s->codec->channel_layout = av_get_default_channel_layout (s->codec->channels);
                        }
@@ -196,7 +197,7 @@ FFmpegDecoder::setup_audio ()
 void
 FFmpegDecoder::setup_subtitle ()
 {
-       if (!_subtitle_stream) {
+       if (!_subtitle_stream || _subtitle_stream->id() >= _format_context->nb_streams) {
                return;
        }
 
@@ -235,18 +236,20 @@ FFmpegDecoder::pass ()
 
                int frame_finished;
 
-               while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
-                       filter_and_emit_video (_frame);
+               if (_opt.decode_video) {
+                       while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
+                               filter_and_emit_video (_frame);
+                       }
                }
 
-               if (_audio_stream && _opt->decode_audio) {
+               if (_audio_stream && _opt.decode_audio) {
                        while (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
                                int const data_size = av_samples_get_buffer_size (
                                        0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1
                                        );
 
                                assert (_audio_codec_context->channels == _film->audio_channels());
-                               Audio (deinterleave_audio (_frame->data[0], data_size));
+                               Audio (deinterleave_audio (_frame->data, data_size));
                        }
                }
 
@@ -257,7 +260,7 @@ FFmpegDecoder::pass ()
 
        shared_ptr<FFmpegAudioStream> ffa = dynamic_pointer_cast<FFmpegAudioStream> (_audio_stream);
 
-       if (_packet.stream_index == _video_stream) {
+       if (_packet.stream_index == _video_stream && _opt.decode_video) {
 
                int frame_finished;
                int const r = avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet);
@@ -267,14 +270,14 @@ FFmpegDecoder::pass ()
                                _film->log()->log (String::compose ("Used only %1 bytes of %2 in packet", r, _packet.size));
                        }
 
-                       if (_opt->video_sync) {
+                       if (_opt.video_sync) {
                                out_with_sync ();
                        } else {
                                filter_and_emit_video (_frame);
                        }
                }
 
-       } else if (ffa && _packet.stream_index == ffa->id() && _opt->decode_audio) {
+       } else if (ffa && _packet.stream_index == ffa->id() && _opt.decode_audio) {
 
                int frame_finished;
                if (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) {
@@ -287,9 +290,9 @@ FFmpegDecoder::pass ()
                           was before this packet.  Until then audio is thrown away.
                        */
                                
-                       if (_first_video && _first_video.get() <= source_pts_seconds) {
+                       if ((_first_video && _first_video.get() <= source_pts_seconds) || !_opt.decode_video) {
 
-                               if (!_first_audio) {
+                               if (!_first_audio && _opt.decode_video) {
                                        _first_audio = source_pts_seconds;
                                        
                                        /* This is our first audio frame, and if we've arrived here we must have had our
@@ -319,11 +322,11 @@ FFmpegDecoder::pass ()
                                        );
                                
                                assert (_audio_codec_context->channels == _film->audio_channels());
-                               Audio (deinterleave_audio (_frame->data[0], data_size));
+                               Audio (deinterleave_audio (_frame->data, data_size));
                        }
                }
                        
-       } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt->decode_subtitles && _first_video) {
+       } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt.decode_subtitles && _first_video) {
 
                int got_subtitle;
                AVSubtitle sub;
@@ -349,8 +352,11 @@ FFmpegDecoder::pass ()
        return false;
 }
 
+/** @param data pointer to array of pointers to buffers.
+ *  Only the first buffer will be used for non-planar data, otherwise there will be one per channel.
+ */
 shared_ptr<AudioBuffers>
-FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
+FFmpegDecoder::deinterleave_audio (uint8_t** data, int size)
 {
        assert (_film->audio_channels());
        assert (bytes_per_audio_sample());
@@ -369,7 +375,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
        switch (audio_sample_format()) {
        case AV_SAMPLE_FMT_S16:
        {
-               int16_t* p = reinterpret_cast<int16_t *> (data);
+               int16_t* p = reinterpret_cast<int16_t *> (data[0]);
                int sample = 0;
                int channel = 0;
                for (int i = 0; i < total_samples; ++i) {
@@ -386,10 +392,10 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
 
        case AV_SAMPLE_FMT_S16P:
        {
-               int16_t* p = reinterpret_cast<int16_t *> (data);
+               int16_t** p = reinterpret_cast<int16_t **> (data);
                for (int i = 0; i < _film->audio_channels(); ++i) {
                        for (int j = 0; j < frames; ++j) {
-                               audio->data(i)[j] = static_cast<float>(*p++) / (1 << 15);
+                               audio->data(i)[j] = static_cast<float>(p[i][j]) / (1 << 15);
                        }
                }
        }
@@ -397,7 +403,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
        
        case AV_SAMPLE_FMT_S32:
        {
-               int32_t* p = reinterpret_cast<int32_t *> (data);
+               int32_t* p = reinterpret_cast<int32_t *> (data[0]);
                int sample = 0;
                int channel = 0;
                for (int i = 0; i < total_samples; ++i) {
@@ -414,7 +420,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
 
        case AV_SAMPLE_FMT_FLT:
        {
-               float* p = reinterpret_cast<float*> (data);
+               float* p = reinterpret_cast<float*> (data[0]);
                int sample = 0;
                int channel = 0;
                for (int i = 0; i < total_samples; ++i) {
@@ -431,10 +437,9 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size)
                
        case AV_SAMPLE_FMT_FLTP:
        {
-               float* p = reinterpret_cast<float*> (data);
+               float** p = reinterpret_cast<float**> (data);
                for (int i = 0; i < _film->audio_channels(); ++i) {
-                       memcpy (audio->data(i), p, frames * sizeof(float));
-                       p += frames;
+                       memcpy (audio->data(i), p[i], frames * sizeof(float));
                }
        }
        break;
@@ -468,10 +473,10 @@ FFmpegDecoder::audio_sample_format () const
        return _audio_codec_context->sample_fmt;
 }
 
-Size
+libdcp::Size
 FFmpegDecoder::native_size () const
 {
-       return Size (_video_codec_context->width, _video_codec_context->height);
+       return libdcp::Size (_video_codec_context->width, _video_codec_context->height);
 }
 
 PixelFormat
@@ -558,12 +563,12 @@ FFmpegDecoder::filter_and_emit_video (AVFrame* frame)
        shared_ptr<FilterGraph> graph;
 
        list<shared_ptr<FilterGraph> >::iterator i = _filter_graphs.begin();
-       while (i != _filter_graphs.end() && !(*i)->can_process (Size (frame->width, frame->height), (AVPixelFormat) frame->format)) {
+       while (i != _filter_graphs.end() && !(*i)->can_process (libdcp::Size (frame->width, frame->height), (AVPixelFormat) frame->format)) {
                ++i;
        }
 
        if (i == _filter_graphs.end ()) {
-               graph.reset (new FilterGraph (_film, this, Size (frame->width, frame->height), (AVPixelFormat) frame->format));
+               graph.reset (new FilterGraph (_film, this, libdcp::Size (frame->width, frame->height), (AVPixelFormat) frame->format));
                _filter_graphs.push_back (graph);
                _film->log()->log (String::compose ("New graph for %1x%2, pixel format %3", frame->width, frame->height, frame->format));
        } else {