Merge branch 'master' into trimming
[dcpomatic.git] / src / lib / ffmpeg_decoder.cc
index 1f11f70a0a462d9d5ac265c8b6e3961879bd46c7..ac25844e34687032ea6c47c1808a0672704b9f44 100644 (file)
@@ -50,6 +50,8 @@ extern "C" {
 #include "filter_graph.h"
 #include "subtitle.h"
 
+#include "i18n.h"
+
 using std::cout;
 using std::string;
 using std::vector;
@@ -113,7 +115,7 @@ FFmpegDecoder::setup_general ()
        }
 
        if (avformat_find_stream_info (_format_context, 0) < 0) {
-               throw DecodeError ("could not find stream information");
+               throw DecodeError (_("could not find stream information"));
        }
 
        /* Find video, audio and subtitle streams and choose the first of each */
@@ -127,7 +129,7 @@ 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);
                        }
@@ -148,12 +150,12 @@ FFmpegDecoder::setup_general ()
        }
 
        if (_video_stream < 0) {
-               throw DecodeError ("could not find video stream");
+               throw DecodeError (N_("could not find video stream"));
        }
 
        _frame = avcodec_alloc_frame ();
        if (_frame == 0) {
-               throw DecodeError ("could not allocate frame");
+               throw DecodeError (N_("could not allocate frame"));
        }
 }
 
@@ -164,11 +166,11 @@ FFmpegDecoder::setup_video ()
        _video_codec = avcodec_find_decoder (_video_codec_context->codec_id);
 
        if (_video_codec == 0) {
-               throw DecodeError ("could not find video decoder");
+               throw DecodeError (_("could not find video decoder"));
        }
 
        if (avcodec_open2 (_video_codec_context, _video_codec, 0) < 0) {
-               throw DecodeError ("could not open video decoder");
+               throw DecodeError (N_("could not open video decoder"));
        }
 }
 
@@ -186,18 +188,18 @@ FFmpegDecoder::setup_audio ()
        _audio_codec = avcodec_find_decoder (_audio_codec_context->codec_id);
 
        if (_audio_codec == 0) {
-               throw DecodeError ("could not find audio decoder");
+               throw DecodeError (_("could not find audio decoder"));
        }
 
        if (avcodec_open2 (_audio_codec_context, _audio_codec, 0) < 0) {
-               throw DecodeError ("could not open audio decoder");
+               throw DecodeError (N_("could not open audio decoder"));
        }
 }
 
 void
 FFmpegDecoder::setup_subtitle ()
 {
-       if (!_subtitle_stream) {
+       if (!_subtitle_stream || _subtitle_stream->id() >= int (_format_context->nb_streams)) {
                return;
        }
 
@@ -205,11 +207,11 @@ FFmpegDecoder::setup_subtitle ()
        _subtitle_codec = avcodec_find_decoder (_subtitle_codec_context->codec_id);
 
        if (_subtitle_codec == 0) {
-               throw DecodeError ("could not find subtitle decoder");
+               throw DecodeError (_("could not find subtitle decoder"));
        }
        
        if (avcodec_open2 (_subtitle_codec_context, _subtitle_codec, 0) < 0) {
-               throw DecodeError ("could not open subtitle decoder");
+               throw DecodeError (N_("could not open subtitle decoder"));
        }
 }
 
@@ -224,7 +226,7 @@ FFmpegDecoder::pass ()
                        /* Maybe we should fail here, but for now we'll just finish off instead */
                        char buf[256];
                        av_strerror (r, buf, sizeof(buf));
-                       _film->log()->log (String::compose ("error on av_read_frame (%1) (%2)", buf, r));
+                       _film->log()->log (String::compose (N_("error on av_read_frame (%1) (%2)"), buf, r));
                }
                
                /* Get any remaining frames */
@@ -236,19 +238,14 @@ 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) {
-                       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));
-                       }
+                       decode_audio_packet ();
                }
 
                return true;
@@ -258,14 +255,14 @@ 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);
                if (r >= 0 && frame_finished) {
 
                        if (r != _packet.size) {
-                               _film->log()->log (String::compose ("Used only %1 bytes of %2 in packet", r, _packet.size));
+                               _film->log()->log (String::compose (N_("Used only %1 bytes of %2 in packet"), r, _packet.size));
                        }
 
                        if (_opt.video_sync) {
@@ -276,54 +273,7 @@ FFmpegDecoder::pass ()
                }
 
        } 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) {
-
-                       /* Where we are in the source, in seconds */
-                       double const source_pts_seconds = av_q2d (_format_context->streams[_packet.stream_index]->time_base)
-                               * av_frame_get_best_effort_timestamp(_frame);
-
-                       /* We only decode audio if we've had our first video packet through, and if it
-                          was before this packet.  Until then audio is thrown away.
-                       */
-                               
-                       if (_first_video && _first_video.get() <= source_pts_seconds) {
-
-                               if (!_first_audio) {
-                                       _first_audio = source_pts_seconds;
-                                       
-                                       /* This is our first audio frame, and if we've arrived here we must have had our
-                                          first video frame.  Push some silence to make up any gap between our first
-                                          video frame and our first audio.
-                                       */
-                       
-                                       /* frames of silence that we must push */
-                                       int const s = rint ((_first_audio.get() - _first_video.get()) * ffa->sample_rate ());
-                                       
-                                       _film->log()->log (
-                                               String::compose (
-                                                       "First video at %1, first audio at %2, pushing %3 audio frames of silence for %4 channels (%5 bytes per sample)",
-                                                       _first_video.get(), _first_audio.get(), s, ffa->channels(), bytes_per_audio_sample()
-                                                       )
-                                               );
-                                       
-                                       if (s) {
-                                               shared_ptr<AudioBuffers> audio (new AudioBuffers (ffa->channels(), s));
-                                               audio->make_silent ();
-                                               Audio (audio);
-                                       }
-                               }
-
-                               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));
-                       }
-               }
-                       
+               decode_audio_packet ();
        } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt.decode_subtitles && _first_video) {
 
                int got_subtitle;
@@ -350,8 +300,11 @@ 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());
@@ -370,7 +323,7 @@ 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) {
@@ -387,10 +340,10 @@ 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);
                        }
                }
        }
@@ -398,7 +351,7 @@ 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) {
@@ -415,7 +368,7 @@ 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) {
@@ -432,16 +385,15 @@ 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;
 
        default:
-               throw DecodeError (String::compose ("Unrecognised audio sample format (%1)", static_cast<int> (audio_sample_format())));
+               throw DecodeError (String::compose (_("Unrecognised audio sample format (%1)"), static_cast<int> (audio_sample_format())));
        }
 
        return audio;
@@ -510,21 +462,21 @@ FFmpegDecoder::stream_name (AVStream* s) const
 {
        stringstream n;
        
-       AVDictionaryEntry const * lang = av_dict_get (s->metadata, "language", 0, 0);
+       AVDictionaryEntry const * lang = av_dict_get (s->metadata, N_("language"), 0, 0);
        if (lang) {
                n << lang->value;
        }
        
-       AVDictionaryEntry const * title = av_dict_get (s->metadata, "title", 0, 0);
+       AVDictionaryEntry const * title = av_dict_get (s->metadata, N_("title"), 0, 0);
        if (title) {
                if (!n.str().empty()) {
-                       n << " ";
+                       n << N_(" ");
                }
                n << title->value;
        }
 
        if (n.str().empty()) {
-               n << "unknown";
+               n << N_("unknown");
        }
 
        return n.str ();
@@ -566,7 +518,7 @@ FFmpegDecoder::filter_and_emit_video (AVFrame* frame)
        if (i == _filter_graphs.end ()) {
                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));
+               _film->log()->log (String::compose (N_("New graph for %1x%2, pixel format %3"), frame->width, frame->height, frame->format));
        } else {
                graph = *i;
        }
@@ -620,7 +572,7 @@ FFmpegAudioStream::create (string t, optional<int> v)
        stringstream s (t);
        string type;
        s >> type;
-       if (type != "ffmpeg") {
+       if (type != N_("ffmpeg")) {
                return shared_ptr<FFmpegAudioStream> ();
        }
 
@@ -642,7 +594,7 @@ FFmpegAudioStream::FFmpegAudioStream (string t, optional<int> version)
                string type;
                /* Current (marked version 1) */
                n >> type >> _id >> _sample_rate >> _channel_layout;
-               assert (type == "ffmpeg");
+               assert (type == N_("ffmpeg"));
        }
 
        for (int i = 0; i < name_index; ++i) {
@@ -658,7 +610,7 @@ FFmpegAudioStream::FFmpegAudioStream (string t, optional<int> version)
 string
 FFmpegAudioStream::to_string () const
 {
-       return String::compose ("ffmpeg %1 %2 %3 %4", _id, _sample_rate, _channel_layout, _name);
+       return String::compose (N_("ffmpeg %1 %2 %3 %4"), _id, _sample_rate, _channel_layout, _name);
 }
 
 void
@@ -672,7 +624,7 @@ FFmpegDecoder::out_with_sync ()
                * av_frame_get_best_effort_timestamp(_frame);
        
        _film->log()->log (
-               String::compose ("Source video frame ready; source at %1, output at %2", source_pts_seconds, out_pts_seconds),
+               String::compose (N_("Source video frame ready; source at %1, output at %2"), source_pts_seconds, out_pts_seconds),
                Log::VERBOSE
                );
        
@@ -691,7 +643,7 @@ FFmpegDecoder::out_with_sync ()
                        repeat_last_video ();
                        _film->log()->log (
                                String::compose (
-                                       "Extra video frame inserted at %1s; source frame %2, source PTS %3 (at %4 fps)",
+                                       N_("Extra video frame inserted at %1s; source frame %2, source PTS %3 (at %4 fps)"),
                                        out_pts_seconds, video_frame(), source_pts_seconds, frames_per_second()
                                        )
                                );
@@ -703,7 +655,7 @@ FFmpegDecoder::out_with_sync ()
                filter_and_emit_video (_frame);
        } else {
                /* Otherwise we are omitting a frame to keep things right */
-               _film->log()->log (String::compose ("Frame removed at %1s", out_pts_seconds));
+               _film->log()->log (String::compose (N_("Frame removed at %1s"), out_pts_seconds));
        }
 }
 
@@ -738,3 +690,71 @@ FFmpegDecoder::frame_time () const
        return av_frame_get_best_effort_timestamp(_frame) * av_q2d (_format_context->streams[_video_stream]->time_base);
 }
 
+void
+FFmpegDecoder::decode_audio_packet ()
+{
+       shared_ptr<FFmpegAudioStream> ffa = dynamic_pointer_cast<FFmpegAudioStream> (_audio_stream);
+       assert (ffa);
+
+       /* Audio packets can contain multiple frames, so we may have to call avcodec_decode_audio4
+          several times.
+       */
+       
+       AVPacket copy_packet = _packet;
+
+       while (copy_packet.size > 0) {
+
+               int frame_finished;
+               int const decode_result = avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &copy_packet);
+               if (decode_result >= 0 && frame_finished) {
+                       
+                       /* Where we are in the source, in seconds */
+                       double const source_pts_seconds = av_q2d (_format_context->streams[copy_packet.stream_index]->time_base)
+                               * av_frame_get_best_effort_timestamp(_frame);
+                       
+                       /* We only decode audio if we've had our first video packet through, and if it
+                          was before this packet.  Until then audio is thrown away.
+                       */
+                       
+                       if ((_first_video && _first_video.get() <= source_pts_seconds) || !_opt.decode_video) {
+                               
+                               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
+                                          first video frame.  Push some silence to make up any gap between our first
+                                          video frame and our first audio.
+                                       */
+                                       
+                                       /* frames of silence that we must push */
+                                       int const s = rint ((_first_audio.get() - _first_video.get()) * ffa->sample_rate ());
+                                       
+                                       _film->log()->log (
+                                               String::compose (
+                                                       N_("First video at %1, first audio at %2, pushing %3 audio frames of silence for %4 channels (%5 bytes per sample)"),
+                                                       _first_video.get(), _first_audio.get(), s, ffa->channels(), bytes_per_audio_sample()
+                                                       )
+                                               );
+                                       
+                                       if (s) {
+                                               shared_ptr<AudioBuffers> audio (new AudioBuffers (ffa->channels(), s));
+                                               audio->make_silent ();
+                                               Audio (audio);
+                                       }
+                               }
+                               
+                               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, data_size));
+                       }
+               }
+
+               if (decode_result >= 0) {
+                       copy_packet.data += decode_result;
+                       copy_packet.size -= decode_result;
+               }
+       }
+}