Various subtitle fixes.
[dcpomatic.git] / src / lib / decoder.cc
index 6bebe800c753b615e4a486769746979b4031fa69..44079edf7bf4d1f846e481f350c54f0b73525ef8 100644 (file)
 
 #include <iostream>
 #include <stdint.h>
+#include <boost/lexical_cast.hpp>
 extern "C" {
 #include <libavfilter/avfiltergraph.h>
 #include <libavfilter/buffersrc.h>
-#ifndef DVDOMATIC_FFMPEG_0_8_3
+#if (LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 77) || LIBAVFILTER_VERSION_MAJOR == 3
 #include <libavfilter/avcodec.h>
 #include <libavfilter/buffersink.h>
+#elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
+#include <libavfilter/vsrc_buffer.h>
 #endif
 #include <libavformat/avio.h>
 }
-#include <samplerate.h>
 #include "film.h"
 #include "format.h"
 #include "job.h"
@@ -68,7 +70,6 @@ Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const
        , _video_frame (0)
        , _buffer_src_context (0)
        , _buffer_sink_context (0)
-       , _swr_context (0)
        , _have_setup_video_filters (false)
        , _delay_line (0)
        , _delay_in_bytes (0)
@@ -84,26 +85,10 @@ Decoder::~Decoder ()
        delete _delay_line;
 }
 
+/** Start off a decode processing run */
 void
 Decoder::process_begin ()
 {
-       if (_fs->audio_sample_rate != dcp_audio_sample_rate (_fs->audio_sample_rate)) {
-               _swr_context = swr_alloc_set_opts (
-                       0,
-                       audio_channel_layout(),
-                       audio_sample_format(),
-                       dcp_audio_sample_rate (_fs->audio_sample_rate),
-                       audio_channel_layout(),
-                       audio_sample_format(),
-                       _fs->audio_sample_rate,
-                       0, 0
-                       );
-               
-               swr_init (_swr_context);
-       } else {
-               _swr_context = 0;
-       }
-
        _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
        delete _delay_line;
        _delay_line = new DelayLine (_delay_in_bytes);
@@ -111,42 +96,14 @@ Decoder::process_begin ()
        _audio_frames_processed = 0;
 }
 
+/** Finish off a decode processing run */
 void
 Decoder::process_end ()
 {
-       if (_swr_context) {
-
-               int mop = 0;
-               while (1) {
-                       uint8_t buffer[256 * _fs->bytes_per_sample() * _fs->audio_channels];
-                       uint8_t* out[1] = {
-                               buffer
-                       };
-
-                       int const frames = swr_convert (_swr_context, out, 256, 0, 0);
-
-                       if (frames < 0) {
-                               throw DecodeError ("could not run sample-rate converter");
-                       }
-
-                       if (frames == 0) {
-                               break;
-                       }
-
-                       mop += frames;
-                       int available = _delay_line->feed (buffer, frames * _fs->audio_channels * _fs->bytes_per_sample());
-                       Audio (buffer, available);
-               }
-
-               cout << "mopped up " << mop << "\n";
-               
-               swr_free (&_swr_context);
-       }
-       
        if (_delay_in_bytes < 0) {
                uint8_t remainder[-_delay_in_bytes];
                _delay_line->get_remaining (remainder);
-               _audio_frames_processed += _delay_in_bytes / (audio_channels() * _fs->bytes_per_sample());
+               _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
                Audio (remainder, _delay_in_bytes);
        }
 
@@ -154,18 +111,27 @@ Decoder::process_end ()
           in to get it to the right length.
        */
 
-       int const audio_short_by_frames = (decoding_frames() * dcp_audio_sample_rate (_fs->audio_sample_rate) / _fs->frames_per_second) - _audio_frames_processed;
+       int64_t const audio_short_by_frames =
+               ((int64_t) decoding_frames() * _fs->target_sample_rate() / _fs->frames_per_second)
+               - _audio_frames_processed;
 
-       int bytes = audio_short_by_frames * audio_channels() * _fs->bytes_per_sample();
+       if (audio_short_by_frames >= 0) {
 
-       int const silence_size = 64 * 1024;
-       uint8_t silence[silence_size];
-       memset (silence, 0, silence_size);
+               stringstream s;
+               s << "Adding " << audio_short_by_frames << " frames of silence to the end.";
+               _log->log (s.str ());
 
-       while (bytes) {
-               int const t = min (bytes, silence_size);
-               Audio (silence, t);
-               bytes -= t;
+               int64_t bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
+               
+               int64_t const silence_size = 64 * 1024;
+               uint8_t silence[silence_size];
+               memset (silence, 0, silence_size);
+               
+               while (bytes) {
+                       int64_t const t = min (bytes, silence_size);
+                       Audio (silence, t);
+                       bytes -= t;
+               }
        }
 }
 
@@ -218,13 +184,17 @@ Decoder::pass ()
        return do_pass ();
 }
 
-/** Called by subclasses to tell the world that some audio data is ready */
+/** Called by subclasses to tell the world that some audio data is ready
+ *  @param data Interleaved audio data, in FilmState::audio_sample_format.
+ *  @param size Number of bytes of data.
+ */
 void
 Decoder::process_audio (uint8_t* data, int size)
 {
+       /* Samples per channel */
        int const samples = size / _fs->bytes_per_sample();
-       int const frames = samples / _fs->audio_channels;
-       
+
+       /* Maybe apply gain */
        if (_fs->audio_gain != 0) {
                float const linear_gain = pow (10, _fs->audio_gain / 20);
                uint8_t* p = data;
@@ -232,10 +202,20 @@ Decoder::process_audio (uint8_t* data, int size)
                case AV_SAMPLE_FMT_S16:
                        for (int i = 0; i < samples; ++i) {
                                /* XXX: assumes little-endian; also we should probably be dithering here */
+
+                               /* unsigned sample */
                                int const ou = p[0] | (p[1] << 8);
+
+                               /* signed sample */
                                int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
+
+                               /* signed sample with altered gain */
                                int const gs = int (os * linear_gain);
+
+                               /* unsigned sample with altered gain */
                                int const gu = gs > 0 ? gs : (0x10000 + gs);
+
+                               /* write it back */
                                p[0] = gu & 0xff;
                                p[1] = (gu & 0xff00) >> 8;
                                p += 2;
@@ -246,40 +226,12 @@ Decoder::process_audio (uint8_t* data, int size)
                }
        }
 
-       uint8_t* out_buffer = 0;
-
-       if (_swr_context) {
-
-               uint8_t const * in[2] = {
-                       data,
-                       0
-               };
-
-               int const out_buffer_size_frames = ceil (frames * float (dcp_audio_sample_rate (_fs->audio_sample_rate)) / _fs->audio_sample_rate) + 32;
-               int const out_buffer_size_bytes = out_buffer_size_frames * _fs->audio_channels * _fs->bytes_per_sample();
-               out_buffer = new uint8_t[out_buffer_size_bytes];
-
-               uint8_t* out[2] = {
-                       out_buffer, 
-                       0
-               };
-               
-               int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
-               if (out_frames < 0) {
-                       throw DecodeError ("could not run sample-rate converter");
-               }
-
-               data = out_buffer;
-               size = out_frames * _fs->audio_channels * _fs->bytes_per_sample();
-       }
-               
        /* Update the number of audio frames we've pushed to the encoder */
        _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
-       
+
+       /* Push into the delay line and then tell the world what we've got */
        int available = _delay_line->feed (data, size);
        Audio (data, available);
-
-       delete[] out_buffer;
 }
 
 /** Called by subclasses to tell the world that some video data is ready.
@@ -306,8 +258,14 @@ Decoder::process_video (AVFrame* frame)
                return;
        }
 
-#ifdef DVDOMATIC_FFMPEG_0_8_3
-       
+#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
+
+       if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
+               throw DecodeError ("could not push buffer into filter chain.");
+       }
+
+#elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
+
        AVRational par;
        par.num = sample_aspect_ratio_numerator ();
        par.den = sample_aspect_ratio_denominator ();
@@ -324,13 +282,13 @@ Decoder::process_video (AVFrame* frame)
 
 #endif 
        
-#ifdef DVDOMATIC_FFMPEG_0_8_3
+#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61       
        while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
 #else
        while (av_buffersink_read (_buffer_sink_context, 0)) {
 #endif         
 
-#ifdef DVDOMATIC_FFMPEG_0_8_3
+#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
                
                int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
                if (r < 0) {
@@ -356,12 +314,19 @@ Decoder::process_video (AVFrame* frame)
                                image->make_black ();
                        }
 
+                       overlay (image);
+
+                       TIMING ("Decoder emits %1", _video_frame);
                        Video (image, _video_frame);
                        ++_video_frame;
                }
        }
 }
 
+
+/** Set up a video filtering chain to include cropping and any filters that are specified
+ *  by the Film.
+ */
 void
 Decoder::setup_video_filters ()
 {
@@ -370,7 +335,7 @@ Decoder::setup_video_filters ()
        
        if (_opt->apply_crop) {
                size_after_crop = _fs->cropped_size (native_size ());
-               fs << crop_string (Position (_fs->left_crop, _fs->top_crop), size_after_crop);
+               fs << crop_string (Position (_fs->crop.left, _fs->crop.top), size_after_crop);
        } else {
                size_after_crop = native_size ();
                fs << crop_string (Position (0, 0), size_after_crop);
@@ -407,12 +372,18 @@ Decoder::setup_video_filters ()
          << sample_aspect_ratio_denominator();
 
        int r;
+
        if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
                throw DecodeError ("could not create buffer source");
        }
 
-       enum PixelFormat pixel_formats[] = { pixel_format(), PIX_FMT_NONE };
-       if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, pixel_formats, graph) < 0) {
+       AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
+       PixelFormat* pixel_fmts = new PixelFormat[2];
+       pixel_fmts[0] = pixel_format ();
+       pixel_fmts[1] = PIX_FMT_NONE;
+       sink_params->pixel_fmts = pixel_fmts;
+       
+       if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
                throw DecodeError ("could not create buffer sink.");
        }
 
@@ -429,15 +400,17 @@ Decoder::setup_video_filters ()
        inputs->next = 0;
 
        _log->log ("Using filter chain `" + filters + "'");
-#ifdef DVDOMATIC_FFMPEG_0_8_3  
+
+#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
        if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
-#else
+               throw DecodeError ("could not set up filter graph.");
+       }
+#else  
        if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
-#endif         
-               
                throw DecodeError ("could not set up filter graph.");
        }
-
+#endif 
+       
        if (avfilter_graph_config (graph, 0) < 0) {
                throw DecodeError ("could not configure filter graph.");
        }