Make sure at least one position change event is emitted after
[dcpomatic.git] / src / lib / ffmpeg_decoder.cc
index 55ff01046f761ec61ff73990d142ec9720db6620..ce20997c84bcebefdc20dd2f90329155545cc505 100644 (file)
@@ -1,5 +1,5 @@
 /*
-    Copyright (C) 2012-2016 Carl Hetherington <cth@carlh.net>
+    Copyright (C) 2012-2018 Carl Hetherington <cth@carlh.net>
 
     This file is part of DCP-o-matic.
 
@@ -27,8 +27,9 @@
 #include "image.h"
 #include "util.h"
 #include "log.h"
+#include "dcpomatic_log.h"
 #include "ffmpeg_decoder.h"
-#include "subtitle_decoder.h"
+#include "text_decoder.h"
 #include "ffmpeg_audio_stream.h"
 #include "ffmpeg_subtitle_stream.h"
 #include "video_filter_graph.h"
@@ -39,7 +40,7 @@
 #include "film.h"
 #include "audio_decoder.h"
 #include "compose.hpp"
-#include "subtitle_content.h"
+#include "text_content.h"
 #include "audio_content.h"
 #include <dcp/subtitle_string.h>
 #include <sub/ssa_reader.h>
@@ -58,10 +59,6 @@ extern "C" {
 
 #include "i18n.h"
 
-#define LOG_GENERAL(...) _log->log (String::compose (__VA_ARGS__), LogEntry::TYPE_GENERAL);
-#define LOG_ERROR(...) _log->log (String::compose (__VA_ARGS__), LogEntry::TYPE_ERROR);
-#define LOG_WARNING_NC(...) _log->log (__VA_ARGS__, LogEntry::TYPE_WARNING);
-#define LOG_WARNING(...) _log->log (String::compose (__VA_ARGS__), LogEntry::TYPE_WARNING);
 
 using std::cout;
 using std::string;
@@ -78,14 +75,14 @@ using boost::optional;
 using boost::dynamic_pointer_cast;
 using dcp::Size;
 
-FFmpegDecoder::FFmpegDecoder (shared_ptr<const FFmpegContent> c, shared_ptr<Log> log, bool fast)
+FFmpegDecoder::FFmpegDecoder (shared_ptr<const Film> film, shared_ptr<const FFmpegContent> c, bool fast)
        : FFmpeg (c)
-       , _log (log)
+       , Decoder (film)
        , _have_current_subtitle (false)
 {
        if (c->video) {
-               video.reset (new VideoDecoder (this, c, log));
-               _pts_offset = pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->active_video_frame_rate());
+               video.reset (new VideoDecoder (this, c));
+               _pts_offset = pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->active_video_frame_rate(film));
                /* It doesn't matter what size or pixel format this is, it just needs to be black */
                _black_image.reset (new Image (AV_PIX_FMT_RGB24, dcp::Size (128, 128), true));
                _black_image->make_black ();
@@ -94,12 +91,12 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const FFmpegContent> c, shared_ptr<Log>
        }
 
        if (c->audio) {
-               audio.reset (new AudioDecoder (this, c->audio, log, fast));
+               audio.reset (new AudioDecoder (this, c->audio, fast));
        }
 
-       if (c->subtitle) {
+       if (c->only_text()) {
                /* XXX: this time here should be the time of the first subtitle, not 0 */
-               subtitle.reset (new SubtitleDecoder (this, c->subtitle, log, ContentTime()));
+               text.push_back (shared_ptr<TextDecoder> (new TextDecoder (this, c->only_text(), ContentTime())));
        }
 
        _next_time.resize (_format_context->nb_streams);
@@ -115,7 +112,7 @@ FFmpegDecoder::flush ()
 
        /* XXX: should we reset _packet.data and size after each *_decode_* call? */
 
-       while (video && decode_video_packet ()) {}
+       while (video && decode_video_packet()) {}
 
        if (audio) {
                decode_audio_packet ();
@@ -123,21 +120,21 @@ FFmpegDecoder::flush ()
 
        /* Make sure all streams are the same length and round up to the next video frame */
 
-       FrameRateChange const frc = _ffmpeg_content->film()->active_frame_rate_change(_ffmpeg_content->position());
-       ContentTime full_length (_ffmpeg_content->full_length(), frc);
+       FrameRateChange const frc = film()->active_frame_rate_change(_ffmpeg_content->position());
+       ContentTime full_length (_ffmpeg_content->full_length(film()), frc);
        full_length = full_length.ceil (frc.source);
        if (video) {
                double const vfr = _ffmpeg_content->video_frame_rate().get();
                Frame const f = full_length.frames_round (vfr);
-               Frame v = video->position().frames_round (vfr) + 1;
+               Frame v = video->position(film()).frames_round(vfr) + 1;
                while (v < f) {
-                       video->emit (shared_ptr<const ImageProxy> (new RawImageProxy (_black_image)), v);
+                       video->emit (film(), shared_ptr<const ImageProxy> (new RawImageProxy (_black_image)), v);
                        ++v;
                }
        }
 
        BOOST_FOREACH (shared_ptr<FFmpegAudioStream> i, _ffmpeg_content->ffmpeg_audio_streams ()) {
-               ContentTime a = audio->stream_position(i);
+               ContentTime a = audio->stream_position(film(), i);
                /* Unfortunately if a is 0 that really means that we don't know the stream position since
                   there has been no data on it since the last seek.  In this case we'll just do nothing
                   here.  I'm not sure if that's the right idea.
@@ -147,7 +144,7 @@ FFmpegDecoder::flush ()
                                ContentTime to_do = min (full_length - a, ContentTime::from_seconds (0.1));
                                shared_ptr<AudioBuffers> silence (new AudioBuffers (i->channels(), to_do.frames_ceil (i->frame_rate())));
                                silence->make_silent ();
-                               audio->emit (i, silence, a);
+                               audio->emit (film(), i, silence, a);
                                a += to_do;
                        }
                }
@@ -184,7 +181,7 @@ FFmpegDecoder::pass ()
 
        if (_video_stream && si == _video_stream.get() && !video->ignore()) {
                decode_video_packet ();
-       } else if (fc->subtitle_stream() && fc->subtitle_stream()->uses_index(_format_context, si) && !subtitle->ignore()) {
+       } else if (fc->subtitle_stream() && fc->subtitle_stream()->uses_index(_format_context, si) && !only_text()->ignore()) {
                decode_subtitle_packet ();
        } else {
                decode_audio_packet ();
@@ -479,7 +476,7 @@ FFmpegDecoder::decode_audio_packet ()
 
                        /* Give this data provided there is some, and its time is sane */
                        if (ct >= ContentTime() && data->frames() > 0) {
-                               audio->emit (*stream, data, ct);
+                               audio->emit (film(), *stream, data, ct);
                        }
                }
 
@@ -525,9 +522,11 @@ FFmpegDecoder::decode_video_packet ()
 
                if (i->second != AV_NOPTS_VALUE) {
                        double const pts = i->second * av_q2d (_format_context->streams[_video_stream.get()]->time_base) + _pts_offset.seconds ();
+
                        video->emit (
+                               film(),
                                shared_ptr<ImageProxy> (new RawImageProxy (image)),
-                               llrint (pts * _ffmpeg_content->active_video_frame_rate ())
+                               llrint(pts * _ffmpeg_content->active_video_frame_rate(film()))
                                );
                } else {
                        LOG_WARNING_NC ("Dropping frame without PTS");
@@ -549,9 +548,9 @@ FFmpegDecoder::decode_subtitle_packet ()
        /* Stop any current subtitle, either at the time it was supposed to stop, or now if now is sooner */
        if (_have_current_subtitle) {
                if (_current_subtitle_to) {
-                       subtitle->emit_stop (min(*_current_subtitle_to, subtitle_period(sub).from + _pts_offset));
+                       only_text()->emit_stop (min(*_current_subtitle_to, subtitle_period(sub).from + _pts_offset));
                } else {
-                       subtitle->emit_stop (subtitle_period(sub).from + _pts_offset);
+                       only_text()->emit_stop (subtitle_period(sub).from + _pts_offset);
                }
                _have_current_subtitle = false;
        }
@@ -567,11 +566,11 @@ FFmpegDecoder::decode_subtitle_packet ()
        FFmpegSubtitlePeriod sub_period = subtitle_period (sub);
        ContentTime from;
        from = sub_period.from + _pts_offset;
-       _have_current_subtitle = true;
        if (sub_period.to) {
                _current_subtitle_to = *sub_period.to + _pts_offset;
        } else {
                _current_subtitle_to = optional<ContentTime>();
+               _have_current_subtitle = true;
        }
 
        for (unsigned int i = 0; i < sub.num_rects; ++i) {
@@ -593,7 +592,7 @@ FFmpegDecoder::decode_subtitle_packet ()
        }
 
        if (_current_subtitle_to) {
-               subtitle->emit_stop (*_current_subtitle_to);
+               only_text()->emit_stop (*_current_subtitle_to);
        }
 
        avsubtitle_free (&sub);
@@ -660,8 +659,19 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTime
                out_p += image->stride()[0] / sizeof (uint32_t);
        }
 
-       int const target_width = subtitle_codec_context()->width;
-       int const target_height = subtitle_codec_context()->height;
+       int target_width = subtitle_codec_context()->width;
+       if (target_width == 0 && video_codec_context()) {
+               /* subtitle_codec_context()->width == 0 has been seen in the wild but I don't
+                  know if it's supposed to mean something from FFmpeg's point of view.
+               */
+               target_width = video_codec_context()->width;
+       }
+       int target_height = subtitle_codec_context()->height;
+       if (target_height == 0 && video_codec_context()) {
+               target_height = video_codec_context()->height;
+       }
+       DCPOMATIC_ASSERT (target_width);
+       DCPOMATIC_ASSERT (target_height);
        dcpomatic::Rect<double> const scaled_rect (
                static_cast<double> (rect->x) / target_width,
                static_cast<double> (rect->y) / target_height,
@@ -669,7 +679,7 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTime
                static_cast<double> (rect->h) / target_height
                );
 
-       subtitle->emit_image_start (from, image, scaled_rect);
+       only_text()->emit_bitmap_start (from, image, scaled_rect);
 }
 
 void
@@ -679,21 +689,29 @@ FFmpegDecoder::decode_ass_subtitle (string ass, ContentTime from)
           produces a single format of Dialogue: lines...
        */
 
-       vector<string> bits;
-       split (bits, ass, is_any_of (","));
-       if (bits.size() < 10) {
+       int commas = 0;
+       string text;
+       for (size_t i = 0; i < ass.length(); ++i) {
+               if (commas < 9 && ass[i] == ',') {
+                       ++commas;
+               } else if (commas == 9) {
+                       text += ass[i];
+               }
+       }
+
+       if (text.empty ()) {
                return;
        }
 
        sub::RawSubtitle base;
        list<sub::RawSubtitle> raw = sub::SSAReader::parse_line (
                base,
-               bits[9],
+               text,
                _ffmpeg_content->video->size().width,
                _ffmpeg_content->video->size().height
                );
 
        BOOST_FOREACH (sub::Subtitle const & i, sub::collect<list<sub::Subtitle> > (raw)) {
-               subtitle->emit_text_start (from, i);
+               only_text()->emit_plain_start (from, i);
        }
 }