Move position variables into the video/audio/subtitle decoder classes.

[dcpomatic.git] / src / lib / ffmpeg_decoder.cc

diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc
index 69ed7ac125f961828a748ce68d2886e5849eea36..ddc563b9855b2b5d5fdd7575d69bfbc9475960e9 100644 (file)
--- a/src/lib/ffmpeg_decoder.cc
+++ b/src/lib/ffmpeg_decoder.cc
@@ -40,6 +40,7 @@
 #include "audio_decoder.h"
 #include "compose.hpp"
 #include "subtitle_content.h"
+#include "audio_content.h"
 #include <dcp/subtitle_string.h>
 #include <sub/ssa_reader.h>
 #include <sub/subtitle.h>
@@ -50,7 +51,6 @@ extern "C" {
 }
 #include <boost/foreach.hpp>
 #include <boost/algorithm/string.hpp>
-#include <boost/make_shared.hpp>
 #include <vector>
 #include <iomanip>
 #include <iostream>
@@ -72,12 +72,13 @@ using std::pair;
 using std::max;
 using std::map;
 using boost::shared_ptr;
-using boost::make_shared;
 using boost::is_any_of;
 using boost::split;
+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 FFmpegContent> c, shared_ptr<Log> log)
        : FFmpeg (c)
        , _log (log)
 {
@@ -89,7 +90,7 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const FFmpegContent> c, shared_ptr<Log>
        }
 
        if (c->audio) {
-               audio.reset (new AudioDecoder (this, c->audio, fast, log));
+               audio.reset (new AudioDecoder (this, c->audio, log));
        }
 
        if (c->subtitle) {
@@ -136,7 +137,7 @@ FFmpegDecoder::pass (PassReason reason, bool accurate)
                        /* Maybe we should fail here, but for now we'll just finish off instead */
                        char buf[256];
                        av_strerror (r, buf, sizeof(buf));
-                       LOG_ERROR (N_("error on av_read_frame (%1) (%2)"), buf, r);
+                       LOG_ERROR (N_("error on av_read_frame (%1) (%2)"), &buf[0], r);
                }
 
                flush ();
@@ -177,7 +178,7 @@ FFmpegDecoder::deinterleave_audio (shared_ptr<FFmpegAudioStream> stream) const
        */
        int const total_samples = size / bytes_per_audio_sample (stream);
        int const frames = total_samples / stream->channels();
-       shared_ptr<AudioBuffers> audio = make_shared<AudioBuffers> (stream->channels(), frames);
+       shared_ptr<AudioBuffers> audio (new AudioBuffers (stream->channels(), frames));
 
        switch (audio_sample_format (stream)) {
        case AV_SAMPLE_FMT_U8:
@@ -328,7 +329,18 @@ FFmpegDecoder::seek (ContentTime time, bool accurate)
           http://www.mjbshaw.com/2012/04/seeking-in-ffmpeg-know-your-timestamp.html
        */
 
-       DCPOMATIC_ASSERT (_video_stream);
+       optional<int> stream;
+
+       if (_video_stream) {
+               stream = _video_stream;
+       } else {
+               shared_ptr<FFmpegAudioStream> s = dynamic_pointer_cast<FFmpegAudioStream> (_ffmpeg_content->audio->stream ());
+               if (s) {
+                       stream = s->index (_format_context);
+               }
+       }
+
+       DCPOMATIC_ASSERT (stream);
 
        ContentTime u = time - _pts_offset;
        if (u < ContentTime ()) {
@@ -336,12 +348,14 @@ FFmpegDecoder::seek (ContentTime time, bool accurate)
        }
        av_seek_frame (
                _format_context,
-               _video_stream.get(),
-               u.seconds() / av_q2d (_format_context->streams[_video_stream.get()]->time_base),
+               stream.get(),
+               u.seconds() / av_q2d (_format_context->streams[stream.get()]->time_base),
                AVSEEK_FLAG_BACKWARD
                );
 
-       avcodec_flush_buffers (video_codec_context());
+       if (video_codec_context ()) {
+               avcodec_flush_buffers (video_codec_context());
+       }
 
        /* XXX: should be flushing audio buffers? */
 
@@ -378,7 +392,7 @@ FFmpegDecoder::decode_audio_packet ()
                if (decode_result < 0) {
                        /* avcodec_decode_audio4 can sometimes return an error even though it has decoded
                           some valid data; for example dca_subframe_footer can return AVERROR_INVALIDDATA
-                          if it overreads the auxiliary data.  ffplay carries on if frame_finished is true,
+                          if it overreads the auxiliary data.  ffplay carries on if frame_finished is true,
                           even in the face of such an error, so I think we should too.
 
                           Returning from the method here caused mantis #352.
@@ -407,7 +421,14 @@ FFmpegDecoder::decode_audio_packet ()
                                ct += ContentTime::from_frames (remove, (*stream)->frame_rate ());
                        }
 
-                       if (data->frames() > 0) {
+                       if (ct < ContentTime()) {
+                               LOG_WARNING ("Crazy timestamp %s", to_string (ct));
+                       }
+
+                       audio->set_position (ct);
+
+                       /* Give this data provided there is some, and its time is sane */
+                       if (ct >= ContentTime() && data->frames() > 0) {
                                audio->give (*stream, data, ct);
                        }
                }
@@ -454,9 +475,10 @@ 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->give (
-                               make_shared<RawImageProxy> (image),
+                               shared_ptr<ImageProxy> (new RawImageProxy (image)),
                                llrint (pts * _ffmpeg_content->active_video_frame_rate ())
                                );
+                       video->set_position (ContentTime::from_seconds (pts));
                } else {
                        LOG_WARNING_NC ("Dropping frame without PTS");
                }
@@ -487,6 +509,7 @@ FFmpegDecoder::decode_subtitle_packet ()
        FFmpegSubtitlePeriod sub_period = subtitle_period (sub);
        ContentTimePeriod period;
        period.from = sub_period.from + _pts_offset;
+       subtitle->set_position (period.from);
        if (sub_period.to) {
                /* We already know the subtitle period `to' time */
                period.to = sub_period.to.get() + _pts_offset;
@@ -534,7 +557,7 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTimeP
        /* Note RGBA is expressed little-endian, so the first byte in the word is R, second
           G, third B, fourth A.
        */
-       shared_ptr<Image> image = make_shared<Image> (AV_PIX_FMT_RGBA, dcp::Size (rect->w, rect->h), true);
+       shared_ptr<Image> image (new Image (AV_PIX_FMT_RGBA, dcp::Size (rect->w, rect->h), true));
 
 #ifdef DCPOMATIC_HAVE_AVSUBTITLERECT_PICT
        /* Start of the first line in the subtitle */
@@ -589,12 +612,13 @@ FFmpegDecoder::decode_bitmap_subtitle (AVSubtitleRect const * rect, ContentTimeP
                out_p += image->stride()[0] / sizeof (uint32_t);
        }
 
-       dcp::Size const vs = _ffmpeg_content->video->size ();
+       int const target_width = subtitle_codec_context()->width;
+       int const target_height = subtitle_codec_context()->height;
        dcpomatic::Rect<double> const scaled_rect (
-               static_cast<double> (rect->x) / vs.width,
-               static_cast<double> (rect->y) / vs.height,
-               static_cast<double> (rect->w) / vs.width,
-               static_cast<double> (rect->h) / vs.height
+               static_cast<double> (rect->x) / target_width,
+               static_cast<double> (rect->y) / target_height,
+               static_cast<double> (rect->w) / target_width,
+               static_cast<double> (rect->h) / target_height
                );
 
        subtitle->give_image (period, image, scaled_rect);