X-Git-Url: https://git.carlh.net/gitweb/?p=dcpomatic.git;a=blobdiff_plain;f=src%2Flib%2Fffmpeg_decoder.cc;h=a5b36d04f442ee71f8ca76adbc718f7c08f4ab3e;hp=30a4b1823e8873193051ec2c6e91f9bbfe68f4c8;hb=dba7e1137282b52a1bd6ad1d56fe6371a8c97e30;hpb=d311043bf3c1e3e7f41b314f7ab7c91ed7e5aa7f

diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc
index 30a4b1823..a5b36d04f 100644
--- a/src/lib/ffmpeg_decoder.cc
+++ b/src/lib/ffmpeg_decoder.cc
@@ -93,8 +93,10 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr<const Film> film, shared_ptr<const FFmp
 	}
 
 	if (c->only_text()) {
-		/* XXX: this time here should be the time of the first subtitle, not 0 */
-		text.push_back (make_shared<TextDecoder>(this, c->only_text(), ContentTime()));
+		text.push_back (make_shared<TextDecoder>(this, c->only_text()));
+		/* XXX: we should be calling maybe_set_position() on this TextDecoder, but we can't easily find
+		 * the time of the first subtitle at this point.
+		 */
 	}
 
 	for (auto i: c->ffmpeg_audio_streams()) {
@@ -217,8 +219,9 @@ FFmpegDecoder::pass ()
 /** @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.
  */
+static
 shared_ptr<AudioBuffers>
-FFmpegDecoder::deinterleave_audio (AVFrame* frame)
+deinterleave_audio(shared_ptr<FFmpegAudioStream> stream, AVFrame* frame)
 {
 	auto format = static_cast<AVSampleFormat>(frame->format);
 
@@ -324,12 +327,12 @@ FFmpegDecoder::deinterleave_audio (AVFrame* frame)
 	case AV_SAMPLE_FMT_FLTP:
 	{
 		auto p = reinterpret_cast<float**> (frame->data);
-		DCPOMATIC_ASSERT (frame->channels <= channels);
+		DCPOMATIC_ASSERT(channels <= stream->channels());
 		/* Sometimes there aren't as many channels in the frame as in the stream */
-		for (int i = 0; i < frame->channels; ++i) {
+		for (int i = 0; i < channels; ++i) {
 			memcpy (data[i], p[i], frames * sizeof(float));
 		}
-		for (int i = frame->channels; i < channels; ++i) {
+		for (int i = channels; i < stream->channels(); ++i) {
 			audio->make_silent (i);
 		}
 	}
@@ -448,7 +451,9 @@ void
 FFmpegDecoder::process_audio_frame (shared_ptr<FFmpegAudioStream> stream)
 {
 	auto frame = audio_frame (stream);
-	auto data = deinterleave_audio (frame);
+	auto data = deinterleave_audio(stream, frame);
+
+	auto const time_base = stream->stream(_format_context)->time_base;
 
 	ContentTime ct;
 	if (frame->pts == AV_NOPTS_VALUE) {
@@ -462,8 +467,16 @@ FFmpegDecoder::process_audio_frame (shared_ptr<FFmpegAudioStream> stream)
 	} else {
 		ct = ContentTime::from_seconds (
 			frame->best_effort_timestamp *
-			av_q2d (stream->stream(_format_context)->time_base))
+			av_q2d(time_base))
 			+ _pts_offset;
+		LOG_DEBUG_PLAYER(
+			"Process audio with timestamp %1 (BET %2, timebase %3/%4, (PTS offset %5)",
+			to_string(ct),
+			frame->best_effort_timestamp,
+			time_base.num,
+			time_base.den,
+			to_string(_pts_offset)
+			);
 	}
 
 	_next_time[stream] = ct + ContentTime::from_frames(data->frames(), stream->frame_rate());
@@ -483,7 +496,7 @@ FFmpegDecoder::process_audio_frame (shared_ptr<FFmpegAudioStream> stream)
 			data->frames(),
 			stream->id(),
 			frame->best_effort_timestamp,
-			av_q2d(stream->stream(_format_context)->time_base),
+			av_q2d(time_base),
 			to_string(_pts_offset)
 			);
 	}
@@ -506,6 +519,7 @@ FFmpegDecoder::decode_and_process_audio_packet (AVPacket* packet)
 	auto context = _codec_context[stream->index(_format_context)];
 	auto frame = audio_frame (stream);
 
+	LOG_DEBUG_PLAYER("Send audio packet on stream %1", stream->index(_format_context));
 	int r = avcodec_send_packet (context, packet);
 	if (r < 0) {
 		LOG_WARNING("avcodec_send_packet returned %1 for an audio packet", r);
@@ -514,6 +528,7 @@ FFmpegDecoder::decode_and_process_audio_packet (AVPacket* packet)
 		r = avcodec_receive_frame (context, frame);
 		if (r == AVERROR(EAGAIN)) {
 			/* More input is required */
+			LOG_DEBUG_PLAYER_NC("EAGAIN after trying to receive audio frame");
 			return;
 		}
 
@@ -644,6 +659,7 @@ FFmpegDecoder::decode_and_process_subtitle_packet (AVPacket* packet)
 		_have_current_subtitle = true;
 	}
 
+	ContentBitmapText bitmap_text(from);
 	for (unsigned int i = 0; i < sub.num_rects; ++i) {
 		auto const rect = sub.rects[i];
 
@@ -651,7 +667,7 @@ FFmpegDecoder::decode_and_process_subtitle_packet (AVPacket* packet)
 		case SUBTITLE_NONE:
 			break;
 		case SUBTITLE_BITMAP:
-			process_bitmap_subtitle (rect, from);
+			bitmap_text.subs.push_back(process_bitmap_subtitle(rect));
 			break;
 		case SUBTITLE_TEXT:
 			cout << "XXX: SUBTITLE_TEXT " << rect->text << "\n";
@@ -662,6 +678,10 @@ FFmpegDecoder::decode_and_process_subtitle_packet (AVPacket* packet)
 		}
 	}
 
+	if (!bitmap_text.subs.empty()) {
+		only_text()->emit_bitmap_start(bitmap_text);
+	}
+
 	if (_current_subtitle_to) {
 		only_text()->emit_stop (*_current_subtitle_to);
 	}
@@ -670,8 +690,8 @@ FFmpegDecoder::decode_and_process_subtitle_packet (AVPacket* packet)
 }
 
 
-void
-FFmpegDecoder::process_bitmap_subtitle (AVSubtitleRect const * rect, ContentTime from)
+BitmapText
+FFmpegDecoder::process_bitmap_subtitle (AVSubtitleRect const * rect)
 {
 	/* Note BGRA is expressed little-endian, so the first byte in the word is B, second
 	   G, third R, fourth A.
@@ -754,7 +774,7 @@ FFmpegDecoder::process_bitmap_subtitle (AVSubtitleRect const * rect, ContentTime
 		static_cast<double>(rect->h) / target_height
 		);
 
-	only_text()->emit_bitmap_start (from, image, scaled_rect);
+	return { image, scaled_rect };
 }
 
 
@@ -784,7 +804,8 @@ FFmpegDecoder::process_ass_subtitle (string ass, ContentTime from)
 		base,
 		text,
 		_ffmpeg_content->video->size().width,
-		_ffmpeg_content->video->size().height
+		_ffmpeg_content->video->size().height,
+		sub::Colour(1, 1, 1)
 		);
 
 	for (auto const& i: sub::collect<vector<sub::Subtitle>>(raw)) {