X-Git-Url: https://git.carlh.net/gitweb/?p=dcpomatic.git;a=blobdiff_plain;f=src%2Flib%2Fffmpeg_decoder.cc;h=ba96d71ff5da78a9fe084d1bd56bbb2d82be56d4;hp=ea961a894d5446b5a81731ea86909a14cc918635;hb=182b9d2e2feb6545592868606aaf0f0146095481;hpb=ca44e3542214050de6cb8bbb223138765ac4bdb7 diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc index ea961a894..ba96d71ff 100644 --- a/src/lib/ffmpeg_decoder.cc +++ b/src/lib/ffmpeg_decoder.cc @@ -93,8 +93,10 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr film, shared_ptronly_text()) { - /* XXX: this time here should be the time of the first subtitle, not 0 */ - text.push_back (make_shared(this, c->only_text(), ContentTime())); + text.push_back (make_shared(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()) { @@ -122,7 +124,7 @@ FFmpegDecoder::flush () /* EOF can happen if we've already sent a flush packet */ throw DecodeError (N_("avcodec_send_packet"), N_("FFmpegDecoder::flush"), r); } - r = avcodec_receive_frame (context, _frame); + r = avcodec_receive_frame (context, audio_frame(i)); if (r >= 0) { process_audio_frame (i); did_something = true; @@ -144,7 +146,7 @@ FFmpegDecoder::flush () auto const f = full_length.frames_round (vfr); auto v = video->position(film()).get_value_or(ContentTime()).frames_round(vfr) + 1; while (v < f) { - video->emit (film(), shared_ptr (new RawImageProxy (_black_image)), v); + video->emit (film(), make_shared(_black_image), v); ++v; } } @@ -447,10 +449,13 @@ FFmpegDecoder::audio_stream_from_index (int index) const void FFmpegDecoder::process_audio_frame (shared_ptr stream) { - auto data = deinterleave_audio (_frame); + auto frame = audio_frame (stream); + auto data = deinterleave_audio (frame); + + auto const time_base = stream->stream(_format_context)->time_base; ContentTime ct; - if (_frame->pts == AV_NOPTS_VALUE) { + if (frame->pts == AV_NOPTS_VALUE) { /* In some streams we see not every frame coming through with a timestamp; for those that have AV_NOPTS_VALUE we need to work out the timestamp ourselves. This is particularly noticeable with TrueHD streams (see #1111). @@ -460,9 +465,17 @@ FFmpegDecoder::process_audio_frame (shared_ptr stream) } } else { ct = ContentTime::from_seconds ( - _frame->best_effort_timestamp * - av_q2d (stream->stream(_format_context)->time_base)) + frame->best_effort_timestamp * + 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()); @@ -481,8 +494,8 @@ FFmpegDecoder::process_audio_frame (shared_ptr stream) to_string(ct), data->frames(), stream->id(), - _frame->best_effort_timestamp, - av_q2d(stream->stream(_format_context)->time_base), + frame->best_effort_timestamp, + av_q2d(time_base), to_string(_pts_offset) ); } @@ -503,15 +516,18 @@ 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); } while (r >= 0) { - r = avcodec_receive_frame (context, _frame); + 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; } @@ -531,43 +547,58 @@ FFmpegDecoder::decode_and_process_video_packet (AVPacket* packet) auto context = video_codec_context(); - int r = avcodec_send_packet (context, packet); - if (r < 0) { - LOG_WARNING("avcodec_send_packet returned %1 for a video packet", r); - } + bool pending = false; + do { + int r = avcodec_send_packet (context, packet); + if (r < 0) { + LOG_WARNING("avcodec_send_packet returned %1 for a video packet", r); + } - r = avcodec_receive_frame (context, _frame); - if (r == AVERROR(EAGAIN) || r == AVERROR_EOF || (r < 0 && !packet)) { - /* More input is required, no more frames are coming, or we are flushing and there was - * some error which we just want to ignore. - */ - return false; - } else if (r < 0) { - throw DecodeError (N_("avcodec_receive_frame"), N_("FFmpeg::decode_and_process_video_packet"), r); - } + /* EAGAIN means we should call avcodec_receive_frame and then re-send the same packet */ + pending = r == AVERROR(EAGAIN); + + while (true) { + r = avcodec_receive_frame (context, _video_frame); + if (r == AVERROR(EAGAIN) || r == AVERROR_EOF || (r < 0 && !packet)) { + /* More input is required, no more frames are coming, or we are flushing and there was + * some error which we just want to ignore. + */ + return false; + } else if (r < 0) { + throw DecodeError (N_("avcodec_receive_frame"), N_("FFmpeg::decode_and_process_video_packet"), r); + } + + process_video_frame (); + } + } while (pending); + + return true; +} - /* We assume we'll only get one frame here, which I think is safe */ +void +FFmpegDecoder::process_video_frame () +{ boost::mutex::scoped_lock lm (_filter_graphs_mutex); shared_ptr graph; auto i = _filter_graphs.begin(); - while (i != _filter_graphs.end() && !(*i)->can_process (dcp::Size (_frame->width, _frame->height), (AVPixelFormat) _frame->format)) { + while (i != _filter_graphs.end() && !(*i)->can_process(dcp::Size(_video_frame->width, _video_frame->height), (AVPixelFormat) _video_frame->format)) { ++i; } if (i == _filter_graphs.end ()) { dcp::Fraction vfr (lrint(_ffmpeg_content->video_frame_rate().get() * 1000), 1000); - graph = make_shared(dcp::Size (_frame->width, _frame->height), (AVPixelFormat) _frame->format, vfr); + graph = make_shared(dcp::Size(_video_frame->width, _video_frame->height), (AVPixelFormat) _video_frame->format, vfr); graph->setup (_ffmpeg_content->filters ()); _filter_graphs.push_back (graph); - LOG_GENERAL (N_("New graph for %1x%2, pixel format %3"), _frame->width, _frame->height, _frame->format); + LOG_GENERAL (N_("New graph for %1x%2, pixel format %3"), _video_frame->width, _video_frame->height, _video_frame->format); } else { graph = *i; } - auto images = graph->process (_frame); + auto images = graph->process (_video_frame); for (auto const& i: images) { @@ -585,8 +616,6 @@ FFmpegDecoder::decode_and_process_video_packet (AVPacket* packet) LOG_WARNING_NC ("Dropping frame without PTS"); } } - - return true; } @@ -629,6 +658,7 @@ 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]; @@ -636,7 +666,7 @@ 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"; @@ -647,6 +677,10 @@ 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); } @@ -655,8 +689,8 @@ 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. @@ -739,7 +773,7 @@ FFmpegDecoder::process_bitmap_subtitle (AVSubtitleRect const * rect, ContentTime static_cast(rect->h) / target_height ); - only_text()->emit_bitmap_start (from, image, scaled_rect); + return { image, scaled_rect }; } @@ -769,7 +803,8 @@ 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>(raw)) {