X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Fffmpeg_decoder.cc;h=253272e9638a3387f529b7d5548eff86321ddaf6;hb=5d6e2ffca8e4b0d587eff8723716003a6d81be47;hp=698078b5b4339eae552aa67d249d24d24bd82d9d;hpb=e5eec6c5e7b96cc6e7697bbf42d3f27546c5ad52;p=dcpomatic.git diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc index 698078b5b..253272e96 100644 --- a/src/lib/ffmpeg_decoder.cc +++ b/src/lib/ffmpeg_decoder.cc @@ -1,19 +1,20 @@ /* Copyright (C) 2012-2016 Carl Hetherington - This program is free software; you can redistribute it and/or modify + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. - This program is distributed in the hope that it will be useful, + DCP-o-matic is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + along with DCP-o-matic. If not, see . */ @@ -27,15 +28,19 @@ #include "util.h" #include "log.h" #include "ffmpeg_decoder.h" +#include "subtitle_decoder.h" #include "ffmpeg_audio_stream.h" #include "ffmpeg_subtitle_stream.h" #include "video_filter_graph.h" #include "audio_buffers.h" #include "ffmpeg_content.h" #include "raw_image_proxy.h" +#include "video_decoder.h" #include "film.h" -#include "md5_digester.h" +#include "audio_decoder.h" #include "compose.hpp" +#include "subtitle_content.h" +#include "audio_content.h" #include #include #include @@ -69,17 +74,35 @@ using std::map; using boost::shared_ptr; using boost::is_any_of; using boost::split; +using boost::optional; +using boost::dynamic_pointer_cast; using dcp::Size; -FFmpegDecoder::FFmpegDecoder (shared_ptr c, shared_ptr log, bool fast) - : VideoDecoder (c) - , AudioDecoder (c, fast) - , SubtitleDecoder (c) - , FFmpeg (c) +FFmpegDecoder::FFmpegDecoder (shared_ptr c, shared_ptr log) + : FFmpeg (c) , _log (log) - , _pts_offset (pts_offset (c->ffmpeg_audio_streams(), c->first_video(), c->video_frame_rate())) { + 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()); + } else { + _pts_offset = ContentTime (); + } + + if (c->audio) { + audio.reset (new AudioDecoder (this, c->audio, log)); + } + if (c->subtitle) { + subtitle.reset ( + new SubtitleDecoder ( + this, + c->subtitle, + bind (&FFmpegDecoder::image_subtitles_during, this, _1, _2), + bind (&FFmpegDecoder::text_subtitles_during, this, _1, _2) + ) + ); + } } void @@ -92,10 +115,12 @@ FFmpegDecoder::flush () /* XXX: should we reset _packet.data and size after each *_decode_* call? */ - while (decode_video_packet ()) {} + while (video && decode_video_packet ()) {} - decode_audio_packet (); - AudioDecoder::flush (); + if (audio) { + decode_audio_packet (); + audio->flush (); + } } bool @@ -112,7 +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 (); @@ -122,7 +147,7 @@ FFmpegDecoder::pass (PassReason reason, bool accurate) int const si = _packet.stream_index; shared_ptr fc = _ffmpeg_content; - if (si == _video_stream && !_ignore_video && (accurate || reason != PASS_REASON_SUBTITLE)) { + if (_video_stream && si == _video_stream.get() && !video->ignore() && (accurate || reason != PASS_REASON_SUBTITLE)) { decode_video_packet (); } else if (fc->subtitle_stream() && fc->subtitle_stream()->uses_index (_format_context, si)) { decode_subtitle_packet (); @@ -207,7 +232,7 @@ FFmpegDecoder::deinterleave_audio (shared_ptr stream) const int sample = 0; int channel = 0; for (int i = 0; i < total_samples; ++i) { - audio->data(channel)[sample] = static_cast(*p++) / (1 << 31); + audio->data(channel)[sample] = static_cast(*p++) / 2147483648; ++channel; if (channel == stream->channels()) { @@ -223,7 +248,7 @@ FFmpegDecoder::deinterleave_audio (shared_ptr stream) const int32_t** p = reinterpret_cast (_frame->data); for (int i = 0; i < stream->channels(); ++i) { for (int j = 0; j < frames; ++j) { - audio->data(i)[j] = static_cast(p[i][j]) / (1 << 31); + audio->data(i)[j] = static_cast(p[i][j]) / 2147483648; } } } @@ -281,9 +306,17 @@ FFmpegDecoder::bytes_per_audio_sample (shared_ptr stream) con void FFmpegDecoder::seek (ContentTime time, bool accurate) { - VideoDecoder::seek (time, accurate); - AudioDecoder::seek (time, accurate); - SubtitleDecoder::seek (time, accurate); + if (video) { + video->seek (time, accurate); + } + + if (audio) { + audio->seek (time, accurate); + } + + if (subtitle) { + subtitle->seek (time, accurate); + } /* If we are doing an `accurate' seek, we need to use pre-roll, as we don't really know what the seek will give us. @@ -296,13 +329,33 @@ FFmpegDecoder::seek (ContentTime time, bool accurate) http://www.mjbshaw.com/2012/04/seeking-in-ffmpeg-know-your-timestamp.html */ + optional stream; + + if (_video_stream) { + stream = _video_stream; + } else { + shared_ptr s = dynamic_pointer_cast (_ffmpeg_content->audio->stream ()); + if (s) { + stream = s->index (_format_context); + } + } + + DCPOMATIC_ASSERT (stream); + ContentTime u = time - _pts_offset; if (u < ContentTime ()) { u = ContentTime (); } - av_seek_frame (_format_context, _video_stream, u.seconds() / av_q2d (_format_context->streams[_video_stream]->time_base), AVSEEK_FLAG_BACKWARD); + av_seek_frame ( + _format_context, + 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? */ @@ -339,7 +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. @@ -368,8 +421,15 @@ FFmpegDecoder::decode_audio_packet () ct += ContentTime::from_frames (remove, (*stream)->frame_rate ()); } - if (data->frames() > 0) { - audio (*stream, data, ct); + if (ct < ContentTime()) { + LOG_WARNING ("Crazy timestamp %s", to_string (ct)); + } + + update_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); } } @@ -381,6 +441,8 @@ FFmpegDecoder::decode_audio_packet () bool FFmpegDecoder::decode_video_packet () { + DCPOMATIC_ASSERT (_video_stream); + int frame_finished; if (avcodec_decode_video2 (video_codec_context(), _frame, &frame_finished, &_packet) < 0 || !frame_finished) { return false; @@ -411,11 +473,12 @@ FFmpegDecoder::decode_video_packet () shared_ptr image = i->first; if (i->second != AV_NOPTS_VALUE) { - double const pts = i->second * av_q2d (_format_context->streams[_video_stream]->time_base) + _pts_offset.seconds (); - video ( + double const pts = i->second * av_q2d (_format_context->streams[_video_stream.get()]->time_base) + _pts_offset.seconds (); + video->give ( shared_ptr (new RawImageProxy (image)), - llrint (pts * _ffmpeg_content->video_frame_rate ()) + llrint (pts * _ffmpeg_content->active_video_frame_rate ()) ); + update_position (ContentTime::from_seconds (pts)); } else { LOG_WARNING_NC ("Dropping frame without PTS"); } @@ -446,6 +509,7 @@ FFmpegDecoder::decode_subtitle_packet () FFmpegSubtitlePeriod sub_period = subtitle_period (sub); ContentTimePeriod period; period.from = sub_period.from + _pts_offset; + update_position (period.from); if (sub_period.to) { /* We already know the subtitle period `to' time */ period.to = sub_period.to.get() + _pts_offset; @@ -548,15 +612,16 @@ 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 const scaled_rect ( - static_cast (rect->x) / vs.width, - static_cast (rect->y) / vs.height, - static_cast (rect->w) / vs.width, - static_cast (rect->h) / vs.height + static_cast (rect->x) / target_width, + static_cast (rect->y) / target_height, + static_cast (rect->w) / target_width, + static_cast (rect->h) / target_height ); - image_subtitle (period, image, scaled_rect); + subtitle->give_image (period, image, scaled_rect); } void @@ -574,54 +639,21 @@ FFmpegDecoder::decode_ass_subtitle (string ass, ContentTimePeriod period) sub::RawSubtitle base; list raw = sub::SSAReader::parse_line (base, bits[9]); - list subs = sub::collect > (raw); - /* XXX: lots of this is copied from TextSubtitle; there should probably be some sharing */ - - /* Highest line index in this subtitle */ - int highest = 0; - BOOST_FOREACH (sub::Subtitle i, subs) { - BOOST_FOREACH (sub::Line j, i.lines) { - DCPOMATIC_ASSERT (j.vertical_position.reference && j.vertical_position.reference.get() == sub::TOP_OF_SUBTITLE); - DCPOMATIC_ASSERT (j.vertical_position.line); - highest = max (highest, j.vertical_position.line.get()); - } + BOOST_FOREACH (sub::Subtitle const & i, sub::collect > (raw)) { + subtitle->give_text (period, i); } +} - list ss; - - BOOST_FOREACH (sub::Subtitle i, sub::collect > (sub::SSAReader::parse_line (base, bits[9]))) { - BOOST_FOREACH (sub::Line j, i.lines) { - BOOST_FOREACH (sub::Block k, j.blocks) { - ss.push_back ( - dcp::SubtitleString ( - boost::optional (), - k.italic, - k.bold, - dcp::Colour (255, 255, 255), - /* 48pt is 1/22nd of the screen height */ - 48, - 1, - dcp::Time (i.from.seconds(), 1000), - dcp::Time (i.to.seconds(), 1000), - 0, - dcp::HALIGN_CENTER, - /* This 1.015 is an arbitrary value to lift the bottom sub off the bottom - of the screen a bit to a pleasing degree. - */ - 1.015 - ((1 + highest - j.vertical_position.line.get()) * 1.5 / 22), - dcp::VALIGN_TOP, - dcp::DIRECTION_LTR, - k.text, - static_cast (0), - dcp::Colour (255, 255, 255), - dcp::Time (), - dcp::Time () - ) - ); - } - } +void +FFmpegDecoder::update_position (ContentTime p) +{ + /* _position should err on the side of being too big, as then there is less + chance that we will erroneously decide not to seek when _position > request. + */ + if (!_position) { + _position = p; + } else { + _position = max (*_position, p); } - - text_subtitle (period, ss); }