X-Git-Url: https://main.carlh.net/gitweb/?p=dcpomatic.git;a=blobdiff_plain;f=src%2Flib%2Fvideo_decoder.cc;h=508ed90b71c71f62e874f0a763b4b19f2d9db8d7;hp=02f8fa0ac661a6eb946dd9ff498a09fdeb606813;hb=5eb8b5c3a1566aef638e9d9df03b88d320735092;hpb=b67a3505b58ffd204ee1a2560be3026688be8f10 diff --git a/src/lib/video_decoder.cc b/src/lib/video_decoder.cc index 02f8fa0ac..508ed90b7 100644 --- a/src/lib/video_decoder.cc +++ b/src/lib/video_decoder.cc @@ -1,19 +1,20 @@ /* - Copyright (C) 2012-2015 Carl Hetherington + Copyright (C) 2012-2018 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 . */ @@ -23,6 +24,7 @@ #include "film.h" #include "log.h" #include "compose.hpp" +#include #include #include "i18n.h" @@ -33,308 +35,135 @@ using std::max; using std::back_inserter; using boost::shared_ptr; using boost::optional; +using namespace dcpomatic; -VideoDecoder::VideoDecoder (shared_ptr c) -#ifdef DCPOMATIC_DEBUG - : test_gaps (0) - , _video_content (c) -#else - : _video_content (c) -#endif - , _last_seek_accurate (true) - , _ignore_video (false) -{ - _black_image.reset (new Image (AV_PIX_FMT_RGB24, _video_content->video_size(), true)); - _black_image->make_black (); -} - -list -VideoDecoder::decoded_video (Frame frame) +VideoDecoder::VideoDecoder (Decoder* parent, shared_ptr c) + : DecoderPart (parent) + , _content (c) { - list output; - for (list::const_iterator i = _decoded_video.begin(); i != _decoded_video.end(); ++i) { - if (i->frame == frame) { - output.push_back (*i); - } - } - - return output; } -/** Get all frames which exist in the content at a given frame index. - * @param frame Frame index. - * @param accurate true to try hard to return frames at the precise time that was requested, otherwise frames nearby may be returned. - * @return Frames; there may be none (if there is no video there), 1 for 2D or 2 for 3D. - */ -list -VideoDecoder::get_video (Frame frame, bool accurate) -{ - /* At this stage, if we have get_video()ed before, _decoded_video will contain the last frame that this - method returned (and possibly a few more). If the requested frame is not in _decoded_video and it is not the next - one after the end of _decoded_video we need to seek. - */ - - _video_content->film()->log()->log (String::compose ("VD has request for %1", frame), LogEntry::TYPE_DEBUG_DECODE); - - if (_decoded_video.empty() || frame < _decoded_video.front().frame || frame > (_decoded_video.back().frame + 1)) { - seek (ContentTime::from_frames (frame, _video_content->video_frame_rate()), accurate); - } - - list dec; - - /* Now enough pass() calls should either: - * (a) give us what we want, or - * (b) give us something after what we want, indicating that we will never get what we want, or - * (c) hit the end of the decoder. - */ - if (accurate) { - /* We are being accurate, so we want the right frame. - * This could all be one statement but it's split up for clarity. - */ - while (true) { - if (!decoded_video(frame).empty ()) { - /* We got what we want */ - break; - } - - if (pass ()) { - /* The decoder has nothing more for us */ - break; - } - - if (!_decoded_video.empty() && _decoded_video.front().frame > frame) { - /* We're never going to get the frame we want. Perhaps the caller is asking - * for a video frame before the content's video starts (if its audio - * begins before its video, for example). - */ - break; - } - } - - dec = decoded_video (frame); - } else { - /* Any frame will do: use the first one that comes out of pass() */ - while (_decoded_video.empty() && !pass ()) {} - if (!_decoded_video.empty ()) { - dec.push_back (_decoded_video.front ()); - } - } - - /* Clean up _decoded_video; keep the frame we are returning (which may have two images - for 3D), but nothing before that */ - while (!_decoded_video.empty() && _decoded_video.front().frame < dec.front().frame) { - _decoded_video.pop_front (); - } - - return dec; -} - -/** Fill _decoded_video from `from' up to, but not including, `to' with - * a frame for one particular Eyes value (which could be EYES_BOTH, - * EYES_LEFT or EYES_RIGHT) +/** Called by decoder classes when they have a video frame ready. + * @param frame Frame index within the content; this does not take into account 3D + * so for 3D_ALTERNATE this value goes: + * 0: frame 0 left + * 1: frame 0 right + * 2: frame 1 left + * 3: frame 1 right + * and so on. */ void -VideoDecoder::fill_one_eye (Frame from, Frame to, Eyes eye) +VideoDecoder::emit (shared_ptr film, shared_ptr image, Frame decoder_frame) { - if (to == 0) { - /* Already OK */ + if (ignore ()) { return; } - /* Fill with black... */ - shared_ptr filler_image (new RawImageProxy (_black_image)); - Part filler_part = PART_WHOLE; - - /* ...unless there's some video we can fill with */ - if (!_decoded_video.empty ()) { - filler_image = _decoded_video.back().image; - filler_part = _decoded_video.back().part; - } - - for (Frame i = from; i < to; ++i) { -#ifdef DCPOMATIC_DEBUG - test_gaps++; -#endif - _decoded_video.push_back ( - ContentVideo (filler_image, eye, filler_part, i) - ); - } -} - -/** Fill _decoded_video from `from' up to, but not including, `to' - * adding both left and right eye frames. - */ -void -VideoDecoder::fill_both_eyes (Frame from, Frame to, Eyes eye) -{ - if (to == 0 && eye == EYES_LEFT) { - /* Already OK */ - return; - } - - /* Fill with black... */ - shared_ptr filler_left_image (new RawImageProxy (_black_image)); - shared_ptr filler_right_image (new RawImageProxy (_black_image)); - Part filler_left_part = PART_WHOLE; - Part filler_right_part = PART_WHOLE; + /* Before we `re-write' the frame indexes of these incoming data we need to check for + the case where the user has some 2D content which they have marked as 3D. With 3D + we should get two frames for each frame index, but in this `bad' case we only get + one. We need to throw an exception if this happens. + */ - /* ...unless there's some video we can fill with */ - for (list::const_reverse_iterator i = _decoded_video.rbegin(); i != _decoded_video.rend(); ++i) { - if (i->eyes == EYES_LEFT && !filler_left_image) { - filler_left_image = i->image; - filler_left_part = i->part; - } else if (i->eyes == EYES_RIGHT && !filler_right_image) { - filler_right_image = i->image; - filler_right_part = i->part; + if (_content->video->frame_type() == VIDEO_FRAME_TYPE_3D) { + if (_last_threed_frames.size() > 4) { + _last_threed_frames.erase (_last_threed_frames.begin()); } - - if (filler_left_image && filler_right_image) { - break; + _last_threed_frames.push_back (decoder_frame); + if (_last_threed_frames.size() == 4) { + if (_last_threed_frames[0] != _last_threed_frames[1] || _last_threed_frames[2] != _last_threed_frames[3]) { + boost::throw_exception ( + DecodeError( + String::compose( + _("The content file %1 is set as 3D but does not appear to contain 3D images. Please set it to 2D. " + "You can still make a 3D DCP from this content by ticking the 3D option in the DCP video tab."), + _content->path(0) + ) + ) + ); + } } } - Frame filler_frame = from; - Eyes filler_eye = _decoded_video.empty() ? EYES_LEFT : _decoded_video.back().eyes; - - if (_decoded_video.empty ()) { - filler_frame = 0; - filler_eye = EYES_LEFT; - } else if (_decoded_video.back().eyes == EYES_LEFT) { - filler_frame = _decoded_video.back().frame; - filler_eye = EYES_RIGHT; - } else if (_decoded_video.back().eyes == EYES_RIGHT) { - filler_frame = _decoded_video.back().frame + 1; - filler_eye = EYES_LEFT; - } - - while (filler_frame != to || filler_eye != eye) { - -#ifdef DCPOMATIC_DEBUG - test_gaps++; -#endif - - _decoded_video.push_back ( - ContentVideo ( - filler_eye == EYES_LEFT ? filler_left_image : filler_right_image, - filler_eye, - filler_eye == EYES_LEFT ? filler_left_part : filler_right_part, - filler_frame - ) - ); - - if (filler_eye == EYES_LEFT) { - filler_eye = EYES_RIGHT; + double const afr = _content->active_video_frame_rate(film); + + Frame frame; + Eyes eyes = EYES_BOTH; + if (!_position) { + /* This is the first data we have received since initialisation or seek. Set + the position based on the frame that was given. After this first time + we just cound frames, since (as with audio) it seems that ContentTimes + are unreliable from FFmpegDecoder. They are much better than audio times + but still we get the occasional one which is duplicated. In this case + ffmpeg seems to carry on regardless, processing the video frame as normal. + If we drop the frame with the duplicated timestamp we obviously lose sync. + */ + _position = ContentTime::from_frames (decoder_frame, afr); + if (_content->video->frame_type() == VIDEO_FRAME_TYPE_3D_ALTERNATE) { + frame = decoder_frame / 2; + _last_emitted_eyes = EYES_RIGHT; } else { - filler_eye = EYES_LEFT; - ++filler_frame; + frame = decoder_frame; + } + } else { + VideoFrameType const ft = _content->video->frame_type (); + if (ft == VIDEO_FRAME_TYPE_3D_ALTERNATE || ft == VIDEO_FRAME_TYPE_3D) { + DCPOMATIC_ASSERT (_last_emitted_eyes); + if (_last_emitted_eyes.get() == EYES_RIGHT) { + frame = _position->frames_round(afr) + 1; + eyes = EYES_LEFT; + } else { + frame = _position->frames_round(afr); + eyes = EYES_RIGHT; + } + } else { + frame = _position->frames_round(afr) + 1; } } -} - -/** Called by subclasses when they have a video frame ready */ -void -VideoDecoder::video (shared_ptr image, Frame frame) -{ - if (_ignore_video) { - return; - } - - _video_content->film()->log()->log (String::compose ("VD receives %1", frame), LogEntry::TYPE_DEBUG_DECODE); - /* We may receive the same frame index twice for 3D, and we need to know - when that happens. - */ - bool const same = (!_decoded_video.empty() && frame == _decoded_video.back().frame); - - /* Work out what we are going to push into _decoded_video next */ - list to_push; - switch (_video_content->video_frame_type ()) { + switch (_content->video->frame_type ()) { case VIDEO_FRAME_TYPE_2D: - to_push.push_back (ContentVideo (image, EYES_BOTH, PART_WHOLE, frame)); + Data (ContentVideo (image, frame, EYES_BOTH, PART_WHOLE)); + break; + case VIDEO_FRAME_TYPE_3D: + { + Data (ContentVideo (image, frame, eyes, PART_WHOLE)); + _last_emitted_frame = frame; + _last_emitted_eyes = eyes; break; + } case VIDEO_FRAME_TYPE_3D_ALTERNATE: - to_push.push_back (ContentVideo (image, same ? EYES_RIGHT : EYES_LEFT, PART_WHOLE, frame)); + { + Data (ContentVideo (image, frame, eyes, PART_WHOLE)); + _last_emitted_eyes = eyes; break; + } case VIDEO_FRAME_TYPE_3D_LEFT_RIGHT: - to_push.push_back (ContentVideo (image, EYES_LEFT, PART_LEFT_HALF, frame)); - to_push.push_back (ContentVideo (image, EYES_RIGHT, PART_RIGHT_HALF, frame)); + Data (ContentVideo (image, frame, EYES_LEFT, PART_LEFT_HALF)); + Data (ContentVideo (image, frame, EYES_RIGHT, PART_RIGHT_HALF)); break; case VIDEO_FRAME_TYPE_3D_TOP_BOTTOM: - to_push.push_back (ContentVideo (image, EYES_LEFT, PART_TOP_HALF, frame)); - to_push.push_back (ContentVideo (image, EYES_RIGHT, PART_BOTTOM_HALF, frame)); + Data (ContentVideo (image, frame, EYES_LEFT, PART_TOP_HALF)); + Data (ContentVideo (image, frame, EYES_RIGHT, PART_BOTTOM_HALF)); break; case VIDEO_FRAME_TYPE_3D_LEFT: - to_push.push_back (ContentVideo (image, EYES_LEFT, PART_WHOLE, frame)); + Data (ContentVideo (image, frame, EYES_LEFT, PART_WHOLE)); break; case VIDEO_FRAME_TYPE_3D_RIGHT: - to_push.push_back (ContentVideo (image, EYES_RIGHT, PART_WHOLE, frame)); + Data (ContentVideo (image, frame, EYES_RIGHT, PART_WHOLE)); break; default: DCPOMATIC_ASSERT (false); } - /* Now VideoDecoder is required never to have gaps in the frames that it presents - via get_video(). Hence we need to fill in any gap between the last thing in _decoded_video - and the things we are about to push. - */ - - optional from; - optional to; - - if (_decoded_video.empty() && _last_seek_time && _last_seek_accurate) { - from = _last_seek_time->frames_round (_video_content->video_frame_rate ()); - to = to_push.front().frame; - } else if (!_decoded_video.empty ()) { - from = _decoded_video.back().frame + 1; - to = to_push.front().frame; - } - - /* If we've pre-rolled on a seek we may now receive out-of-order frames - (frames before the last seek time) which we can just ignore. - */ - - if (from && to && from.get() > to.get()) { - return; - } - - if (from) { - switch (_video_content->video_frame_type ()) { - case VIDEO_FRAME_TYPE_2D: - fill_one_eye (from.get(), to.get (), EYES_BOTH); - break; - case VIDEO_FRAME_TYPE_3D_LEFT_RIGHT: - case VIDEO_FRAME_TYPE_3D_TOP_BOTTOM: - case VIDEO_FRAME_TYPE_3D_ALTERNATE: - fill_both_eyes (from.get(), to.get(), to_push.front().eyes); - break; - case VIDEO_FRAME_TYPE_3D_LEFT: - fill_one_eye (from.get(), to.get (), EYES_LEFT); - break; - case VIDEO_FRAME_TYPE_3D_RIGHT: - fill_one_eye (from.get(), to.get (), EYES_RIGHT); - } - } - - copy (to_push.begin(), to_push.end(), back_inserter (_decoded_video)); - - /* We can't let this build up too much or we will run out of memory. We need to allow - the most frames that can exist between blocks of sound in a multiplexed file. - */ - DCPOMATIC_ASSERT (_decoded_video.size() <= 96); -} - -void -VideoDecoder::seek (ContentTime s, bool accurate) -{ - _decoded_video.clear (); - _last_seek_time = s; - _last_seek_accurate = accurate; + _position = ContentTime::from_frames (frame, afr); } -/** Set this player never to produce any video data */ void -VideoDecoder::set_ignore_video () +VideoDecoder::seek () { - _ignore_video = true; + _position = boost::optional(); + _last_emitted_frame.reset (); + _last_emitted_eyes.reset (); }