Fix DCP playback when an incorrect KDM is added, followed by
[dcpomatic.git] / src / lib / dcp_decoder.cc
index 4e28dc7582bea298c5641b054ad4ee585069460e..ab724cd6c75bbb7a5f02326e15b92f3a964a0322 100644 (file)
 /*
-    Copyright (C) 2014-2016 Carl Hetherington <cth@carlh.net>
+    Copyright (C) 2014-2019 Carl Hetherington <cth@carlh.net>
 
-    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 <http://www.gnu.org/licenses/>.
 
 */
 
 #include "dcp_decoder.h"
 #include "dcp_content.h"
 #include "audio_content.h"
+#include "video_decoder.h"
+#include "audio_decoder.h"
 #include "j2k_image_proxy.h"
+#include "text_decoder.h"
+#include "ffmpeg_image_proxy.h"
 #include "image.h"
 #include "config.h"
+#include "digester.h"
+#include "frame_interval_checker.h"
 #include <dcp/dcp.h>
-#include <dcp/decrypted_kdm.h>
 #include <dcp/cpl.h>
 #include <dcp/reel.h>
 #include <dcp/mono_picture_asset.h>
+#include <dcp/mono_picture_asset_reader.h>
 #include <dcp/stereo_picture_asset.h>
+#include <dcp/stereo_picture_asset_reader.h>
 #include <dcp/reel_picture_asset.h>
 #include <dcp/reel_sound_asset.h>
 #include <dcp/reel_subtitle_asset.h>
+#include <dcp/reel_closed_caption_asset.h>
 #include <dcp/mono_picture_frame.h>
 #include <dcp/stereo_picture_frame.h>
 #include <dcp/sound_frame.h>
+#include <dcp/sound_asset_reader.h>
+#include <dcp/subtitle_image.h>
+#include <dcp/decrypted_kdm.h>
 #include <boost/foreach.hpp>
 #include <iostream>
 
+#include "i18n.h"
+
 using std::list;
 using std::cout;
+using std::string;
 using boost::shared_ptr;
 using boost::dynamic_pointer_cast;
+using boost::optional;
+using namespace dcpomatic;
 
-DCPDecoder::DCPDecoder (shared_ptr<const DCPContent> c, shared_ptr<Log> log, bool fast)
-       : VideoDecoder (c->video, log)
-       , AudioDecoder (c->audio, fast, log)
-       , SubtitleDecoder (c->subtitle)
-       , _dcp_content (c)
+DCPDecoder::DCPDecoder (shared_ptr<const Film> film, shared_ptr<const DCPContent> c, bool fast, bool tolerant, shared_ptr<DCPDecoder> old)
+       : DCP (c, tolerant)
+       , Decoder (film)
+       , _decode_referenced (false)
 {
-       dcp::DCP dcp (c->directory ());
-       dcp.read (false, 0, true);
-       if (c->kdm ()) {
-               dcp.add (dcp::DecryptedKDM (c->kdm().get (), Config::instance()->decryption_chain()->key().get ()));
+       if (c->can_be_played()) {
+               if (c->video) {
+                       video.reset (new VideoDecoder (this, c));
+               }
+               if (c->audio) {
+                       audio.reset (new AudioDecoder (this, c->audio, fast));
+               }
+               BOOST_FOREACH (shared_ptr<TextContent> i, c->text) {
+                       /* XXX: this time here should be the time of the first subtitle, not 0 */
+                       text.push_back (shared_ptr<TextDecoder> (new TextDecoder (this, i, ContentTime())));
+               }
        }
-       DCPOMATIC_ASSERT (dcp.cpls().size() == 1);
-       _reels = dcp.cpls().front()->reels ();
+
+       /* We try to avoid re-scanning the DCP's files every time we make a new DCPDecoder; we do this
+          by re-using the _reels list.  Before we do this we need to check that nothing too serious
+          has changed in the DCPContent.
+
+          We do this by storing a digest of the important bits of the DCPContent and then checking that's
+          the same before we re-use _reels.
+       */
+
+       _lazy_digest = calculate_lazy_digest (c);
+
+       if (old && old->lazy_digest() == _lazy_digest) {
+               _reels = old->_reels;
+       } else {
+
+               list<shared_ptr<dcp::CPL> > cpl_list = cpls ();
+
+               if (cpl_list.empty()) {
+                       throw DCPError (_("No CPLs found in DCP."));
+               }
+
+               shared_ptr<dcp::CPL> cpl;
+               BOOST_FOREACH (shared_ptr<dcp::CPL> i, cpl_list) {
+                       if (_dcp_content->cpl() && i->id() == _dcp_content->cpl().get()) {
+                               cpl = i;
+                       }
+               }
+
+               if (!cpl) {
+                       /* No CPL found; probably an old file that doesn't specify it;
+                          just use the first one.
+                       */
+                       cpl = cpls().front ();
+               }
+
+               _reels = cpl->reels ();
+       }
+
+       set_decode_referenced (false);
+
        _reel = _reels.begin ();
+       _offset = 0;
+       get_readers ();
 }
 
+
 bool
-DCPDecoder::pass (PassReason reason, bool)
+DCPDecoder::pass ()
 {
-       if (_reel == _reels.end () || !_dcp_content->can_be_played ()) {
+       if (!_dcp_content->can_be_played()) {
                return true;
        }
 
-       /* Offset of the start of the current reel from the start of the content in frames */
-       int offset = 0;
-       list<shared_ptr<dcp::Reel> >::const_iterator i = _reels.begin();
-       while (i != _reel) {
-               offset += (*i)->main_picture()->duration ();
-               ++i;
+       if (_reel == _reels.end()) {
+               if (audio) {
+                       audio->flush ();
+               }
+               return true;
        }
 
-       double const vfr = _dcp_content->video->video_frame_rate ();
+       double const vfr = _dcp_content->active_video_frame_rate (film());
 
        /* Frame within the (played part of the) reel that is coming up next */
        int64_t const frame = _next.frames_round (vfr);
 
-       if ((*_reel)->main_picture () && reason != PASS_REASON_SUBTITLE) {
-               shared_ptr<dcp::PictureAsset> asset = (*_reel)->main_picture()->asset ();
-               shared_ptr<dcp::MonoPictureAsset> mono = dynamic_pointer_cast<dcp::MonoPictureAsset> (asset);
-               shared_ptr<dcp::StereoPictureAsset> stereo = dynamic_pointer_cast<dcp::StereoPictureAsset> (asset);
-               int64_t const entry_point = (*_reel)->main_picture()->entry_point ();
-               if (mono) {
-                       video (shared_ptr<ImageProxy> (new J2KImageProxy (mono->get_frame (entry_point + frame), asset->size())), offset + frame);
+       shared_ptr<dcp::PictureAsset> picture_asset = (*_reel)->main_picture()->asset();
+       DCPOMATIC_ASSERT (picture_asset);
+
+       /* We must emit texts first as when we emit the video for this frame
+          it will expect already to have the texts.
+       */
+       pass_texts (_next, picture_asset->size());
+
+       if ((_mono_reader || _stereo_reader) && (_decode_referenced || !_dcp_content->reference_video())) {
+               int64_t const entry_point = (*_reel)->main_picture()->entry_point().get_value_or(0);
+               if (_mono_reader) {
+                       video->emit (
+                               film(),
+                               shared_ptr<ImageProxy> (
+                                       new J2KImageProxy (
+                                               _mono_reader->get_frame (entry_point + frame),
+                                               picture_asset->size(),
+                                               AV_PIX_FMT_XYZ12LE,
+                                               _forced_reduction
+                                               )
+                                       ),
+                               _offset + frame
+                               );
                } else {
-                       video (
-                               shared_ptr<ImageProxy> (new J2KImageProxy (stereo->get_frame (entry_point + frame), asset->size(), dcp::EYE_LEFT)),
-                               offset + frame
+                       video->emit (
+                               film(),
+                               shared_ptr<ImageProxy> (
+                                       new J2KImageProxy (
+                                               _stereo_reader->get_frame (entry_point + frame),
+                                               picture_asset->size(),
+                                               dcp::EYE_LEFT,
+                                               AV_PIX_FMT_XYZ12LE,
+                                               _forced_reduction
+                                               )
+                                       ),
+                               _offset + frame
                                );
 
-                       video (
-                               shared_ptr<ImageProxy> (new J2KImageProxy (stereo->get_frame (entry_point + frame), asset->size(), dcp::EYE_RIGHT)),
-                               offset + frame
+                       video->emit (
+                               film(),
+                               shared_ptr<ImageProxy> (
+                                       new J2KImageProxy (
+                                               _stereo_reader->get_frame (entry_point + frame),
+                                               picture_asset->size(),
+                                               dcp::EYE_RIGHT,
+                                               AV_PIX_FMT_XYZ12LE,
+                                               _forced_reduction
+                                               )
+                                       ),
+                               _offset + frame
                                );
                }
        }
 
-       if ((*_reel)->main_sound () && reason != PASS_REASON_SUBTITLE) {
-               int64_t const entry_point = (*_reel)->main_sound()->entry_point ();
-               shared_ptr<const dcp::SoundFrame> sf = (*_reel)->main_sound()->asset()->get_frame (entry_point + frame);
+       if (_sound_reader && (_decode_referenced || !_dcp_content->reference_audio())) {
+               int64_t const entry_point = (*_reel)->main_sound()->entry_point().get_value_or(0);
+               shared_ptr<const dcp::SoundFrame> sf = _sound_reader->get_frame (entry_point + frame);
                uint8_t const * from = sf->data ();
 
                int const channels = _dcp_content->audio->stream()->channels ();
                int const frames = sf->size() / (3 * channels);
                shared_ptr<AudioBuffers> data (new AudioBuffers (channels, frames));
+               float** data_data = data->data();
                for (int i = 0; i < frames; ++i) {
                        for (int j = 0; j < channels; ++j) {
-                               data->data()[j][i] = static_cast<int> ((from[0] << 8) | (from[1] << 16) | (from[2] << 24)) / static_cast<float> (INT_MAX - 256);
+                               data_data[j][i] = static_cast<int> ((from[0] << 8) | (from[1] << 16) | (from[2] << 24)) / static_cast<float> (INT_MAX - 256);
                                from += 3;
                        }
                }
 
-               audio (_dcp_content->audio->stream(), data, ContentTime::from_frames (offset, vfr) + _next);
+               audio->emit (film(), _dcp_content->audio->stream(), data, ContentTime::from_frames (_offset, vfr) + _next);
+       }
+
+       _next += ContentTime::from_frames (1, vfr);
+
+       if ((*_reel)->main_picture ()) {
+               if (_next.frames_round (vfr) >= (*_reel)->main_picture()->duration()) {
+                       next_reel ();
+                       _next = ContentTime ();
+               }
+       }
+
+       return false;
+}
+
+void
+DCPDecoder::pass_texts (ContentTime next, dcp::Size size)
+{
+       list<shared_ptr<TextDecoder> >::const_iterator decoder = text.begin ();
+       if (decoder == text.end()) {
+               /* It's possible that there is now a main subtitle but no TextDecoders, for example if
+                  the CPL has just changed but the TextContent's texts have not been recreated yet.
+               */
+               return;
+       }
+
+       if ((*_reel)->main_subtitle()) {
+               pass_texts (
+                       next,
+                       (*_reel)->main_subtitle()->asset(),
+                       _dcp_content->reference_text(TEXT_OPEN_SUBTITLE),
+                       (*_reel)->main_subtitle()->entry_point().get_value_or(0),
+                       *decoder,
+                       size
+                       );
+               ++decoder;
        }
 
-       if ((*_reel)->main_subtitle ()) {
-               int64_t const entry_point = (*_reel)->main_subtitle()->entry_point ();
-               list<dcp::SubtitleString> subs = (*_reel)->main_subtitle()->asset()->subtitles_during (
+       BOOST_FOREACH (shared_ptr<dcp::ReelClosedCaptionAsset> i, (*_reel)->closed_captions()) {
+               pass_texts (
+                       next, i->asset(), _dcp_content->reference_text(TEXT_CLOSED_CAPTION), i->entry_point().get_value_or(0), *decoder, size
+                       );
+               ++decoder;
+       }
+}
+
+void
+DCPDecoder::pass_texts (
+       ContentTime next, shared_ptr<dcp::SubtitleAsset> asset, bool reference, int64_t entry_point, shared_ptr<TextDecoder> decoder, dcp::Size size
+       )
+{
+       double const vfr = _dcp_content->active_video_frame_rate (film());
+       /* Frame within the (played part of the) reel that is coming up next */
+       int64_t const frame = next.frames_round (vfr);
+
+       if (_decode_referenced || !reference) {
+               list<shared_ptr<dcp::Subtitle> > subs = asset->subtitles_during (
                        dcp::Time (entry_point + frame, vfr, vfr),
                        dcp::Time (entry_point + frame + 1, vfr, vfr),
                        true
                        );
 
-               if (!subs.empty ()) {
-                       /* XXX: assuming that all `subs' are at the same time; maybe this is ok */
-                       text_subtitle (
+               list<dcp::SubtitleString> strings;
+
+               BOOST_FOREACH (shared_ptr<dcp::Subtitle> i, subs) {
+                       shared_ptr<dcp::SubtitleString> is = dynamic_pointer_cast<dcp::SubtitleString> (i);
+                       if (is) {
+                               if (!strings.empty() && (strings.back().in() != is->in() || strings.back().out() != is->out())) {
+                                       dcp::SubtitleString b = strings.back();
+                                       decoder->emit_plain (
+                                               ContentTimePeriod (
+                                                       ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
+                                                       ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
+                                                       ),
+                                               strings
+                                               );
+                                       strings.clear ();
+                               }
+
+                               strings.push_back (*is);
+                       }
+
+                       /* XXX: perhaps these image subs should also be collected together like the string ones are;
+                          this would need to be done both here and in DCPSubtitleDecoder.
+                       */
+
+                       shared_ptr<dcp::SubtitleImage> ii = dynamic_pointer_cast<dcp::SubtitleImage> (i);
+                       if (ii) {
+                               emit_subtitle_image (
+                                       ContentTimePeriod (
+                                               ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->in().as_seconds ()),
+                                               ContentTime::from_frames (_offset - entry_point, vfr) + ContentTime::from_seconds (i->out().as_seconds ())
+                                               ),
+                                       *ii,
+                                       size,
+                                       decoder
+                                       );
+                       }
+               }
+
+               if (!strings.empty()) {
+                       dcp::SubtitleString b = strings.back();
+                       decoder->emit_plain (
                                ContentTimePeriod (
-                                       ContentTime::from_frames (offset - entry_point, vfr) + ContentTime::from_seconds (subs.front().in().as_seconds ()),
-                                       ContentTime::from_frames (offset - entry_point, vfr) + ContentTime::from_seconds (subs.front().out().as_seconds ())
+                                       ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.in().as_seconds()),
+                                       ContentTime::from_frames(_offset - entry_point, vfr) + ContentTime::from_seconds(b.out().as_seconds())
                                        ),
-                               subs
+                               strings
                                );
+                       strings.clear ();
                }
        }
+}
 
-       _next += ContentTime::from_frames (1, vfr);
+void
+DCPDecoder::next_reel ()
+{
+       _offset += (*_reel)->main_picture()->actual_duration();
+       ++_reel;
+       get_readers ();
+}
 
-       if ((*_reel)->main_picture ()) {
-               if (_next.frames_round (vfr) >= (*_reel)->main_picture()->duration()) {
-                       ++_reel;
-                       _next = ContentTime ();
+void
+DCPDecoder::get_readers ()
+{
+       if (_reel == _reels.end() || !_dcp_content->can_be_played ()) {
+               _mono_reader.reset ();
+               _stereo_reader.reset ();
+               _sound_reader.reset ();
+               return;
+       }
+
+       if ((*_reel)->main_picture()) {
+               shared_ptr<dcp::PictureAsset> asset = (*_reel)->main_picture()->asset ();
+               shared_ptr<dcp::MonoPictureAsset> mono = dynamic_pointer_cast<dcp::MonoPictureAsset> (asset);
+               shared_ptr<dcp::StereoPictureAsset> stereo = dynamic_pointer_cast<dcp::StereoPictureAsset> (asset);
+               DCPOMATIC_ASSERT (mono || stereo);
+               if (mono) {
+                       _mono_reader = mono->start_read ();
+                       _stereo_reader.reset ();
+               } else {
+                       _stereo_reader = stereo->start_read ();
+                       _mono_reader.reset ();
                }
+       } else {
+               _mono_reader.reset ();
+               _stereo_reader.reset ();
        }
 
-       return false;
+       if ((*_reel)->main_sound()) {
+               _sound_reader = (*_reel)->main_sound()->asset()->start_read ();
+       } else {
+               _sound_reader.reset ();
+       }
 }
 
 void
 DCPDecoder::seek (ContentTime t, bool accurate)
 {
-       VideoDecoder::seek (t, accurate);
-       AudioDecoder::seek (t, accurate);
-       SubtitleDecoder::seek (t, accurate);
+       if (!_dcp_content->can_be_played ()) {
+               return;
+       }
+
+       Decoder::seek (t, accurate);
 
        _reel = _reels.begin ();
-       while (_reel != _reels.end() && t >= ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->video->video_frame_rate ())) {
-               t -= ContentTime::from_frames ((*_reel)->main_picture()->duration(), _dcp_content->video->video_frame_rate ());
-               ++_reel;
+       _offset = 0;
+       get_readers ();
+
+       int const pre_roll_seconds = 2;
+
+       /* Pre-roll for subs */
+
+       ContentTime pre = t - ContentTime::from_seconds (pre_roll_seconds);
+       if (pre < ContentTime()) {
+               pre = ContentTime ();
        }
 
-       _next = t;
-}
+       /* Seek to pre-roll position */
 
+       while (
+               _reel != _reels.end() &&
+               pre >= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()))
+               ) {
 
-list<ContentTimePeriod>
-DCPDecoder::image_subtitles_during (ContentTimePeriod, bool) const
-{
-       return list<ContentTimePeriod> ();
-}
+               ContentTime rd = ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()));
+               pre -= rd;
+               t -= rd;
+               next_reel ();
+       }
 
-list<ContentTimePeriod>
-DCPDecoder::text_subtitles_during (ContentTimePeriod period, bool starting) const
-{
-       /* XXX: inefficient */
+       /* Pass texts in the pre-roll */
 
-       list<ContentTimePeriod> ctp;
-       double const vfr = _dcp_content->video->video_frame_rate ();
+       double const vfr = _dcp_content->active_video_frame_rate (film());
+       for (int i = 0; i < pre_roll_seconds * vfr; ++i) {
+               pass_texts (pre, (*_reel)->main_picture()->asset()->size());
+               pre += ContentTime::from_frames (1, vfr);
+       }
 
-       BOOST_FOREACH (shared_ptr<dcp::Reel> r, _reels) {
-               if (!r->main_subtitle ()) {
-                       continue;
-               }
+       /* Seek to correct position */
 
-               int64_t const entry_point = r->main_subtitle()->entry_point ();
+       while (
+               _reel != _reels.end() &&
+               t >= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()))
+               ) {
 
-               list<dcp::SubtitleString> subs = r->main_subtitle()->asset()->subtitles_during (
-                       dcp::Time (period.from.seconds(), 1000) - dcp::Time (entry_point, vfr, vfr),
-                       dcp::Time (period.to.seconds(), 1000) - dcp::Time (entry_point, vfr, vfr),
-                       starting
-                       );
+               t -= ContentTime::from_frames ((*_reel)->main_picture()->actual_duration(), _dcp_content->active_video_frame_rate(film()));
+               next_reel ();
+       }
 
-               BOOST_FOREACH (dcp::SubtitleString const & s, subs) {
-                       ctp.push_back (
-                               ContentTimePeriod (
-                                       ContentTime::from_seconds (s.in().as_seconds ()),
-                                       ContentTime::from_seconds (s.out().as_seconds ())
-                                       )
-                               );
-               }
+       _next = t;
+}
+
+void
+DCPDecoder::set_decode_referenced (bool r)
+{
+       _decode_referenced = r;
+
+       if (video) {
+               video->set_ignore (_dcp_content->reference_video() && !_decode_referenced);
        }
+       if (audio) {
+               audio->set_ignore (_dcp_content->reference_audio() && !_decode_referenced);
+       }
+}
+
+void
+DCPDecoder::set_forced_reduction (optional<int> reduction)
+{
+       _forced_reduction = reduction;
+}
 
-       return ctp;
+string
+DCPDecoder::calculate_lazy_digest (shared_ptr<const DCPContent> c) const
+{
+       Digester d;
+       BOOST_FOREACH (boost::filesystem::path i, c->paths()) {
+               d.add (i.string());
+       }
+       if (_dcp_content->kdm()) {
+               d.add(_dcp_content->kdm()->id());
+       }
+       d.add (static_cast<bool>(c->cpl()));
+       if (c->cpl()) {
+               d.add (c->cpl().get());
+       }
+       return d.get ();
+}
+
+ContentTime
+DCPDecoder::position () const
+{
+       return ContentTime::from_frames(_offset, _dcp_content->active_video_frame_rate(film())) + _next;
 }