From: Carl Hetherington Date: Sun, 5 May 2019 20:49:51 +0000 (+0000) Subject: Fix case where the is FFmpegContent with 24fps video and 44.1kHz audio X-Git-Tag: v2.14.2~3 X-Git-Url: https://main.carlh.net/gitweb/?a=commitdiff_plain;h=870ab8d9295b9d4b6605e8876919d23379dc3a35;p=dcpomatic.git Fix case where the is FFmpegContent with 24fps video and 44.1kHz audio and a start trim of 6724000. With these numbers the start trim is on an integer video frame but halfway through an audio frame. Without this patch the trim would be "corrected" to 6724001, causing video frames to come out of the player at DCPTimes 0, 3999, 5999 etc. It's possible that Frame const position = time.frames_floor(_film->video_frame_rate()); in J2KEncoder::encode should be frames_round, which would also help with this, but that would be a much more risky patch. --- diff --git a/src/lib/content.cc b/src/lib/content.cc index 2ca029d5a..9280d2e61 100644 --- a/src/lib/content.cc +++ b/src/lib/content.cc @@ -219,7 +219,13 @@ Content::set_position (shared_ptr film, DCPTime p, bool force_emit) video->modify_position (film, p); } - if (audio) { + /* Only allow the audio to modify if we have no video; + sometimes p can't be on an integer video AND audio frame, + and in these cases we want the video constraint to be + satisfied since (I think) the audio code is better able to + cope. + */ + if (!video && audio) { audio->modify_position (film, p); } @@ -245,7 +251,8 @@ Content::set_trim_start (ContentTime t) video->modify_trim_start (t); } - if (audio) { + /* See note in ::set_position */ + if (!video && audio) { audio->modify_trim_start (t); }