Previously the code did not account for referenced audio, so far

as I can see.  It decided which reel to write new audio to based on
how many frames had been written to the current reel; this makes
no sense for referred reels for which the player will emit no audio.

This code looks at the audio timestamp instead.

1 files changed, 1 insertions, 2 deletions

diff --git a/src/lib/dcp_encoder.cc b/src/lib/dcp_encoder.cc
index 81464c8dc..76bfba5a2 100644
--- a/src/lib/dcp_encoder.cc
+++ b/src/lib/dcp_encoder.cc
@@ -121,11 +121,10 @@ DCPEncoder::video (shared_ptr<PlayerVideo> data, DCPTime time)
 	_j2k_encoder->encode (data, time);
 }
 
-/** The audio data passed into this method must be contiguous and start from the last accurate seek time */
 void
 DCPEncoder::audio (shared_ptr<AudioBuffers> data, DCPTime time)
 {
-	_writer->write (data);
+	_writer->write (data, time);
 
 	shared_ptr<Job> job = _job.lock ();
 	DCPOMATIC_ASSERT (job);