opus: Handle large comment headers

A comment header may span multiple pages and in such situations we
need to delay parsing until we have read all pages which belong to the
comment header packet. The final page in the comment header packet is
marked by setting "granule position" to zero.

It's common to have comment headers spanning multiple pages when album
art is embedded in an opus file. A single page can contain 65025 bytes
so if tags + picture needs more space then additional pages are
needed.

Author: robho

Changes

@@ -2,6 +2,9 @@ Revision history for Audio::Scan
 
 Note: Bug numbers refer to bugs at http://bugs.slimdevices.com
 
+Upcoming release
+	- opus: Fix parsing large comment headers (such as large embedded images)
+
 1.05    2021-09-10
 	- ID3: Fix v2.4 extended header handling.
 	- WavPack DSD: fix song_length_ms calculation (Kimmo Taskinen).

src/opus.c

@@ -194,7 +194,17 @@ _opus_parse(PerlIO *infile, char *file, HV *info, HV *tags, uint8_t seeking)
     // Copy page into vorbis buffer
     buffer_append( &vorbis_buf, buffer_ptr(&ogg_buf), pagelen );
     DEBUG_TRACE("  Read %d into vorbis buffer\n", pagelen);
-    
+
+    if ( granule_pos != 0 ) {
+      // RFC7845: The granule position MUST be zero for the ID header
+      //          page and the page where the comment header
+      //          completes.
+      //
+      // So, we need to read more pages
+      buffer_consume( &ogg_buf, pagelen );
+      continue;
+    }
+
     // Process vorbis packet
     TOC_byte = buffer_get_char(&vorbis_buf);
     if ( TOC_byte == 'O' ) {

t/opus.t

@@ -2,7 +2,7 @@ use strict;
 
 use File::Spec::Functions;
 use FindBin ();
-use Test::More tests => 74;
+use Test::More tests => 88;
 
 use Audio::Scan;
 
@@ -47,6 +47,29 @@ eval {
     is($info->{audio_md5}, 'bebb6f0f0a90ce4e4e90635a3c7408d0', 'Audio MD5 ok' );
 }
 
+{
+    local $ENV{AUDIO_SCAN_NO_ARTWORK} = 1;
+    my $s = Audio::Scan->scan( _f('large_embedded_picture.opus'), { md5_size => 4096 } );
+
+    my $info = $s->{info};
+    is($info->{bitrate_average}, 72640, 'Bitrate ok');
+    is($info->{channels}, 1, 'Channels ok');
+    is($info->{file_size}, 205606, 'File size ok');
+    is($info->{stereo}, 0, 'Stereo ok');
+    is($info->{samplerate}, 48000, 'Sample Rate ok');
+    is($info->{input_samplerate}, 44100, 'Input Sample Rate ok');
+    is($info->{song_length_ms}, 100, 'Song length ok');
+    is($info->{audio_offset}, 204698, 'Audio offset ok');
+    is($info->{audio_size}, 908, 'Audio size ok');
+    is($info->{audio_md5}, 'c050e1584b6284c230708b8b0fb2f31e', 'Audio MD5 ok');
+
+    my $tags = $s->{tags};
+    is($tags->{ALLPICTURES}[0]{image_data}, 152291, 'Image size ok');
+    is($tags->{ALLPICTURES}[0]{width}, 500, 'Image width ok');
+    is($tags->{ALLPICTURES}[0]{height}, 500, 'Image height ok');
+    is($tags->{ALLPICTURES}[0]{mime_type}, 'image/png', 'Image type ok');
+}
+
 ## A few of the official Opus test files from https://people.xiph.org/~greg/opus_testvectors/
 
 {