Added support for iTunSMPB to MP3 parser; Replaced crappy MP3 seek function with accurate full file parse with 8 packet pre-roll; Changed top level decoder plugin to truncate output at the specified duration

Frameworks/FFMPEG/ffmpeg-minimal/libavformat/id3v2.c

@@ -309,6 +309,47 @@ static void read_ttag(AVFormatContext *s, AVIOContext *pb, int taglen,
         av_dict_set(metadata, key, (const char *) dst, dict_flags);
 }
 
+/**
+ * Parse a comment tag.
+ */
+static void read_comm(AVFormatContext *s, AVIOContext *pb, int taglen,
+                      AVDictionary **metadata)
+{
+    const char *key = "comment";
+    uint8_t *dst;
+    int encoding, dict_flags = AV_DICT_DONT_OVERWRITE | AV_DICT_DONT_STRDUP_VAL;
+    int language;
+    
+    if (taglen < 4)
+        return;
+    
+    encoding = avio_r8(pb);
+    taglen--;
+    
+    language = avio_rl24(pb);
+    taglen -= 3;
+    
+    if (decode_str(s, pb, encoding, &dst, &taglen) < 0) {
+        av_log(s, AV_LOG_ERROR, "Error reading comment frame, skipped\n");
+        return;
+    }
+    
+    if (dst && dst[0]) {
+        key = (const char *) dst;
+        dict_flags |= AV_DICT_DONT_STRDUP_KEY;
+    }
+    
+    if (decode_str(s, pb, encoding, &dst, &taglen) < 0) {
+        av_log(s, AV_LOG_ERROR, "Error reading comment frame, skipped\n");
+        if (dict_flags & AV_DICT_DONT_STRDUP_KEY)
+            av_freep((void*)&key);
+        return;
+    }
+    
+    if (dst)
+        av_dict_set(metadata, key, (const char *) dst, dict_flags);
+}
+
 /**
  * Parse GEOB tag into a ID3v2ExtraMetaGEOB struct.
  */
@@ -720,7 +761,7 @@ static void id3v2_parse(AVFormatContext *s, int len, uint8_t version,
             av_log(s, AV_LOG_WARNING, "Skipping %s ID3v2 frame %s.\n", type, tag);
             avio_skip(s->pb, tlen);
         /* check for text tag or supported special meta tag */
-        } else if (tag[0] == 'T' ||
+        } else if (tag[0] == 'T' || memcmp(tag, "COMM", 4) == 0 ||
                    (extra_meta &&
                     (extra_func = get_extra_meta_func(tag, isv34)))) {
             pbx = s->pb;
@@ -786,6 +827,8 @@ static void id3v2_parse(AVFormatContext *s, int len, uint8_t version,
             if (tag[0] == 'T')
                 /* parse text tag */
                 read_ttag(s, pbx, tlen, &s->metadata, tag);
+            else if (memcmp(tag, "COMM", 4) == 0)
+                read_comm(s, pbx, tlen, &s->metadata);
             else
                 /* parse special meta tag */
                 extra_func->read(s, pbx, tlen, tag, extra_meta);

Frameworks/FFMPEG/ffmpeg-minimal/libavformat/mp3dec.c

@@ -129,6 +129,8 @@ static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
     MPADecodeHeader c;
     int vbrtag_size = 0;
     int is_cbr;
+    AVDictionaryEntry *de;
+    uint64_t duration = 0;
 
     v = avio_rb32(s->pb);
     if(ff_mpa_check_header(v) < 0)
@@ -139,6 +141,9 @@ static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
     if(c.layer != 3)
         return -1;
     
+    mp3->start_pad = 0;
+    mp3->end_pad = 0;
+    
     spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
 
     /* Check for Xing / Info tag */
@@ -164,6 +169,8 @@ static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
             mp3->start_pad = v>>12;
             mp3->  end_pad = v&4095;
             st->skip_samples = mp3->start_pad + 528 + 1;
+            if (mp3->end_pad >= 528 + 1)
+                mp3->end_pad -= 528 + 1;
             av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3->  end_pad);
         }
     }
@@ -181,17 +188,57 @@ static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
         }
     }
     
-    if(!frames && !size)
+    if (!frames)
+        vbrtag_size = 0;
+
+    if (s->metadata && (de = av_dict_get(s->metadata, "iTunSMPB", NULL, 0))) {
+        uint32_t zero, start_pad, end_pad;
+        uint64_t last_eight_frames_offset;
+        if (sscanf(de->value, "%x %x %x %llx %x %llx", &zero, &start_pad, &end_pad, &duration, &zero, &last_eight_frames_offset) < 6) {
+            duration = 0;
+        }
+        else {
+            mp3->start_pad = start_pad;
+            mp3->end_pad = end_pad;
+            if (end_pad >= 528 + 1)
+                mp3->end_pad = end_pad - (528 + 1);
+            st->skip_samples = mp3->start_pad + 528 + 1;
+            av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3->end_pad);
+            if (s->pb->seekable) {
+                int i;
+                size = last_eight_frames_offset;
+                avio_seek(s->pb, base + vbrtag_size + last_eight_frames_offset, SEEK_SET);
+                for (i = 0; i < 8; ++i) {
+                    v = avio_rb32(s->pb);
+                    if (ff_mpa_check_header(v) < 0)
+                        return -1;
+                    if (avpriv_mpegaudio_decode_header(&c, v) != 0)
+                        break;
+                    size += c.frame_size;
+                    avio_skip(s->pb, c.frame_size - 4);
+                }
+            }
+        }
+    }
+    
+    if(!frames && !size && !duration)
         return -1;
 
     /* Skip the vbr tag frame */
     avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
 
-    if(frames)
+    if (duration)
+        st->duration = av_rescale_q(duration, (AVRational){1, c.sample_rate}, st->time_base);
+    else if(frames)
         st->duration = av_rescale_q(frames, (AVRational){spf, c.sample_rate},
-                                    st->time_base);
+                                    st->time_base) - av_rescale_q(mp3->end_pad, (AVRational){1, c.sample_rate}, st->time_base);
-    if (size && frames && !is_cbr)
+    
-        st->codec->bit_rate = av_rescale(size, 8 * c.sample_rate, frames * (int64_t)spf);
+    if (size) {
+        if (duration)
+            st->codec->bit_rate = av_rescale(size, 8 * c.sample_rate, duration);
+        else if (frames)
+            st->codec->bit_rate = av_rescale(size, 8 * c.sample_rate, frames * (int64_t)spf);
+    }
 
     mp3->is_cbr          = is_cbr;
     mp3->header_filesize = size;
@@ -285,57 +332,58 @@ static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
                     int flags)
 {
     MP3DecContext *mp3 = s->priv_data;
-    AVIndexEntry *ie, ie1;
     AVStream *st = s->streams[0];
-    int64_t ret  = av_index_search_timestamp(st, timestamp, flags);
+    int64_t timestamp_samples = 0;
-    int i, j;
+    uint32_t v, spf;
 
-    if (mp3->is_cbr && st->duration > 0 && mp3->header_filesize > s->data_offset) {
+    timestamp = av_clip64(timestamp, 0, st->duration);
-        int64_t filesize = avio_size(s->pb);
-        int64_t duration;
-        if (filesize <= s->data_offset)
-            filesize = mp3->header_filesize;
-        filesize -= s->data_offset;
-        duration = av_rescale(st->duration, filesize, mp3->header_filesize - s->data_offset);
-        ie = &ie1;
-        timestamp = av_clip64(timestamp, 0, duration);
-        ie->timestamp = timestamp;
-        ie->pos       = av_rescale(timestamp, filesize, duration) + s->data_offset;
-    } else if (mp3->xing_toc) {
-        if (ret < 0)
-            return ret;
     
-        ie = &st->index_entries[ret];
+    /* Screw it, we're doing a full stream parse! */
-    } else {
+    avio_seek(s->pb, s->data_offset, SEEK_SET);
-        st->skip_samples = timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
     
-        return -1;
+    st->skip_samples = 0;
+    
+    if (timestamp > 0) {
+        int64_t skipped = 0;
+        int64_t skip_extra = 0;
+        do {
+            MPADecodeHeader c;
+            
+            v = avio_rb32(s->pb);
+            avio_seek(s->pb, -4, SEEK_CUR);
+            
+            if(ff_mpa_check_header(v) < 0)
+                return -1;
+            
+            if (avpriv_mpegaudio_decode_header(&c, v) != 0)
+                return -1;
+
+            spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
+            
+            if (!timestamp_samples) {
+                timestamp_samples = av_rescale_q(timestamp, st->time_base, (AVRational){1, c.sample_rate}) + mp3->start_pad + 528 + 1;
+                if (timestamp_samples >= spf * 8 ) {
+                    timestamp_samples -= spf * 8;
+                    skip_extra = spf * 8;
+                }
+                else {
+                    skip_extra = timestamp_samples;
+                    timestamp_samples = 0;
+                }
+            }
+            
+            if ( skipped + spf > timestamp_samples ) break;
+            
+            skipped += spf;
+            
+            avio_skip(s->pb, c.frame_size);
+        } while ( skipped < timestamp_samples );
+        
+        st->skip_samples = timestamp_samples - skipped + skip_extra;
     }
     
-    ret = avio_seek(s->pb, ie->pos, SEEK_SET);
+    ff_update_cur_dts(s, st, timestamp);
-    if (ret < 0)
-        return ret;
     
-#define MIN_VALID 3
-    for(i=0; i<4096; i++) {
-        int64_t pos = ie->pos + i;
-        for(j=0; j<MIN_VALID; j++) {
-            ret = check(s, pos);
-            if(ret < 0)
-                break;
-            pos += ret;
-        }
-        if(j==MIN_VALID)
-            break;
-    }
-    if(j!=MIN_VALID)
-        i=0;
-
-    ret = avio_seek(s->pb, ie->pos + i, SEEK_SET);
-    if (ret < 0)
-        return ret;
-    ff_update_cur_dts(s, st, ie->timestamp);
-    st->skip_samples = ie->timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
     return 0;
 }
 

Plugins/FFMPEG/FFMPEGDecoder.h

@@ -35,6 +35,7 @@
 	float frequency;
 	long totalFrames;
 	long framesPlayed;
+    long framesToSkip;
 }
 
 @end

Plugins/FFMPEG/FFMPEGDecoder.m

@@ -128,7 +128,7 @@ int lockmgr_callback(void ** mutex, enum AVLockOp op)
 	NSLog(@"channels: %d", c->channels);
 	
 	channels = c->channels;
-	bitrate = ic->bit_rate / 1000;
+	bitrate = c->bit_rate / 1000;
     floatingPoint = NO;
     switch (c->sample_fmt) {
         case AV_SAMPLE_FMT_U8:
@@ -162,6 +162,7 @@ int lockmgr_callback(void ** mutex, enum AVLockOp op)
             return NO;
     }
 	totalFrames = c->sample_rate * ((float)ic->duration/AV_TIME_BASE);
+    framesPlayed = 0;
 	frequency = c->sample_rate;
 	seekable = YES;
 	
@@ -204,8 +205,12 @@ int lockmgr_callback(void ** mutex, enum AVLockOp op)
 		}
 		if (frames > 0)
 		{
+            if (framesPlayed >= totalFrames)
+                break;
+            
             size_t sampleBufferOffset = 0;
             
+            
 			if (av_read_frame(ic, &framePacket) < 0)
 			{
 				NSLog(@"Uh oh... av_read_frame returned negative");
@@ -266,6 +271,11 @@ int lockmgr_callback(void ** mutex, enum AVLockOp op)
 			
 			numFrames = sampleBufferOffset / bytesPerFrame;
 			samplePos = 0;
+            
+            if (numFrames + framesPlayed > totalFrames)
+                numFrames = totalFrames - framesPlayed;
+            
+            framesPlayed += numFrames;
         }
 	}
 	
@@ -278,6 +288,8 @@ int lockmgr_callback(void ** mutex, enum AVLockOp op)
 	NSLog(@"frame: %ld", frame);
     AVRational time_base = ic->streams[streamIndex]->time_base;
     av_seek_frame(ic, streamIndex, frame * time_base.den / time_base.num / frequency, 0);
+    numFrames = 0;
+    framesPlayed = frame;
 	return frame;
 }