]> git.sesse.net Git - ffmpeg/blobdiff - libavformat/matroskadec.c
Use full internal pathname in doxygen @file directives.
[ffmpeg] / libavformat / matroskadec.c
index d7a1e39846c4cc278b7b6eab69a7818e6b9c0d7b..7751db88eb131d22f5ceedc9c4dc4c22d24b2418 100644 (file)
@@ -20,7 +20,7 @@
  */
 
 /**
- * @file matroskadec.c
+ * @file libavformat/matroskadec.c
  * Matroska file demuxer
  * by Ronald Bultje <rbultje@ronald.bitfreak.net>
  * with a little help from Moritz Bunkus <moritz@bunkus.org>
@@ -28,6 +28,7 @@
  * Specs available on the Matroska project page: http://www.matroska.org/.
  */
 
+#include <stdio.h>
 #include "avformat.h"
 /* For codec_get_id(). */
 #include "riff.h"
 #include "matroska.h"
 #include "libavcodec/mpeg4audio.h"
 #include "libavutil/intfloat_readwrite.h"
+#include "libavutil/intreadwrite.h"
 #include "libavutil/avstring.h"
 #include "libavutil/lzo.h"
-#ifdef CONFIG_ZLIB
+#if CONFIG_ZLIB
 #include <zlib.h>
 #endif
-#ifdef CONFIG_BZLIB
+#if CONFIG_BZLIB
 #include <bzlib.h>
 #endif
 
@@ -138,6 +140,7 @@ typedef struct {
     EbmlList encodings;
 
     AVStream *stream;
+    int64_t end_timecode;
 } MatroskaTrack;
 
 typedef struct {
@@ -163,6 +166,12 @@ typedef struct {
     EbmlList pos;
 } MatroskaIndex;
 
+typedef struct {
+    char *name;
+    char *string;
+    EbmlList sub;
+} MatroskaTag;
+
 typedef struct {
     uint64_t id;
     uint64_t pos;
@@ -188,21 +197,23 @@ typedef struct {
     EbmlList attachments;
     EbmlList chapters;
     EbmlList index;
+    EbmlList tags;
     EbmlList seekhead;
 
     /* byte position of the segment inside the stream */
-    offset_t segment_start;
+    int64_t segment_start;
 
     /* the packet queue */
     AVPacket **packets;
     int num_packets;
+    AVPacket *prev_pkt;
 
     int done;
     int has_cluster_id;
 
     /* What to skip before effectively reading a packet. */
     int skip_to_keyframe;
-    AVStream *skip_to_stream;
+    uint64_t skip_to_timecode;
 } MatroskaDemuxContext;
 
 typedef struct {
@@ -216,8 +227,6 @@ typedef struct {
     EbmlList blocks;
 } MatroskaCluster;
 
-#define ARRAY_SIZE(x)  (sizeof(x)/sizeof(*x))
-
 static EbmlSyntax ebml_header[] = {
     { EBML_ID_EBMLREADVERSION,        EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
     { EBML_ID_EBMLMAXSIZELENGTH,      EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
@@ -226,7 +235,6 @@ static EbmlSyntax ebml_header[] = {
     { EBML_ID_DOCTYPEREADVERSION,     EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
     { EBML_ID_EBMLVERSION,            EBML_NONE },
     { EBML_ID_DOCTYPEVERSION,         EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -243,7 +251,6 @@ static EbmlSyntax matroska_info[] = {
     { MATROSKA_ID_MUXINGAPP,          EBML_NONE },
     { MATROSKA_ID_DATEUTC,            EBML_NONE },
     { MATROSKA_ID_SEGMENTUID,         EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -254,10 +261,14 @@ static EbmlSyntax matroska_track_video[] = {
     { MATROSKA_ID_VIDEOPIXELWIDTH,    EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
     { MATROSKA_ID_VIDEOPIXELHEIGHT,   EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
     { MATROSKA_ID_VIDEOCOLORSPACE,    EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
+    { MATROSKA_ID_VIDEOPIXELCROPB,    EBML_NONE },
+    { MATROSKA_ID_VIDEOPIXELCROPT,    EBML_NONE },
+    { MATROSKA_ID_VIDEOPIXELCROPL,    EBML_NONE },
+    { MATROSKA_ID_VIDEOPIXELCROPR,    EBML_NONE },
+    { MATROSKA_ID_VIDEODISPLAYUNIT,   EBML_NONE },
     { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
     { MATROSKA_ID_VIDEOSTEREOMODE,    EBML_NONE },
     { MATROSKA_ID_VIDEOASPECTRATIO,   EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -266,14 +277,12 @@ static EbmlSyntax matroska_track_audio[] = {
     { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
     { MATROSKA_ID_AUDIOBITDEPTH,      EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
     { MATROSKA_ID_AUDIOCHANNELS,      EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_track_encoding_compression[] = {
     { MATROSKA_ID_ENCODINGCOMPALGO,   EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
     { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -281,13 +290,12 @@ static EbmlSyntax matroska_track_encoding[] = {
     { MATROSKA_ID_ENCODINGSCOPE,      EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
     { MATROSKA_ID_ENCODINGTYPE,       EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
     { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_ENCODINGORDER,      EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_track_encodings[] = {
     { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -314,13 +322,12 @@ static EbmlSyntax matroska_track[] = {
     { MATROSKA_ID_CODECDOWNLOADURL,     EBML_NONE },
     { MATROSKA_ID_TRACKMINCACHE,        EBML_NONE },
     { MATROSKA_ID_TRACKMAXCACHE,        EBML_NONE },
-    { EBML_ID_VOID,                     EBML_NONE },
+    { MATROSKA_ID_TRACKMAXBLKADDID,     EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_tracks[] = {
     { MATROSKA_ID_TRACKENTRY,         EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -328,20 +335,19 @@ static EbmlSyntax matroska_attachment[] = {
     { MATROSKA_ID_FILENAME,           EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
     { MATROSKA_ID_FILEMIMETYPE,       EBML_STR,  0, offsetof(MatroskaAttachement,mime) },
     { MATROSKA_ID_FILEDATA,           EBML_BIN,  0, offsetof(MatroskaAttachement,bin) },
+    { MATROSKA_ID_FILEDESC,           EBML_NONE },
     { MATROSKA_ID_FILEUID,            EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_attachments[] = {
     { MATROSKA_ID_ATTACHEDFILE,       EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_chapter_display[] = {
     { MATROSKA_ID_CHAPSTRING,         EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_CHAPLANG,           EBML_NONE },
     { 0 }
 };
 
@@ -351,7 +357,9 @@ static EbmlSyntax matroska_chapter_entry[] = {
     { MATROSKA_ID_CHAPTERUID,         EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
     { MATROSKA_ID_CHAPTERDISPLAY,     EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
     { MATROSKA_ID_CHAPTERFLAGHIDDEN,  EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
+    { MATROSKA_ID_CHAPTERPHYSEQUIV,   EBML_NONE },
+    { MATROSKA_ID_CHAPTERATOM,        EBML_NONE },
     { 0 }
 };
 
@@ -360,51 +368,61 @@ static EbmlSyntax matroska_chapter[] = {
     { MATROSKA_ID_EDITIONUID,         EBML_NONE },
     { MATROSKA_ID_EDITIONFLAGHIDDEN,  EBML_NONE },
     { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_chapters[] = {
     { MATROSKA_ID_EDITIONENTRY,       EBML_NEST, 0, 0, {.n=matroska_chapter} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_index_pos[] = {
     { MATROSKA_ID_CUETRACK,           EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
     { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos)   },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_CUEBLOCKNUMBER,     EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_index_entry[] = {
     { MATROSKA_ID_CUETIME,            EBML_UINT, 0, offsetof(MatroskaIndex,time) },
     { MATROSKA_ID_CUETRACKPOSITION,   EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_index[] = {
     { MATROSKA_ID_POINTENTRY,         EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
-    { EBML_ID_VOID,                   EBML_NONE },
+    { 0 }
+};
+
+static EbmlSyntax matroska_simpletag[] = {
+    { MATROSKA_ID_TAGNAME,            EBML_UTF8, 0, offsetof(MatroskaTag,name) },
+    { MATROSKA_ID_TAGSTRING,          EBML_UTF8, 0, offsetof(MatroskaTag,string) },
+    { MATROSKA_ID_SIMPLETAG,          EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
+    { MATROSKA_ID_TAGLANG,            EBML_NONE },
+    { MATROSKA_ID_TAGDEFAULT,         EBML_NONE },
+    { 0 }
+};
+
+static EbmlSyntax matroska_tag[] = {
+    { MATROSKA_ID_SIMPLETAG,          EBML_NEST, sizeof(MatroskaTag), 0, {.n=matroska_simpletag} },
+    { MATROSKA_ID_TAGTARGETS,         EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_tags[] = {
-    { EBML_ID_VOID,                   EBML_NONE },
+    { MATROSKA_ID_TAG,                EBML_NEST, 0, offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
     { 0 }
 };
 
 static EbmlSyntax matroska_seekhead_entry[] = {
     { MATROSKA_ID_SEEKID,             EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
     { MATROSKA_ID_SEEKPOSITION,       EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_seekhead[] = {
     { MATROSKA_ID_SEEKENTRY,          EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
-    { EBML_ID_VOID,                   EBML_NONE },
     { 0 }
 };
 
@@ -417,7 +435,6 @@ static EbmlSyntax matroska_segment[] = {
     { MATROSKA_ID_TAGS,           EBML_NEST, 0, 0, {.n=matroska_tags       } },
     { MATROSKA_ID_SEEKHEAD,       EBML_NEST, 0, 0, {.n=matroska_seekhead   } },
     { MATROSKA_ID_CLUSTER,        EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
-    { EBML_ID_VOID,               EBML_NONE },
     { 0 }
 };
 
@@ -431,7 +448,6 @@ static EbmlSyntax matroska_blockgroup[] = {
     { MATROSKA_ID_SIMPLEBLOCK,    EBML_BIN,  0, offsetof(MatroskaBlock,bin) },
     { MATROSKA_ID_BLOCKDURATION,  EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
     { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
-    { EBML_ID_VOID,               EBML_NONE },
     { 0 }
 };
 
@@ -439,22 +455,46 @@ static EbmlSyntax matroska_cluster[] = {
     { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
     { MATROSKA_ID_BLOCKGROUP,     EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
     { MATROSKA_ID_SIMPLEBLOCK,    EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
-    { EBML_ID_VOID,               EBML_NONE },
+    { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
+    { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
     { 0 }
 };
 
 static EbmlSyntax matroska_clusters[] = {
     { MATROSKA_ID_CLUSTER,        EBML_NEST, 0, 0, {.n=matroska_cluster} },
+    { MATROSKA_ID_INFO,           EBML_NONE },
+    { MATROSKA_ID_CUES,           EBML_NONE },
+    { MATROSKA_ID_TAGS,           EBML_NONE },
+    { MATROSKA_ID_SEEKHEAD,       EBML_NONE },
     { 0 }
 };
 
+#define SIZE_OFF(x) sizeof(((AVFormatContext*)0)->x),offsetof(AVFormatContext,x)
+const struct {
+    const char name[16];
+    int   size;
+    int   offset;
+} metadata[] = {
+    { "TITLE",           SIZE_OFF(title)      },
+    { "ARTIST",          SIZE_OFF(author)     },
+    { "WRITTEN_BY",      SIZE_OFF(author)     },
+    { "LEAD_PERFORMER",  SIZE_OFF(author)     },
+    { "COPYRIGHT",       SIZE_OFF(copyright)  },
+    { "COMMENT",         SIZE_OFF(comment)    },
+    { "ALBUM",           SIZE_OFF(album)      },
+    { "DATE_WRITTEN",    SIZE_OFF(year)       },
+    { "DATE_RELEASED",   SIZE_OFF(year)       },
+    { "PART_NUMBER",     SIZE_OFF(track)      },
+    { "GENRE",           SIZE_OFF(genre)      },
+};
+
 /*
  * Return: Whether we reached the end of a level in the hierarchy or not.
  */
 static int ebml_level_end(MatroskaDemuxContext *matroska)
 {
     ByteIOContext *pb = matroska->ctx->pb;
-    offset_t pos = url_ftell(pb);
+    int64_t pos = url_ftell(pb);
 
     if (matroska->num_levels > 0) {
         MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
@@ -486,7 +526,7 @@ static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
     if (!(total = get_byte(pb))) {
         /* we might encounter EOS here */
         if (!url_feof(pb)) {
-            offset_t pos = url_ftell(pb);
+            int64_t pos = url_ftell(pb);
             av_log(matroska->ctx, AV_LOG_ERROR,
                    "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
                    pos, pos);
@@ -500,7 +540,7 @@ static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
         len_mask >>= 1;
     }
     if (read > max_size) {
-        offset_t pos = url_ftell(pb) - 1;
+        int64_t pos = url_ftell(pb) - 1;
         av_log(matroska->ctx, AV_LOG_ERROR,
                "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
                (uint8_t) total, pos, pos);
@@ -654,7 +694,7 @@ static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
     for (i=0; syntax[i].id; i++)
         if (id == syntax[i].id)
             break;
-    if (!syntax[i].id)
+    if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
         av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
     return ebml_parse_elem(matroska, &syntax[i], data);
 }
@@ -727,7 +767,7 @@ static int ebml_parse_elem(MatroskaDemuxContext *matroska,
                      return ebml_parse_nest(matroska, syntax->def.n, data);
     case EBML_PASS:  return ebml_parse_id(matroska, syntax->def.n, id, data);
     case EBML_STOP:  *(int *)data = 1;      return 1;
-    default:         url_fskip(pb, length); return 0;
+    default:         return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
     }
     if (res == AVERROR_INVALIDDATA)
         av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
@@ -767,7 +807,7 @@ static int matroska_probe(AVProbeData *p)
 {
     uint64_t total = 0;
     int len_mask = 0x80, size = 1, n = 1;
-    char probe_data[] = "matroska";
+    static const char probe_data[] = "matroska";
 
     /* EBML header? */
     if (AV_RB32(p->buf) != EBML_ID_HEADER)
@@ -839,7 +879,7 @@ static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
             goto failed;
         pkt_size -= olen;
         break;
-#ifdef CONFIG_ZLIB
+#if CONFIG_ZLIB
     case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
         z_stream zstream = {0};
         if (inflateInit(&zstream) != Z_OK)
@@ -860,7 +900,7 @@ static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
         break;
     }
 #endif
-#ifdef CONFIG_BZLIB
+#if CONFIG_BZLIB
     case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
         bz_stream bzstream = {0};
         if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
@@ -881,6 +921,8 @@ static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
         break;
     }
 #endif
+    default:
+        return -1;
     }
 
     *buf = pkt_data;
@@ -891,17 +933,78 @@ static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
     return -1;
 }
 
+static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
+                                    AVPacket *pkt, uint64_t display_duration)
+{
+    char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
+    for (; *ptr!=',' && ptr<end-1; ptr++);
+    if (*ptr == ',')
+        layer = ++ptr;
+    for (; *ptr!=',' && ptr<end-1; ptr++);
+    if (*ptr == ',') {
+        int64_t end_pts = pkt->pts + display_duration;
+        int sc = matroska->time_scale * pkt->pts / 10000000;
+        int ec = matroska->time_scale * end_pts  / 10000000;
+        int sh, sm, ss, eh, em, es, len;
+        sh = sc/360000;  sc -= 360000*sh;
+        sm = sc/  6000;  sc -=   6000*sm;
+        ss = sc/   100;  sc -=    100*ss;
+        eh = ec/360000;  ec -= 360000*eh;
+        em = ec/  6000;  ec -=   6000*em;
+        es = ec/   100;  ec -=    100*es;
+        *ptr++ = '\0';
+        len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
+        if (!(line = av_malloc(len)))
+            return;
+        snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
+                 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
+        av_free(pkt->data);
+        pkt->data = line;
+        pkt->size = strlen(line);
+    }
+}
+
+static void matroska_merge_packets(AVPacket *out, AVPacket *in)
+{
+    out->data = av_realloc(out->data, out->size+in->size);
+    memcpy(out->data+out->size, in->data, in->size);
+    out->size += in->size;
+    av_destruct_packet(in);
+    av_free(in);
+}
+
+static void matroska_convert_tags(AVFormatContext *s, EbmlList *list)
+{
+    MatroskaTag *tags = list->elem;
+    int i, j;
+
+    for (i=0; i < list->nb_elem; i++) {
+        for (j=0; j < FF_ARRAY_ELEMS(metadata); j++){
+            if (!strcmp(tags[i].name, metadata[j].name)) {
+                int *ptr = (int *)((char *)s + metadata[j].offset);
+                if (*ptr)  continue;
+                if (metadata[j].size > sizeof(int))
+                    av_strlcpy((char *)ptr, tags[i].string, metadata[j].size);
+                else
+                    *ptr = atoi(tags[i].string);
+            }
+        }
+        if (tags[i].sub.nb_elem)
+            matroska_convert_tags(s, &tags[i].sub);
+    }
+}
+
 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
 {
     EbmlList *seekhead_list = &matroska->seekhead;
     MatroskaSeekhead *seekhead = seekhead_list->elem;
     uint32_t level_up = matroska->level_up;
-    offset_t before_pos = url_ftell(matroska->ctx->pb);
+    int64_t before_pos = url_ftell(matroska->ctx->pb);
     MatroskaLevel level;
     int i;
 
     for (i=0; i<seekhead_list->nb_elem; i++) {
-        offset_t offset = seekhead[i].pos + matroska->segment_start;
+        int64_t offset = seekhead[i].pos + matroska->segment_start;
 
         if (seekhead[i].pos <= before_pos
             || seekhead[i].id == MATROSKA_ID_SEEKHEAD
@@ -943,10 +1046,10 @@ static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
 
 static int matroska_aac_profile(char *codec_id)
 {
-    static const char *aac_profiles[] = { "MAIN", "LC", "SSR" };
+    static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
     int profile;
 
-    for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
+    for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
         if (strstr(codec_id, aac_profiles[profile]))
             break;
     return profile + 1;
@@ -956,7 +1059,7 @@ static int matroska_aac_sri(int samplerate)
 {
     int sri;
 
-    for (sri=0; sri<ARRAY_SIZE(ff_mpeg4audio_sample_rates); sri++)
+    for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
         if (ff_mpeg4audio_sample_rates[sri] == samplerate)
             break;
     return sri;
@@ -972,6 +1075,8 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
     MatroskaTrack *tracks;
     EbmlList *index_list;
     MatroskaIndex *index;
+    int index_scale = 1;
+    uint64_t max_start = 0;
     Ebml ebml = { 0 };
     AVStream *st;
     int i, j;
@@ -1002,6 +1107,7 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
     if (matroska->title)
         strncpy(matroska->ctx->title, matroska->title,
                 sizeof(matroska->ctx->title)-1);
+    matroska_convert_tags(s, &matroska->tags);
 
     tracks = matroska->tracks.elem;
     for (i=0; i < matroska->tracks.nb_elem; i++) {
@@ -1012,6 +1118,7 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
         uint8_t *extradata = NULL;
         int extradata_size = 0;
         int extradata_offset = 0;
+        ByteIOContext b;
 
         /* Apply some sanity checks. */
         if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
@@ -1042,10 +1149,10 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
         } else if (encodings_list->nb_elem == 1) {
             if (encodings[0].type ||
                 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
-#ifdef CONFIG_ZLIB
+#if CONFIG_ZLIB
                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
 #endif
-#ifdef CONFIG_BZLIB
+#if CONFIG_BZLIB
                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
 #endif
                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
@@ -1095,13 +1202,31 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
         } else if (!strcmp(track->codec_id, "A_MS/ACM")
                    && track->codec_priv.size >= 18
                    && track->codec_priv.data != NULL) {
-            uint16_t tag = AV_RL16(track->codec_priv.data);
-            codec_id = codec_get_id(codec_wav_tags, tag);
+            init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
+                          URL_RDONLY, NULL, NULL, NULL, NULL);
+            get_wav_header(&b, st->codec, track->codec_priv.size);
+            codec_id = st->codec->codec_id;
+            extradata_offset = 18;
+            track->codec_priv.size -= extradata_offset;
         } else if (!strcmp(track->codec_id, "V_QUICKTIME")
                    && (track->codec_priv.size >= 86)
                    && (track->codec_priv.data != NULL)) {
             track->video.fourcc = AV_RL32(track->codec_priv.data);
             codec_id=codec_get_id(codec_movvideo_tags, track->video.fourcc);
+        } else if (codec_id == CODEC_ID_PCM_S16BE) {
+            switch (track->audio.bitdepth) {
+            case  8:  codec_id = CODEC_ID_PCM_U8;     break;
+            case 24:  codec_id = CODEC_ID_PCM_S24BE;  break;
+            case 32:  codec_id = CODEC_ID_PCM_S32BE;  break;
+            }
+        } else if (codec_id == CODEC_ID_PCM_S16LE) {
+            switch (track->audio.bitdepth) {
+            case  8:  codec_id = CODEC_ID_PCM_U8;     break;
+            case 24:  codec_id = CODEC_ID_PCM_S24LE;  break;
+            case 32:  codec_id = CODEC_ID_PCM_S32LE;  break;
+            }
+        } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
+            codec_id = CODEC_ID_PCM_F64LE;
         } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
             int profile = matroska_aac_profile(track->codec_id);
             int sri = matroska_aac_sri(track->audio.samplerate);
@@ -1119,7 +1244,6 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             } else
                 extradata_size = 2;
         } else if (codec_id == CODEC_ID_TTA) {
-            ByteIOContext b;
             extradata_size = 30;
             extradata = av_mallocz(extradata_size);
             if (extradata == NULL)
@@ -1141,8 +1265,6 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             track->audio.channels = 1;
         } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
                    codec_id == CODEC_ID_ATRAC3) {
-            ByteIOContext b;
-
             init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
                           0, NULL, NULL, NULL, NULL);
             url_fskip(&b, 24);
@@ -1166,6 +1288,8 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             av_log(matroska->ctx, AV_LOG_INFO,
                    "Unknown/unsupported CodecID %s.\n", track->codec_id);
 
+        if (track->time_scale < 0.01)
+            track->time_scale = 1.0;
         av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
 
         st->codec->codec_id = codec_id;
@@ -1184,7 +1308,8 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             st->codec->extradata = extradata;
             st->codec->extradata_size = extradata_size;
         } else if(track->codec_priv.data && track->codec_priv.size > 0){
-            st->codec->extradata = av_malloc(track->codec_priv.size);
+            st->codec->extradata = av_mallocz(track->codec_priv.size +
+                                              FF_INPUT_BUFFER_PADDING_SIZE);
             if(st->codec->extradata == NULL)
                 return AVERROR(ENOMEM);
             st->codec->extradata_size = track->codec_priv.size;
@@ -1198,8 +1323,8 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             st->codec->codec_tag  = track->video.fourcc;
             st->codec->width  = track->video.pixel_width;
             st->codec->height = track->video.pixel_height;
-            av_reduce(&st->codec->sample_aspect_ratio.num,
-                      &st->codec->sample_aspect_ratio.den,
+            av_reduce(&st->sample_aspect_ratio.num,
+                      &st->sample_aspect_ratio.den,
                       st->codec->height * track->video.display_width,
                       st->codec-> width * track->video.display_height,
                       255);
@@ -1243,13 +1368,21 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
 
     chapters = chapters_list->elem;
     for (i=0; i<chapters_list->nb_elem; i++)
-        if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid)
+        if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
+            && (max_start==0 || chapters[i].start > max_start)) {
             ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
                            chapters[i].start, chapters[i].end,
                            chapters[i].title);
+            max_start = chapters[i].start;
+        }
 
     index_list = &matroska->index;
     index = index_list->elem;
+    if (index_list->nb_elem
+        && index[0].time > 100000000000000/matroska->time_scale) {
+        av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
+        index_scale = matroska->time_scale;
+    }
     for (i=0; i<index_list->nb_elem; i++) {
         EbmlList *pos_list = &index[i].pos;
         MatroskaIndexPos *pos = pos_list->elem;
@@ -1259,8 +1392,8 @@ static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
             if (track && track->stream)
                 av_add_index_entry(track->stream,
                                    pos[j].pos + matroska->segment_start,
-                                   index[i].time*matroska->time_scale/AV_TIME_BASE,
-                                   0, 0, AVINDEX_KEYFRAME);
+                                   index[i].time/index_scale, 0, 0,
+                                   AVINDEX_KEYFRAME);
         }
     }
 
@@ -1311,8 +1444,10 @@ static void matroska_clear_queue(MatroskaDemuxContext *matroska)
 
 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
                                 int size, int64_t pos, uint64_t cluster_time,
-                                uint64_t duration, int is_keyframe)
+                                uint64_t duration, int is_keyframe,
+                                int64_t cluster_pos)
 {
+    uint64_t timecode = AV_NOPTS_VALUE;
     MatroskaTrack *track;
     int res = 0;
     AVStream *st;
@@ -1348,8 +1483,19 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
     if (is_keyframe == -1)
         is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
 
-    if (matroska->skip_to_keyframe) {
-        if (!is_keyframe || st != matroska->skip_to_stream)
+    if (cluster_time != (uint64_t)-1
+        && (block_time >= 0 || cluster_time >= -block_time)) {
+        timecode = cluster_time + block_time;
+        if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
+            && timecode < track->end_timecode)
+            is_keyframe = 0;  /* overlapping subtitles are not key frame */
+        if (is_keyframe)
+            av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
+        track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
+    }
+
+    if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
+        if (!is_keyframe || timecode < matroska->skip_to_timecode)
             return res;
         matroska->skip_to_keyframe = 0;
     }
@@ -1431,12 +1577,6 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
     }
 
     if (res == 0) {
-        uint64_t timecode = AV_NOPTS_VALUE;
-
-        if (cluster_time != (uint64_t)-1
-            && (block_time >= 0 || cluster_time >= -block_time))
-            timecode = cluster_time + block_time;
-
         for (n = 0; n < laces; n++) {
             if (st->codec->codec_id == CODEC_ID_RA_288 ||
                 st->codec->codec_id == CODEC_ID_COOK ||
@@ -1504,9 +1644,23 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
 
                 pkt->pts = timecode;
                 pkt->pos = pos;
-                pkt->duration = duration;
-
-                dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
+                if (st->codec->codec_id == CODEC_ID_TEXT)
+                    pkt->convergence_duration = duration;
+                else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
+                    pkt->duration = duration;
+
+                if (st->codec->codec_id == CODEC_ID_SSA)
+                    matroska_fix_ass_packet(matroska, pkt, duration);
+
+                if (matroska->prev_pkt &&
+                    timecode != AV_NOPTS_VALUE &&
+                    matroska->prev_pkt->pts == timecode &&
+                    matroska->prev_pkt->stream_index == st->index)
+                    matroska_merge_packets(matroska->prev_pkt, pkt);
+                else {
+                    dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
+                    matroska->prev_pkt = pkt;
+                }
             }
 
             if (timecode != AV_NOPTS_VALUE)
@@ -1525,23 +1679,28 @@ static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
     EbmlList *blocks_list;
     MatroskaBlock *blocks;
     int i, res;
+    int64_t pos = url_ftell(matroska->ctx->pb);
+    matroska->prev_pkt = NULL;
     if (matroska->has_cluster_id){
         /* For the first cluster we parse, its ID was already read as
            part of matroska_read_header(), so don't read it again */
         res = ebml_parse_id(matroska, matroska_clusters,
                             MATROSKA_ID_CLUSTER, &cluster);
+        pos -= 4;  /* sizeof the ID which was already read */
         matroska->has_cluster_id = 0;
     } else
         res = ebml_parse(matroska, matroska_clusters, &cluster);
     blocks_list = &cluster.blocks;
     blocks = blocks_list->elem;
-    for (i=0; !res && i<blocks_list->nb_elem; i++)
+    for (i=0; i<blocks_list->nb_elem; i++)
         if (blocks[i].bin.size > 0)
             res=matroska_parse_block(matroska,
                                      blocks[i].bin.data, blocks[i].bin.size,
                                      blocks[i].bin.pos,  cluster.timecode,
-                                     blocks[i].duration, !blocks[i].reference);
+                                     blocks[i].duration, !blocks[i].reference,
+                                     pos);
     ebml_free(matroska_cluster, &cluster);
+    if (res < 0)  matroska->done = 1;
     return res;
 }
 
@@ -1552,8 +1711,7 @@ static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
     while (matroska_deliver_packet(matroska, pkt)) {
         if (matroska->done)
             return AVERROR(EIO);
-        if (matroska_parse_cluster(matroska) < 0)
-            matroska->done = 1;
+        matroska_parse_cluster(matroska);
     }
 
     return 0;
@@ -1563,18 +1721,44 @@ static int matroska_read_seek(AVFormatContext *s, int stream_index,
                               int64_t timestamp, int flags)
 {
     MatroskaDemuxContext *matroska = s->priv_data;
+    MatroskaTrack *tracks = matroska->tracks.elem;
     AVStream *st = s->streams[stream_index];
-    int index;
+    int i, index, index_sub, index_min;
 
-    index = av_index_search_timestamp(st, timestamp, flags);
-    if (index < 0)
+    if (!st->nb_index_entries)
         return 0;
+    timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
+
+    if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
+        url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
+        while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
+            matroska_clear_queue(matroska);
+            if (matroska_parse_cluster(matroska) < 0)
+                break;
+        }
+    }
 
     matroska_clear_queue(matroska);
+    if (index < 0)
+        return 0;
+
+    index_min = index;
+    for (i=0; i < matroska->tracks.nb_elem; i++) {
+        tracks[i].end_timecode = 0;
+        if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
+            && !tracks[i].stream->discard != AVDISCARD_ALL) {
+            index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
+            if (index_sub >= 0
+                && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
+                && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
+                index_min = index_sub;
+        }
+    }
 
-    url_fseek(s->pb, st->index_entries[index].pos, SEEK_SET);
+    url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
     matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
-    matroska->skip_to_stream = st;
+    matroska->skip_to_timecode = st->index_entries[index].timestamp;
+    matroska->done = 0;
     av_update_cur_dts(s, st, st->index_entries[index].timestamp);
     return 0;
 }