]> git.sesse.net Git - ffmpeg/blobdiff - libavformat/oggparsevorbis.c
lavfi: rename vsrc_buffer.c to buffersrc.c
[ffmpeg] / libavformat / oggparsevorbis.c
index ab6c6c74d8ddca12225a3f356b28fb896ee3847f..f72fd26c4c68912b2eb4a91cba4a45926874de04 100644 (file)
 #include <stdlib.h>
 #include "libavutil/avstring.h"
 #include "libavutil/bswap.h"
+#include "libavutil/dict.h"
 #include "libavcodec/get_bits.h"
 #include "libavcodec/bytestream.h"
+#include "libavcodec/vorbis_parser.h"
 #include "avformat.h"
+#include "internal.h"
 #include "oggdec.h"
 #include "vorbiscomment.h"
 
@@ -43,7 +46,7 @@ static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
         if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
             return 0;
 
-        ff_new_chapter(as, cnum, (AVRational){1,1000},
+        avpriv_new_chapter(as, cnum, (AVRational){1,1000},
                        ms + 1000*(s + 60*(m + 60*h)),
                        AV_NOPTS_VALUE, NULL);
         av_free(val);
@@ -56,8 +59,8 @@ static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
         if (!chapter)
             return 0;
 
-        av_metadata_set2(&chapter->metadata, "title", val,
-                         AV_METADATA_DONT_STRDUP_VAL);
+        av_dict_set(&chapter->metadata, "title", val,
+                         AV_DICT_DONT_STRDUP_VAL);
     } else
         return 0;
 
@@ -66,7 +69,7 @@ static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
 }
 
 int
-ff_vorbis_comment(AVFormatContext * as, AVMetadata **m, const uint8_t *buf, int size)
+ff_vorbis_comment(AVFormatContext * as, AVDictionary **m, const uint8_t *buf, int size)
 {
     const uint8_t *p = buf;
     const uint8_t *end = buf + size;
@@ -126,9 +129,9 @@ ff_vorbis_comment(AVFormatContext * as, AVMetadata **m, const uint8_t *buf, int
             ct[vl] = 0;
 
             if (!ogm_chapter(as, tt, ct))
-                av_metadata_set2(m, tt, ct,
-                                   AV_METADATA_DONT_STRDUP_KEY |
-                                   AV_METADATA_DONT_STRDUP_VAL);
+                av_dict_set(m, tt, ct,
+                                   AV_DICT_DONT_STRDUP_KEY |
+                                   AV_DICT_DONT_STRDUP_VAL);
         }
     }
 
@@ -138,7 +141,7 @@ ff_vorbis_comment(AVFormatContext * as, AVMetadata **m, const uint8_t *buf, int
         av_log(as, AV_LOG_INFO,
                "truncated comment header, %i comments not found\n", n);
 
-    metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
+    ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
 
     return 0;
 }
@@ -160,6 +163,9 @@ ff_vorbis_comment(AVFormatContext * as, AVMetadata **m, const uint8_t *buf, int
 struct oggvorbis_private {
     unsigned int len[3];
     unsigned char *packet[3];
+    VorbisParseContext vp;
+    int64_t final_pts;
+    int final_duration;
 };
 
 
@@ -221,6 +227,7 @@ vorbis_header (AVFormatContext * s, int idx)
     if (os->buf[os->pstart] == 1) {
         const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
         unsigned blocksize, bs0, bs1;
+        int srate;
 
         if (os->psize != 30)
             return -1;
@@ -229,7 +236,7 @@ vorbis_header (AVFormatContext * s, int idx)
             return -1;
 
         st->codec->channels = bytestream_get_byte(&p);
-        st->codec->sample_rate = bytestream_get_le32(&p);
+        srate = bytestream_get_le32(&p);
         p += 4; // skip maximum bitrate
         st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
         p += 4; // skip minimum bitrate
@@ -249,21 +256,107 @@ vorbis_header (AVFormatContext * s, int idx)
         st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
         st->codec->codec_id = CODEC_ID_VORBIS;
 
-        st->time_base.num = 1;
-        st->time_base.den = st->codec->sample_rate;
+        if (srate > 0) {
+            st->codec->sample_rate = srate;
+            avpriv_set_pts_info(st, 64, 1, srate);
+        }
     } else if (os->buf[os->pstart] == 3) {
-        if (os->psize > 8)
-            ff_vorbis_comment (s, &st->metadata, os->buf + os->pstart + 7, os->psize - 8);
+        if (os->psize > 8 &&
+            ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7, os->psize - 8) >= 0) {
+            // drop all metadata we parsed and which is not required by libvorbis
+            unsigned new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
+            if (new_len >= 16 && new_len < os->psize) {
+                AV_WL32(priv->packet[1] + new_len - 5, 0);
+                priv->packet[1][new_len - 1] = 1;
+                priv->len[1] = new_len;
+            }
+        }
     } else {
+        int ret;
         st->codec->extradata_size =
             fixup_vorbis_headers(s, priv, &st->codec->extradata);
+        if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
+            av_freep(&st->codec->extradata);
+            st->codec->extradata_size = 0;
+            return ret;
+        }
     }
 
     return 1;
 }
 
+static int vorbis_packet(AVFormatContext *s, int idx)
+{
+    struct ogg *ogg = s->priv_data;
+    struct ogg_stream *os = ogg->streams + idx;
+    struct oggvorbis_private *priv = os->private;
+    int duration;
+
+    /* first packet handling
+       here we parse the duration of each packet in the first page and compare
+       the total duration to the page granule to find the encoder delay and
+       set the first timestamp */
+    if (!os->lastpts) {
+        int seg;
+        uint8_t *last_pkt = os->buf + os->pstart;
+        uint8_t *next_pkt = last_pkt;
+        int first_duration = 0;
+
+        avpriv_vorbis_parse_reset(&priv->vp);
+        duration = 0;
+        for (seg = 0; seg < os->nsegs; seg++) {
+            if (os->segments[seg] < 255) {
+                int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
+                if (d < 0) {
+                    duration = os->granule;
+                    break;
+                }
+                if (!duration)
+                    first_duration = d;
+                duration += d;
+                last_pkt = next_pkt + os->segments[seg];
+            }
+            next_pkt += os->segments[seg];
+        }
+        os->lastpts = os->lastdts   = os->granule - duration;
+        s->streams[idx]->start_time = os->lastpts + first_duration;
+        if (s->streams[idx]->duration)
+            s->streams[idx]->duration -= s->streams[idx]->start_time;
+        s->streams[idx]->cur_dts    = AV_NOPTS_VALUE;
+        priv->final_pts             = AV_NOPTS_VALUE;
+        avpriv_vorbis_parse_reset(&priv->vp);
+    }
+
+    /* parse packet duration */
+    if (os->psize > 0) {
+        duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
+        if (duration <= 0) {
+            os->pflags |= AV_PKT_FLAG_CORRUPT;
+            return 0;
+        }
+        os->pduration = duration;
+    }
+
+    /* final packet handling
+       here we save the pts of the first packet in the final page, sum up all
+       packet durations in the final page except for the last one, and compare
+       to the page granule to find the duration of the final packet */
+    if (os->flags & OGG_FLAG_EOS) {
+        if (os->lastpts != AV_NOPTS_VALUE) {
+            priv->final_pts = os->lastpts;
+            priv->final_duration = 0;
+        }
+        if (os->segp == os->nsegs)
+            os->pduration = os->granule - priv->final_pts - priv->final_duration;
+        priv->final_duration += os->pduration;
+    }
+
+    return 0;
+}
+
 const struct ogg_codec ff_vorbis_codec = {
     .magic = "\001vorbis",
     .magicsize = 7,
-    .header = vorbis_header
+    .header = vorbis_header,
+    .packet = vorbis_packet,
 };