oggdemux: Fix vorbis parsing

Add a granule to granulepos conversion function. Fix the duration function for vorbis. Handle timestamps on header packets differently and be more careful about calculating OFFSET and OFFSET_END. After this change, timestamps for vorbis don't exactly match up with the timestamps that vorbisparse outputs, but it's unclear if vorbisparse is actually correct and it would add a lot more code to make oggdemux match vorbisparse. Fixes #602790.
2024-07-20 05:18:20 +00:00 · 2009-11-24 21:22:03 -08:00 · 2009-11-24 21:22:03 -08:00 · 78aad52cbf
parent 1273909419
commit 78aad52cbf
4 changed files with 114 additions and 31 deletions
--- a/ext/ogg/gstoggdemux.c
+++ b/ext/ogg/gstoggdemux.c
@ -479,14 +479,34 @@ gst_ogg_demux_chain_peer (GstOggPad * pad, ogg_packet * packet)

  duration = gst_ogg_stream_get_packet_duration (&pad->map, packet);

-  GST_BUFFER_TIMESTAMP (buf) = gst_ogg_stream_granule_to_time (&pad->map,
-      pad->current_granule);
-  pad->current_granule += duration;
-  GST_BUFFER_DURATION (buf) = gst_ogg_stream_granule_to_time (&pad->map,
-      pad->current_granule) - GST_BUFFER_TIMESTAMP (buf);
+  if (packet->b_o_s) {
+    GST_BUFFER_TIMESTAMP (buf) = GST_CLOCK_TIME_NONE;
+    GST_BUFFER_DURATION (buf) = GST_CLOCK_TIME_NONE;
+    GST_BUFFER_OFFSET (buf) = 0;
+    GST_BUFFER_OFFSET_END (buf) = -1;
+  } else {
+    GST_BUFFER_TIMESTAMP (buf) = gst_ogg_stream_granule_to_time (&pad->map,
+        pad->current_granule);
+    GST_BUFFER_DURATION (buf) = gst_ogg_stream_granule_to_time (&pad->map,
+        pad->current_granule + duration) - GST_BUFFER_TIMESTAMP (buf);

-  GST_BUFFER_OFFSET (buf) = -1;
-  GST_BUFFER_OFFSET_END (buf) = packet->granulepos;
+    pad->current_granule += duration;
+    if (packet->granulepos != -1) {
+      gint64 granule;
+      granule = gst_ogg_stream_granulepos_to_granule (&pad->map,
+          packet->granulepos);
+      if (granule != pad->current_granule) {
+        GST_WARNING ("calculated granule didn't match actual (%lld != %lld)",
+            pad->current_granule, granule);
+      }
+      pad->current_granule = granule;
+    }
+    GST_BUFFER_OFFSET_END (buf) =
+        gst_ogg_stream_granule_to_granulepos (&pad->map, pad->current_granule,
+        pad->current_granule);
+    GST_BUFFER_OFFSET (buf) =
+        gst_ogg_stream_granule_to_time (&pad->map, pad->current_granule);
+  }

  /* Mark discont on the buffer */
  if (pad->discont) {
@ -586,7 +606,11 @@ gst_ogg_pad_submit_packet (GstOggPad * pad, ogg_packet * packet)
    pad->current_granule = granule;
  }

-  if (!gst_ogg_stream_packet_is_header (&pad->map, packet)) {
+  /* Overload the value of b_o_s in ogg_packet with a flag whether or
+   * not this is a header packet.  Maybe some day this could be cleaned
+   * up.  */
+  packet->b_o_s = gst_ogg_stream_packet_is_header (&pad->map, packet);
+  if (!packet->b_o_s) {
    if (pad->start_time == GST_CLOCK_TIME_NONE) {
      gint64 duration = gst_ogg_stream_get_packet_duration (&pad->map, packet);
      if (duration != -1) {
@ -602,6 +626,10 @@ gst_ogg_pad_submit_packet (GstOggPad * pad, ogg_packet * packet)

        pad->start_time = gst_ogg_stream_granule_to_time (&pad->map,
            start_granule);
+        GST_DEBUG ("start time %" G_GINT64_FORMAT, pad->start_time);
+      } else {
+        packet->granulepos = gst_ogg_stream_granule_to_granulepos (&pad->map,
+            pad->map.accumulated_granule, pad->map.accumulated_granule);
      }
    }
  } else {
@ -848,6 +876,7 @@ gst_ogg_chain_mark_discont (GstOggChain * chain)
    GstOggPad *pad = g_array_index (chain->streams, GstOggPad *, i);

    pad->discont = TRUE;
+    pad->map.last_size = 0;
  }
 }

@ -880,6 +909,7 @@ gst_ogg_chain_new_stream (GstOggChain * chain, glong serialno)

  GST_PAD_DIRECTION (ret) = GST_PAD_SRC;
  ret->discont = TRUE;
+  ret->map.last_size = 0;

  ret->chain = chain;
  ret->ogg = chain->ogg;
@ -1420,6 +1450,7 @@ gst_ogg_demux_activate_chain (GstOggDemux * ogg, GstOggChain * chain,

      /* mark discont */
      pad->discont = TRUE;
+      pad->map.last_size = 0;
      pad->last_ret = GST_FLOW_OK;

      pad->is_sparse =
--- a/ext/ogg/gstoggstream.c
+++ b/ext/ogg/gstoggstream.c
@ -39,6 +39,8 @@ typedef GstClockTime (*GstOggMapToTimeFunc) (GstOggStream * pad,
    gint64 granulepos);
 typedef gint64 (*GstOggMapToGranuleFunc) (GstOggStream * pad,
    gint64 granulepos);
+typedef gint64 (*GstOggMapToGranuleposFunc) (GstOggStream * pad,
+    gint64 granule, gint64 keyframe_granule);

 /* returns TRUE if the granulepos denotes a key frame */
 typedef gboolean (*GstOggMapIsKeyFrameFunc) (GstOggStream * pad,
@ -63,6 +65,7 @@ struct _GstOggMap
  const gchar *media_type;
  GstOggMapSetupFunc setup_func;
  GstOggMapToGranuleFunc granulepos_to_granule_func;
+  GstOggMapToGranuleposFunc granule_to_granulepos_func;
  GstOggMapIsKeyFrameFunc is_key_frame_func;
  GstOggMapIsHeaderPacketFunc is_header_func;
  GstOggMapPacketDurationFunc packet_duration_func;
@ -126,13 +129,30 @@ gst_ogg_stream_granulepos_to_granule (GstOggStream * pad, gint64 granulepos)
  }

  if (mappers[pad->map].granulepos_to_granule_func == NULL) {
-    GST_WARNING ("Failed to convert granulepos to time");
-    return GST_CLOCK_TIME_NONE;
+    GST_WARNING ("Failed to convert granulepos to granule");
+    return -1;
  }

  return mappers[pad->map].granulepos_to_granule_func (pad, granulepos);
 }

+gint64
+gst_ogg_stream_granule_to_granulepos (GstOggStream * pad, gint64 granule,
+    gint64 keyframe_granule)
+{
+  if (granule == -1 || granule == 0) {
+    return granule;
+  }
+
+  if (mappers[pad->map].granule_to_granulepos_func == NULL) {
+    GST_WARNING ("Failed to convert granule to granulepos");
+    return -1;
+  }
+
+  return mappers[pad->map].granule_to_granulepos_func (pad, granule,
+      keyframe_granule);
+}
+
 gboolean
 gst_ogg_stream_packet_granulepos_is_key_frame (GstOggStream * pad,
    gint64 granulepos)
@ -196,6 +216,20 @@ granulepos_to_granule_default (GstOggStream * pad, gint64 granulepos)
  }
 }

+static gint64
+granule_to_granulepos_default (GstOggStream * pad, gint64 granule,
+    gint64 keyframe_granule)
+{
+  gint64 keyoffset;
+
+  if (pad->granuleshift != 0) {
+    keyoffset = granule - keyframe_granule;
+    return (keyframe_granule << pad->granuleshift) | keyoffset;
+  } else {
+    return granule;
+  }
+}
+
 static gboolean
 is_header_unknown (GstOggStream * pad, ogg_packet * packet)
 {
@ -357,6 +391,15 @@ granulepos_to_granule_dirac (GstOggStream * pad, gint64 gp)
  return dt + 4;
 }

+static gint64
+granule_to_granulepos_dirac (GstOggStream * pad, gint64 granule,
+    gint64 keyframe_granule)
+{
+  /* This conversion requires knowing more details about the Dirac
+   * stream. */
+  return -1;
+}
+

 /* vorbis */

@ -373,6 +416,7 @@ setup_vorbis_mapper (GstOggStream * pad, ogg_packet * packet)
  pad->granulerate_n = GST_READ_UINT32_LE (data);
  pad->granulerate_d = 1;
  pad->granuleshift = 0;
+  pad->last_size = 0;
  GST_LOG ("sample rate: %" G_GUINT64_FORMAT, pad->granulerate_n);

  pad->n_header_packets = 3;
@ -408,28 +452,20 @@ packet_duration_vorbis (GstOggStream * pad, ogg_packet * packet)
  int size;
  int duration;

-  mode = (packet->packet[0] >> 1) & ((1 << pad->vorbis_log2_num_modes) - 1);
-  size = pad->vorbis_mode_sizes[mode];
+  if (packet->packet[0] & 1)
+    return 0;

-  if (size) {
-    switch ((packet->packet[0] >> (1 + pad->vorbis_log2_num_modes)) & 3) {
-      case 0:
-      case 3:
-        duration = pad->long_size / 2;
-        break;
-      case 1:
-        duration = (pad->long_size + pad->short_size) / 4;
-        break;
-      case 2:
-        duration = (3 * pad->long_size - pad->short_size) / 4;
-        break;
-      default:
-        duration = -1;
-        break;
-    }
+  mode = (packet->packet[0] >> 1) & ((1 << pad->vorbis_log2_num_modes) - 1);
+  size = pad->vorbis_mode_sizes[mode] ? pad->long_size : pad->short_size;
+
+  if (pad->last_size == 0) {
+    duration = 0;
  } else {
-    duration = pad->short_size / 2;
+    duration = pad->last_size / 4 + size / 4;
  }
+  pad->last_size = size;
+
+  GST_DEBUG ("duration %d", (int) duration);

  return duration;
 }
@ -898,6 +934,7 @@ static const GstOggMap mappers[] = {
    "video/x-theora",
    setup_theora_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    is_keyframe_theora,
    is_header_theora,
    packet_duration_constant
@ -907,6 +944,7 @@ static const GstOggMap mappers[] = {
    "audio/x-vorbis",
    setup_vorbis_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    is_keyframe_true,
    is_header_vorbis,
    packet_duration_vorbis
@ -916,6 +954,7 @@ static const GstOggMap mappers[] = {
    "audio/x-speex",
    setup_speex_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    is_keyframe_true,
    is_header_count,
    packet_duration_constant
@ -926,6 +965,7 @@ static const GstOggMap mappers[] = {
    setup_pcm_mapper,
    NULL,
    NULL,
+    NULL,
    is_header_count,
    NULL
  },
@ -935,6 +975,7 @@ static const GstOggMap mappers[] = {
    setup_cmml_mapper,
    NULL,
    NULL,
+    NULL,
    is_header_count,
    NULL
  },
@ -943,6 +984,7 @@ static const GstOggMap mappers[] = {
    "application/x-annodex",
    setup_fishead_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    NULL,
    is_header_count,
    NULL
@ -953,6 +995,7 @@ static const GstOggMap mappers[] = {
    setup_fishead_mapper,
    NULL,
    NULL,
+    NULL,
    is_header_true,
    NULL
  },
@ -961,6 +1004,7 @@ static const GstOggMap mappers[] = {
    "audio/x-flac",
    setup_fLaC_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    NULL,
    is_header_count,
    NULL
@ -970,6 +1014,7 @@ static const GstOggMap mappers[] = {
    "audio/x-flac",
    setup_flac_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    NULL,
    is_header_count,
    packet_duration_flac
@ -981,12 +1026,14 @@ static const GstOggMap mappers[] = {
    NULL,
    NULL,
    NULL,
+    NULL,
  },
  {
    "CELT    ", 8, 0,
    "audio/x-celt",
    setup_celt_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    NULL,
    is_header_count,
    packet_duration_constant
@ -997,6 +1044,7 @@ static const GstOggMap mappers[] = {
    setup_kate_mapper,
    NULL,
    NULL,
+    NULL,
    is_header_count,
    NULL
  },
@ -1005,6 +1053,7 @@ static const GstOggMap mappers[] = {
    "video/x-dirac",
    setup_dirac_mapper,
    granulepos_to_granule_dirac,
+    granule_to_granulepos_dirac,
    is_keyframe_dirac,
    is_header_count,
    packet_duration_constant
@ -1014,6 +1063,7 @@ static const GstOggMap mappers[] = {
    "application/x-ogm-audio",
    setup_ogmaudio_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    is_keyframe_true,
    is_header_unknown,
    NULL
@ -1023,6 +1073,7 @@ static const GstOggMap mappers[] = {
    "application/x-ogm-video",
    setup_ogmvideo_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    NULL,
    is_header_unknown,
    NULL
@ -1032,6 +1083,7 @@ static const GstOggMap mappers[] = {
    "application/x-ogm-text",
    setup_ogmtext_mapper,
    granulepos_to_granule_default,
+    granule_to_granulepos_default,
    is_keyframe_true,
    is_header_unknown,
    NULL
--- a/ext/ogg/gstoggstream.h
+++ b/ext/ogg/gstoggstream.h
@ -62,6 +62,7 @@ struct _GstOggStream
  int long_size;
  int vorbis_log2_num_modes;
  int vorbis_mode_sizes[256];
+  int last_size;
 };


@ -72,6 +73,7 @@ GstClockTime gst_ogg_stream_get_start_time_for_granulepos (GstOggStream *pad,
    gint64 granulepos);
 GstClockTime gst_ogg_stream_granule_to_time (GstOggStream *pad, gint64 granule);
 gint64 gst_ogg_stream_granulepos_to_granule (GstOggStream * pad, gint64 granulepos);
+gint64 gst_ogg_stream_granule_to_granulepos (GstOggStream * pad, gint64 granule, gint64 keyframe_granule);
 GstClockTime gst_ogg_stream_get_packet_start_time (GstOggStream *pad,
    ogg_packet *packet);
 gboolean gst_ogg_stream_granulepos_is_key_frame (GstOggStream *pad,
--- a/ext/ogg/vorbis_parse.c
+++ b/ext/ogg/vorbis_parse.c
@ -94,8 +94,6 @@ parse_vorbis_header_packet (GstOggStream * pad, ogg_packet * packet)
  pad->short_size = short_size;
  pad->long_size = long_size;
  pad->nsn_increment = short_size >> 1;
-
-  pad->accumulated_granule = -long_size / 2;
 }

 void