mxfdemux: Take temporal reordering from the index table into account

This is needed to know the PTS, without that we only know the DTS and
using that also for the PTS is wrong unless we have an intra-only codec.

If we can't get the temporal reordering from the index table, don't set
any PTS for non-intra-only codecs and let decoders figure out something.

https://bugzilla.gnome.org/show_bug.cgi?id=784027

11 files changed, 154 insertions, 39 deletions

diff --git a/gst/mxf/mxfaes-bwf.c b/gst/mxf/mxfaes-bwf.c
index f92d68279..d128ab270 100644
--- a/gst/mxf/mxfaes-bwf.c
+++ b/gst/mxf/mxfaes-bwf.c
@@ -1233,6 +1233,7 @@ static const MXFUL mxf_sound_essence_compression_alaw =
 static GstCaps *
 mxf_bwf_create_caps (MXFMetadataTimelineTrack * track,
     MXFMetadataGenericSoundEssenceDescriptor * descriptor, GstTagList ** tags,
+    gboolean * intra_only,
     MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
 {
   GstCaps *ret = NULL;
@@ -1348,13 +1349,16 @@ mxf_bwf_create_caps (MXFMetadataTimelineTrack * track,
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_BITRATE,
         wa_descriptor->avg_bps * 8, NULL);
 
+  *intra_only = TRUE;
+
   return ret;
 }
 
 static GstCaps *
 mxf_aes3_create_caps (MXFMetadataTimelineTrack * track,
     MXFMetadataGenericSoundEssenceDescriptor * descriptor, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   GstCaps *ret = NULL;
   MXFMetadataWaveAudioEssenceDescriptor *wa_descriptor = NULL;
@@ -1404,13 +1408,15 @@ mxf_aes3_create_caps (MXFMetadataTimelineTrack * track,
   g_free (codec_name);
 
   *handler = mxf_aes3_handle_essence_element;
+  *intra_only = TRUE;
 
   return ret;
 }
 
 static GstCaps *
 mxf_aes_bwf_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataGenericSoundEssenceDescriptor *s = NULL;
   gboolean bwf = FALSE;
@@ -1454,9 +1460,11 @@ mxf_aes_bwf_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
     GST_ERROR ("No descriptor found for this track");
     return NULL;
   } else if (bwf) {
-    return mxf_bwf_create_caps (track, s, tags, handler, mapping_data);
+    return mxf_bwf_create_caps (track, s, tags, intra_only, handler,
+        mapping_data);
   } else {
-    return mxf_aes3_create_caps (track, s, tags, handler, mapping_data);
+    return mxf_aes3_create_caps (track, s, tags, intra_only, handler,
+        mapping_data);
   }
 
   return NULL;
diff --git a/gst/mxf/mxfalaw.c b/gst/mxf/mxfalaw.c
index 3ffce3ab3..d490033e9 100644
--- a/gst/mxf/mxfalaw.c
+++ b/gst/mxf/mxfalaw.c
@@ -119,7 +119,8 @@ mxf_alaw_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_alaw_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataGenericSoundEssenceDescriptor *s = NULL;
   guint i;
@@ -167,6 +168,8 @@ mxf_alaw_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
 
   }
 
+  *intra_only = TRUE;
+
   return caps;
 }
 
diff --git a/gst/mxf/mxfd10.c b/gst/mxf/mxfd10.c
index b3833c56e..21401cf52 100644
--- a/gst/mxf/mxfd10.c
+++ b/gst/mxf/mxfd10.c
@@ -174,7 +174,8 @@ mxf_d10_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_d10_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataGenericPictureEssenceDescriptor *p = NULL;
   MXFMetadataGenericSoundEssenceDescriptor *s = NULL;
@@ -247,6 +248,8 @@ mxf_d10_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
 
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "SMPTE D-10 Audio", NULL);
+
+    *intra_only = TRUE;
   } else if (p) {
     caps =
         gst_caps_new_simple ("video/mpeg", "systemstream", G_TYPE_BOOLEAN,
@@ -256,6 +259,9 @@ mxf_d10_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
     *handler = mxf_d10_picture_handle_essence_element;
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "SMPTE D-10 Video", NULL);
+
+    /* Does not allow temporal reordering */
+    *intra_only = TRUE;
   }
 
   return caps;
diff --git a/gst/mxf/mxfdemux.c b/gst/mxf/mxfdemux.c
index 677c0677b..9bfb6ec59 100644
--- a/gst/mxf/mxfdemux.c
+++ b/gst/mxf/mxfdemux.c
@@ -821,10 +821,11 @@ gst_mxf_demux_update_essence_tracks (GstMXFDemux * demux)
 
         caps = gst_caps_new_empty_simple (name);
         g_free (name);
+        etrack->intra_only = FALSE;
       } else {
         caps =
             etrack->handler->create_caps (track, &etrack->tags,
-            &etrack->handle_func, &etrack->mapping_data);
+            &etrack->intra_only, &etrack->handle_func, &etrack->mapping_data);
       }
 
       GST_DEBUG_OBJECT (demux, "Created caps %" GST_PTR_FORMAT, caps);
@@ -1622,6 +1623,8 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
   GstBuffer *outbuf = NULL;
   GstMXFDemuxEssenceTrack *etrack = NULL;
   gboolean keyframe = TRUE;
+  /* As in GstMXFDemuxIndex */
+  guint64 pts = G_MAXUINT64, dts = G_MAXUINT64;
 
   GST_DEBUG_OBJECT (demux,
       "Handling generic container essence element of size %" G_GSIZE_FORMAT
@@ -1680,7 +1683,8 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
         GstMXFDemuxIndex *idx =
             &g_array_index (etrack->offsets, GstMXFDemuxIndex, i);
 
-        if (idx->offset != 0 && idx->offset == demux->offset - demux->run_in) {
+        if (idx->initialized && idx->offset != 0
+            && idx->offset == demux->offset - demux->run_in) {
           etrack->position = i;
           break;
         }
@@ -1696,8 +1700,12 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
   if (etrack->offsets && etrack->offsets->len > etrack->position) {
     GstMXFDemuxIndex *index =
         &g_array_index (etrack->offsets, GstMXFDemuxIndex, etrack->position);
-    if (index->offset != 0)
+    if (index->initialized && index->offset != 0)
       keyframe = index->keyframe;
+    if (index->initialized && index->pts != G_MAXUINT64)
+      pts = index->pts;
+    if (index->initialized && index->dts != G_MAXUINT64)
+      dts = index->dts;
   }
 
   /* Create subbuffer to be able to change metadata */
@@ -1733,7 +1741,7 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
     keyframe = !GST_BUFFER_FLAG_IS_SET (outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
 
   /* Prefer keyframe information from index tables over everything else */
-  if (demux->index_tables && outbuf) {
+  if (demux->index_tables) {
     GList *l;
     GstMXFDemuxIndexTable *index_table = NULL;
 
@@ -1751,14 +1759,21 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
       GstMXFDemuxIndex *index =
           &g_array_index (index_table->offsets, GstMXFDemuxIndex,
           etrack->position);
-      if (index->offset != 0) {
+      if (index->initialized && index->offset != 0) {
         keyframe = index->keyframe;
 
-        if (keyframe)
-          GST_BUFFER_FLAG_UNSET (outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
-        else
-          GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
+        if (outbuf) {
+          if (keyframe)
+            GST_BUFFER_FLAG_UNSET (outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
+          else
+            GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
+        }
       }
+
+      if (index->initialized && index->pts != G_MAXUINT64)
+        pts = index->pts;
+      if (index->initialized && index->dts != G_MAXUINT64)
+        dts = index->dts;
     }
   }
 
@@ -1776,6 +1791,9 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
       GstMXFDemuxIndex index;
 
       index.offset = demux->offset - demux->run_in;
+      index.initialized = TRUE;
+      index.pts = pts;
+      index.dts = dts;
       index.keyframe = keyframe;
       if (etrack->offsets->len < etrack->position)
         g_array_set_size (etrack->offsets, etrack->position);
@@ -1826,7 +1844,15 @@ gst_mxf_demux_handle_generic_container_essence_element (GstMXFDemux * demux,
         gst_buffer_get_size (inbuf));
 
     GST_BUFFER_DTS (outbuf) = pad->position;
-    GST_BUFFER_PTS (outbuf) = pad->position;
+    if (pts != G_MAXUINT64)
+      GST_BUFFER_PTS (outbuf) = gst_util_uint64_scale (pts * GST_SECOND,
+          pad->current_essence_track->source_track->edit_rate.d,
+          pad->current_essence_track->source_track->edit_rate.n);
+    else if (etrack->intra_only)
+      GST_BUFFER_PTS (outbuf) = pad->position;
+    else
+      GST_BUFFER_PTS (outbuf) = GST_CLOCK_TIME_NONE;
+
     GST_BUFFER_DURATION (outbuf) =
         gst_util_uint64_scale (GST_SECOND,
         pad->current_essence_track->source_track->edit_rate.d,
@@ -3575,8 +3601,6 @@ collect_index_table_segments (GstMXFDemux * demux)
 
     for (i = 0; i < segment->n_index_entries && start + i < t->offsets->len;
         i++) {
-      GstMXFDemuxIndex *index =
-          &g_array_index (t->offsets, GstMXFDemuxIndex, start + i);
       guint64 offset = segment->index_entries[i].stream_offset;
       GList *m;
       GstMXFDemuxPartition *offset_partition = NULL, *next_partition = NULL;
@@ -3596,8 +3620,7 @@ collect_index_table_segments (GstMXFDemux * demux)
         next_partition = NULL;
       }
 
-      if (offset_partition && offset >= offset_partition->partition.body_offset
-          && (offset - offset_partition->partition.body_offset)) {
+      if (offset_partition && offset >= offset_partition->partition.body_offset) {
         offset =
             offset_partition->partition.this_partition +
             offset_partition->essence_container_offset + (offset -
@@ -3608,9 +3631,42 @@ collect_index_table_segments (GstMXFDemux * demux)
           GST_ERROR_OBJECT (demux,
               "Invalid index table segment going into next unrelated partition");
         } else {
+          GstMXFDemuxIndex *index;
+          gint8 temporal_offset = segment->index_entries[i].temporal_offset;
+          guint64 pts_i = G_MAXUINT64;
+
+          if (temporal_offset > 0 ||
+              (temporal_offset < 0 && start + i >= -(gint) temporal_offset)) {
+            pts_i = start + i + temporal_offset;
+
+            if (t->offsets->len < pts_i)
+              g_array_set_size (t->offsets, pts_i);
+
+            index = &g_array_index (t->offsets, GstMXFDemuxIndex, pts_i);
+            if (!index->initialized) {
+              index->initialized = TRUE;
+              index->offset = 0;
+              index->pts = G_MAXUINT64;
+              index->dts = G_MAXUINT64;
+              index->keyframe = FALSE;
+            }
+
+            index->pts = start + i;
+          }
+
+          index = &g_array_index (t->offsets, GstMXFDemuxIndex, start + i);
+          if (!index->initialized) {
+            index->initialized = TRUE;
+            index->offset = 0;
+            index->pts = G_MAXUINT64;
+            index->dts = G_MAXUINT64;
+            index->keyframe = FALSE;
+          }
+
           index->offset = offset;
           index->keyframe = ! !(segment->index_entries[i].flags & 0x80)
               || (segment->index_entries[i].key_frame_offset == 0);
+          index->dts = pts_i;
         }
       }
     }
diff --git a/gst/mxf/mxfdemux.h b/gst/mxf/mxfdemux.h
index b4d7242a8..48cf9e863 100644
--- a/gst/mxf/mxfdemux.h
+++ b/gst/mxf/mxfdemux.h
@@ -58,12 +58,6 @@ typedef struct
 
 typedef struct
 {
-  guint64 offset;
-  gboolean keyframe;
-} GstMXFDemuxIndex;
-
-typedef struct
-{
   guint32 body_sid;
   guint32 index_sid;
   guint32 track_number;
@@ -86,12 +80,30 @@ typedef struct
   GstTagList *tags;
 
   GstCaps *caps;
+  gboolean intra_only;
 } GstMXFDemuxEssenceTrack;
 
 typedef struct
 {
+  /* 0 if uninitialized */
+  guint64 offset;
+
+  /* PTS edit unit number or G_MAXUINT64 */
+  guint64 pts;
+
+  /* DTS edit unit number if we got here via PTS */
+  guint64 dts;
+
+  gboolean keyframe;
+  gboolean initialized;
+} GstMXFDemuxIndex;
+
+typedef struct
+{
   guint32 body_sid;
   guint32 index_sid;
+
+  /* offsets indexed by DTS */
   GArray *offsets;
 } GstMXFDemuxIndexTable;
 
diff --git a/gst/mxf/mxfdv-dif.c b/gst/mxf/mxfdv-dif.c
index 6f72c902a..dbfe3490a 100644
--- a/gst/mxf/mxfdv-dif.c
+++ b/gst/mxf/mxfdv-dif.c
@@ -140,7 +140,8 @@ mxf_dv_dif_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_dv_dif_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   GstCaps *caps = NULL;
   guint i;
@@ -181,6 +182,8 @@ mxf_dv_dif_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
 
   gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_CODEC, "DV-DIF", NULL);
 
+  *intra_only = TRUE;
+
   return caps;
 }
 
diff --git a/gst/mxf/mxfessence.h b/gst/mxf/mxfessence.h
index 97feab624..57a57a3b2 100644
--- a/gst/mxf/mxfessence.h
+++ b/gst/mxf/mxfessence.h
@@ -37,7 +37,7 @@ typedef GstFlowReturn (*MXFEssenceElementHandleFunc) (const MXFUL *key, GstBuffe
 typedef struct {
   gboolean (*handles_track) (const MXFMetadataTimelineTrack *track);
   MXFEssenceWrapping (*get_track_wrapping) (const MXFMetadataTimelineTrack *track);
-  GstCaps * (*create_caps) (MXFMetadataTimelineTrack *track, GstTagList **tags, MXFEssenceElementHandleFunc *handler, gpointer *mapping_data);
+  GstCaps * (*create_caps) (MXFMetadataTimelineTrack *track, GstTagList **tags, gboolean * intra_only, MXFEssenceElementHandleFunc *handler, gpointer *mapping_data);
 } MXFEssenceElementHandler;
 
 typedef GstFlowReturn (*MXFEssenceElementWriteFunc) (GstBuffer *buffer, gpointer mapping_data, GstAdapter *adapter, GstBuffer **outbuf, gboolean flush);
diff --git a/gst/mxf/mxfjpeg2000.c b/gst/mxf/mxfjpeg2000.c
index 35df1ba06..5d084f24f 100644
--- a/gst/mxf/mxfjpeg2000.c
+++ b/gst/mxf/mxfjpeg2000.c
@@ -126,7 +126,8 @@ mxf_jpeg2000_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_jpeg2000_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataFileDescriptor *f = NULL;
   MXFMetadataGenericPictureEssenceDescriptor *p = NULL;
@@ -234,6 +235,8 @@ mxf_jpeg2000_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
   gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
       "JPEG 2000", NULL);
 
+  *intra_only = TRUE;
+
   return caps;
 }
 
diff --git a/gst/mxf/mxfmpeg.c b/gst/mxf/mxfmpeg.c
index 30ba9cfc0..57dd75d07 100644
--- a/gst/mxf/mxfmpeg.c
+++ b/gst/mxf/mxfmpeg.c
@@ -627,8 +627,8 @@ static const MXFUL sound_essence_compression_aac = { {
 
 static GstCaps *
 mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data,
-    MXFMetadataGenericPictureEssenceDescriptor * p,
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data, MXFMetadataGenericPictureEssenceDescriptor * p,
     MXFMetadataGenericSoundEssenceDescriptor * s)
 {
   GstCaps *caps = NULL;
@@ -648,6 +648,7 @@ mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       codec_name = "MPEG-2 Video";
       t = MXF_MPEG_ESSENCE_TYPE_VIDEO_MPEG2;
       memcpy (mdata, &t, sizeof (MXFMPEGEssenceType));
+      *intra_only = FALSE;
     } else if (p->picture_essence_coding.u[0] != 0x06
         || p->picture_essence_coding.u[1] != 0x0e
         || p->picture_essence_coding.u[2] != 0x2b
@@ -669,12 +670,14 @@ mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       codec_name = "MPEG-2 Video";
       t = MXF_MPEG_ESSENCE_TYPE_VIDEO_MPEG2;
       memcpy (mdata, &t, sizeof (MXFMPEGEssenceType));
+      *intra_only = FALSE;
     } else if (p->picture_essence_coding.u[13] == 0x10) {
       caps = gst_caps_new_simple ("video/mpeg", "mpegversion", G_TYPE_INT, 1,
           "systemstream", G_TYPE_BOOLEAN, FALSE, NULL);
       codec_name = "MPEG-1 Video";
       t = MXF_MPEG_ESSENCE_TYPE_VIDEO_MPEG2;
       memcpy (mdata, &t, sizeof (MXFMPEGEssenceType));
+      *intra_only = TRUE;
     } else if (p->picture_essence_coding.u[13] == 0x20) {
       MXFLocalTag *local_tag =
           (((MXFMetadataBase *) p)->other_tags) ?
@@ -698,6 +701,7 @@ mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       codec_name = "MPEG-4 Video";
       t = MXF_MPEG_ESSENCE_TYPE_VIDEO_MPEG4;
       memcpy (mdata, &t, sizeof (MXFMPEGEssenceType));
+      *intra_only = FALSE;
     } else if ((p->picture_essence_coding.u[13] >> 4) == 0x03) {
       /* RP 2008 */
 
@@ -707,6 +711,7 @@ mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       codec_name = "h.264 Video";
       t = MXF_MPEG_ESSENCE_TYPE_VIDEO_AVC;
       memcpy (mdata, &t, sizeof (MXFMPEGEssenceType));
+      *intra_only = FALSE;
     } else {
       GST_ERROR ("Unsupported MPEG picture essence coding 0x%02x",
           p->picture_essence_coding.u[13]);
@@ -764,6 +769,7 @@ mxf_mpeg_es_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       mxf_metadata_generic_sound_essence_descriptor_set_caps (s, caps);
       *handler = mxf_mpeg_audio_handle_essence_element;
     }
+    *intra_only = TRUE;
   }
 
   if (caps) {
@@ -823,7 +829,8 @@ mxf_mpeg_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataFileDescriptor *f = NULL;
   MXFMetadataGenericPictureEssenceDescriptor *p = NULL;
@@ -866,7 +873,9 @@ mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
   if (f->essence_container.u[13] == 0x04) {
     GST_DEBUG ("Found MPEG ES stream");
 
-    caps = mxf_mpeg_es_create_caps (track, tags, handler, mapping_data, p, s);
+    caps =
+        mxf_mpeg_es_create_caps (track, tags, intra_only, handler, mapping_data,
+        p, s);
   } else if (f->essence_container.u[13] == 0x07) {
     GST_ERROR ("MPEG PES streams not supported yet");
     return NULL;
@@ -880,6 +889,7 @@ mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       *tags = gst_tag_list_new_empty ();
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "MPEG PS", NULL);
+    *intra_only = FALSE;
   } else if (f->essence_container.u[13] == 0x09) {
     GST_DEBUG ("Found MPEG TS stream");
     caps = gst_caps_new_empty_simple ("video/mpegts");
@@ -888,6 +898,7 @@ mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       *tags = gst_tag_list_new_empty ();
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "MPEG TS", NULL);
+    *intra_only = FALSE;
   } else if (f->essence_container.u[13] == 0x0f) {
     GST_DEBUG ("Found h264 NAL unit stream");
     /* RP 2008 */
@@ -899,6 +910,7 @@ mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       *tags = gst_tag_list_new_empty ();
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "h.264 Video", NULL);
+    *intra_only = FALSE;
   } else if (f->essence_container.u[13] == 0x10) {
     GST_DEBUG ("Found h264 byte-stream stream");
     /* RP 2008 */
@@ -910,6 +922,7 @@ mxf_mpeg_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
       *tags = gst_tag_list_new_empty ();
     gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
         "h.264 Video", NULL);
+    *intra_only = FALSE;
   }
 
   if (p && caps)
diff --git a/gst/mxf/mxfup.c b/gst/mxf/mxfup.c
index afb6d950f..d8b6664da 100644
--- a/gst/mxf/mxfup.c
+++ b/gst/mxf/mxfup.c
@@ -240,7 +240,8 @@ mxf_up_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 static GstCaps *
 mxf_up_rgba_create_caps (MXFMetadataTimelineTrack * track,
     MXFMetadataRGBAPictureEssenceDescriptor * d, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   GstCaps *caps = NULL;
   guint i;
@@ -280,6 +281,7 @@ mxf_up_rgba_create_caps (MXFMetadataTimelineTrack * track,
         ((MXFMetadataGenericPictureEssenceDescriptor *) d)->image_end_offset;
 
     *mapping_data = data;
+    *intra_only = TRUE;
   } else {
     GST_WARNING ("Unsupported pixel layout");
   }
@@ -290,7 +292,8 @@ mxf_up_rgba_create_caps (MXFMetadataTimelineTrack * track,
 static GstCaps *
 mxf_up_cdci_create_caps (MXFMetadataTimelineTrack * track,
     MXFMetadataCDCIPictureEssenceDescriptor * d, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   GstCaps *caps = NULL;
   guint i;
@@ -326,6 +329,7 @@ mxf_up_cdci_create_caps (MXFMetadataTimelineTrack * track,
         ((MXFMetadataGenericPictureEssenceDescriptor *) d)->image_end_offset;
 
     *mapping_data = data;
+    *intra_only = TRUE;
   } else {
     GST_WARNING ("Unsupported CDCI format");
   }
@@ -335,7 +339,8 @@ mxf_up_cdci_create_caps (MXFMetadataTimelineTrack * track,
 
 static GstCaps *
 mxf_up_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataGenericPictureEssenceDescriptor *p = NULL;
   MXFMetadataCDCIPictureEssenceDescriptor *c = NULL;
@@ -378,9 +383,13 @@ mxf_up_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
   *handler = mxf_up_handle_essence_element;
 
   if (r) {
-    caps = mxf_up_rgba_create_caps (track, r, tags, handler, mapping_data);
+    caps =
+        mxf_up_rgba_create_caps (track, r, tags, intra_only, handler,
+        mapping_data);
   } else if (c) {
-    caps = mxf_up_cdci_create_caps (track, c, tags, handler, mapping_data);
+    caps =
+        mxf_up_cdci_create_caps (track, c, tags, intra_only, handler,
+        mapping_data);
   } else {
     return NULL;
   }
diff --git a/gst/mxf/mxfvc3.c b/gst/mxf/mxfvc3.c
index 2f2e5b074..638cffa3a 100644
--- a/gst/mxf/mxfvc3.c
+++ b/gst/mxf/mxfvc3.c
@@ -138,7 +138,8 @@ mxf_vc3_get_track_wrapping (const MXFMetadataTimelineTrack * track)
 
 static GstCaps *
 mxf_vc3_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
-    MXFEssenceElementHandleFunc * handler, gpointer * mapping_data)
+    gboolean * intra_only, MXFEssenceElementHandleFunc * handler,
+    gpointer * mapping_data)
 {
   MXFMetadataFileDescriptor *f = NULL;
   MXFMetadataGenericPictureEssenceDescriptor *p = NULL;
@@ -186,6 +187,7 @@ mxf_vc3_create_caps (MXFMetadataTimelineTrack * track, GstTagList ** tags,
     *tags = gst_tag_list_new_empty ();
   gst_tag_list_add (*tags, GST_TAG_MERGE_APPEND, GST_TAG_VIDEO_CODEC,
       "VC-3 Video", NULL);
+  *intra_only = TRUE;
 
   return caps;
 }