On 7/30/2024 11:56 AM, jiangjie wrote:
---
  libavformat/mov.c | 7 +++----
  1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/libavformat/mov.c b/libavformat/mov.c
index 1052691936..38f015a766 100644
--- a/libavformat/mov.c
+++ b/libavformat/mov.c
@@ -5964,11 +5964,10 @@ static int mov_read_sidx(MOVContext *c, AVIOContext 
*pb, MOVAtom atom)
          int index;
          MOVFragmentStreamInfo * frag_stream_info;
          uint32_t size = avio_rb32(pb);
+        // ReferenceType = (size & (1 << 31)) ? 1 : 0
+        size = size & 0x7FFFFFFF;
+
          uint32_t duration = avio_rb32(pb);
-        if (size & 0x80000000) {
-            avpriv_request_sample(c->fc, "sidx reference_type 1");
-            return AVERROR_PATCHWELCOME;
-        }

Per the spec:

"Each entry in the SegmentIndexBox contains a reference type that indicates whether the reference points directly to the media bytes of a referenced leaf subsegment, or to a SegmentIndexBox that describes how the referenced subsegment is further subdivided; as a result, the segment may be indexed in a ‘hierarchical’ or ‘daisy-chain’ or other form by documenting time and byte offset information for other SegmentIndexBoxes applying to portions of the same (sub)segment"

And

"reference_type set to 1 indicates that the reference is to a SegmentIndexBox; otherwise the reference is to media content (e.g., in the case of files based on this document, to a MovieFragmentBox); if a separate index segment is used, then entries with reference type 1 are in the index segment, and entries with reference type 0 are in the media file".

All the code below this part assumes movie fragment, not another sidx box, so this patch is just ignoring the reference_type bit and thus is not correct.
Proper support for reference_type == 1 is needed.

          avio_rb32(pb); // sap_flags
          timestamp = av_rescale_q(pts, timescale, st->time_base);
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

Reply via email to