fix: proper handling of add-sidx for multi-fragment segment

Eyevinn · Apr 19, 2024 · e2dfbde · e2dfbde
1 parent b1b0dfb
commit e2dfbde
Show file tree

Hide file tree

Showing 3 changed files with 63 additions and 56 deletions.
diff --git a/examples/add-sidx/main.go b/examples/add-sidx/main.go
@@ -92,11 +92,6 @@ func run(in io.Reader, out io.Writer, nonZeroEPT, removeEncBoxes, segOnMoof bool
 		return err
 	}
 	fmt.Printf("found %d segments\n", len(mp4Root.Segments))
-	for i, seg := range mp4Root.Segments {
-		for j, frag := range seg.Fragments {
-			fmt.Printf("segment: %d, fragment: %d, sequenceNr: %d\n", i, j, frag.Moof.Mfhd.SequenceNumber)
-		}
-	}
 
 	if removeEncBoxes {
 		removeEncryptionBoxes(mp4Root)

diff --git a/examples/add-sidx/main_test.go b/examples/add-sidx/main_test.go
@@ -11,40 +11,45 @@ import (
 func TestAddSidx(t *testing.T) {
 	inPath := "testdata/clear_with_enc_boxes.mp4"
 	testCases := []struct {
-		desc         string
-		inPath       string
-		removeEnc    bool
-		segOnMoof    bool
-		wantedNrSegs uint32
-		wantedSize   uint32
+		desc           string
+		inPath         string
+		removeEnc      bool
+		segOnMoof      bool
+		wantedNrSegs   uint32
+		wantedSize     uint32
+		wantedFirstDur uint32
 	}{
 		{
-			desc:         "sidx, enc boxes, 1 segment",
-			inPath:       inPath,
-			removeEnc:    false,
-			segOnMoof:    false,
-			wantedNrSegs: 1,
+			desc:           "sidx, enc boxes, 1 segment",
+			inPath:         inPath,
+			removeEnc:      false,
+			segOnMoof:      false,
+			wantedNrSegs:   1,
+			wantedFirstDur: 2 * 144144,
 		},
 		{
-			desc:         "sidx, enc boxes, many segments",
-			inPath:       inPath,
-			removeEnc:    false,
-			segOnMoof:    true,
-			wantedNrSegs: 2,
+			desc:           "sidx, enc boxes, many segments",
+			inPath:         inPath,
+			removeEnc:      false,
+			segOnMoof:      true,
+			wantedNrSegs:   2,
+			wantedFirstDur: 144144,
 		},
 		{
-			desc:         "sidx, no enc boxes, many segments",
-			inPath:       inPath,
-			removeEnc:    true,
-			segOnMoof:    true,
-			wantedNrSegs: 2,
+			desc:           "sidx, no enc boxes, many segments",
+			inPath:         inPath,
+			removeEnc:      true,
+			segOnMoof:      true,
+			wantedNrSegs:   2,
+			wantedFirstDur: 144144,
 		},
 		{
-			desc:         "normal file with styp",
-			inPath:       "../resegmenter/testdata/testV300.mp4",
-			removeEnc:    false,
-			segOnMoof:    false,
-			wantedNrSegs: 4,
+			desc:           "normal file with styp",
+			inPath:         "../resegmenter/testdata/testV300.mp4",
+			removeEnc:      false,
+			segOnMoof:      false,
+			wantedNrSegs:   4,
+			wantedFirstDur: 180000,
 		},
 	}
 
@@ -66,10 +71,14 @@ func TestAddSidx(t *testing.T) {
 			if decOut.Sidx == nil {
 				t.Error("no sidx box")
 			}
-			gotNrEntries := len(decOut.Sidx.SidxRefs)
+			sidxEntries := decOut.Sidx.SidxRefs
+			gotNrEntries := len(sidxEntries)
 			if gotNrEntries != int(tc.wantedNrSegs) {
 				t.Errorf("got %d sidx entries instead of %d", gotNrEntries, tc.wantedNrSegs)
 			}
+			if sidxEntries[0].SubSegmentDuration != tc.wantedFirstDur {
+				t.Errorf("got first duration %d instead of %d", sidxEntries[0].SubSegmentDuration, tc.wantedFirstDur)
+			}
 			if tc.removeEnc {
 				for _, seg := range decOut.Segments {
 					for _, frag := range seg.Fragments {

diff --git a/mp4/file.go b/mp4/file.go
@@ -780,38 +780,41 @@ type segData struct {
 	size             uint32
 }
 
+// findSegmentData returns a slice of segment media data using a reference track.
 func findSegmentData(segs []*MediaSegment, refTrak *TrakBox, trex *TrexBox) ([]segData, error) {
 	segDatas := make([]segData, 0, len(segs))
 	for _, seg := range segs {
-		frag := seg.Fragments[0]
-		for _, traf := range frag.Moof.Trafs {
-			tfhd := traf.Tfhd
-			if tfhd.TrackID == refTrak.Tkhd.TrackID {
-				// Found the track that the sidx should be based on
-				baseTime := traf.Tfdt.BaseMediaDecodeTime()
-				dur := uint32(0)
-				var firstCompositionTimeOffest int64
-				for i, trun := range traf.Truns {
-					trun.AddSampleDefaultValues(tfhd, trex)
-					samples := trun.GetSamples()
-					for j, sample := range samples {
-						if i == 0 && j == 0 {
-							firstCompositionTimeOffest = int64(sample.CompositionTimeOffset)
+		var firstCompositionTimeOffest int64
+		dur := uint32(0)
+		var baseTime uint64
+		for fIdx, frag := range seg.Fragments {
+			for _, traf := range frag.Moof.Trafs {
+				tfhd := traf.Tfhd
+				if tfhd.TrackID == refTrak.Tkhd.TrackID { // Find track that gives sidx time values
+					if fIdx == 0 {
+						baseTime = traf.Tfdt.BaseMediaDecodeTime()
+					}
+					for i, trun := range traf.Truns {
+						trun.AddSampleDefaultValues(tfhd, trex)
+						samples := trun.GetSamples()
+						for j, sample := range samples {
+							if fIdx == 0 && i == 0 && j == 0 {
+								firstCompositionTimeOffest = int64(sample.CompositionTimeOffset)
+							}
+							dur += sample.Dur
 						}
-						dur += sample.Dur
 					}
 				}
-				sd := segData{
-					startPos:         seg.StartPos,
-					presentationTime: uint64(int64(baseTime) + firstCompositionTimeOffest),
-					baseDecodeTime:   baseTime,
-					dur:              dur,
-					size:             uint32(seg.Size()),
-				}
-				segDatas = append(segDatas, sd)
-				break
 			}
 		}
+		sd := segData{
+			startPos:         seg.StartPos,
+			presentationTime: uint64(int64(baseTime) + firstCompositionTimeOffest),
+			baseDecodeTime:   baseTime,
+			dur:              dur,
+			size:             uint32(seg.Size()),
+		}
+		segDatas = append(segDatas, sd)
 	}
 	return segDatas, nil
 }