mux_isom sometimes repeates ID3 emsg boxes in fragment files

[jadarve]

Thanks for reporting your issue. Please make sure these boxes are checked before submitting your issue - thank you!

• I looked for a similar issue and couldn't find any.
• I tried with the latest version of GPAC. Installers available at https://gpac.io/downloads/gpac-nightly-builds/
• I give enough information for contributors to reproduce my issue (meaningful title, GitHub labels, platform, and compiler, command-line ...).

Running GPAC for generating an HLS/DASH stream with CMAF from a live TS stream containing ID3 tags, sometimes duplicates tags in the CMAF segments. It's hard to diagnose, but I managed to find a test file that can be used to reproduce the problem. I will share it privately.

gpac/bin/gcc/gpac \
-i example.ts \
-o gpac_local/playlist.mpd:gpac:dual:gpac:buf=-150:ll_part_hb=-1:seg_sync=no:segdur=1.0:cdur=0.2:llhls=sf:hlsc=true:cmaf=cmf2:dmode=dynamic:pssh=mv:template=segment:keep_segs=true

With the sample file, segment 1270 has 2 ID3 tags with the same content.

Adding some log lines here https://github.com/gpac/gpac/blob/master/src/filters/mux_isom.c#L6425 I don't see the logic for deduplicating emsg boxes being activated. The moof box always arrives with zero emsg boxes here.

Related to: #2865

[jadarve]

Looking further into this, it looks as the problem of duplicated tags only appears when creating CMAF streams in fragmented mode. The original command:

gpac/bin/gcc/gpac \
-i example.ts \
-o gpac_local/playlist.mpd:gpac:dual:gpac:buf=-150:ll_part_hb=-1:seg_sync=no:segdur=1.0:cdur=0.2:llhls=sf:hlsc=true:cmaf=cmf2:dmode=dynamic:pssh=mv:template=segment:keep_segs=true

yields the following duplicated tag for one of the segments:

[styp] size=24
 - majorBrand: msdh
 - minorVersion: 0
 - compatibleBrand: msdh
 - compatibleBrand: msix
[moof] size=172
  [mfhd] size=16 version=0 flags=000000
   - sequenceNumber: 6346
  [traf] size=148
    [tfhd] size=24 version=0 flags=020022
     - trackID: 66
     - defaultBaseIsMoof: true
     - sampleDescriptionIndex: 1
     - defaultSampleFlags: 02000000 (isLeading=0 dependsOn=2 isDependedOn=0 hasRedundancy=0 padding=0 isNonSync=false degradationPriority=0)
    [tfdt] size=16 version=0 flags=000000
     - baseMediaDecodeTime: 438209280
    [trun] size=100 version=0 flags=000301
     - sampleCount: 10
[mdat] size=3464
[moof] size=172
  [mfhd] size=16 version=0 flags=000000
   - sequenceNumber: 6347
  [traf] size=148
    [tfhd] size=24 version=0 flags=020022
     - trackID: 66
     - defaultBaseIsMoof: true
     - sampleDescriptionIndex: 1
     - defaultSampleFlags: 02000000 (isLeading=0 dependsOn=2 isDependedOn=0 hasRedundancy=0 padding=0 isNonSync=false degradationPriority=0)
    [tfdt] size=16 version=0 flags=000000
     - baseMediaDecodeTime: 438228479
    [trun] size=100 version=0 flags=000301
     - sampleCount: 10
[mdat] size=3426
[emsg] size=355 version=1 flags=000000
 - timeScale: 90000
 - presentationTime: 438264456
 - eventDuration: 4294967295
 - id: 0
 - schedIdURI: https://aomedia.org/emsg/ID3
 - value: www.nielsen.com:id3:v1
 - messageData size=271: 49443304002000000205505249560000017b0000777777...
[moof] size=164
  [mfhd] size=16 version=0 flags=000000
   - sequenceNumber: 6348
  [traf] size=140
    [tfhd] size=24 version=0 flags=020022
     - trackID: 66
     - defaultBaseIsMoof: true
     - sampleDescriptionIndex: 1
     - defaultSampleFlags: 02000000 (isLeading=0 dependsOn=2 isDependedOn=0 hasRedundancy=0 padding=0 isNonSync=false degradationPriority=0)
    [tfdt] size=16 version=0 flags=000000
     - baseMediaDecodeTime: 438247679
    [trun] size=92 version=0 flags=000301
     - sampleCount: 9
[mdat] size=3285
[emsg] size=355 version=1 flags=000000
 - timeScale: 90000
 - presentationTime: 438264456
 - eventDuration: 4294967295
 - id: 0
 - schedIdURI: https://aomedia.org/emsg/ID3
 - value: www.nielsen.com:id3:v1
 - messageData size=271: 49443304002000000205505249560000017b0000777777..
[moof] size=164
  [mfhd] size=16 version=0 flags=000000
   - sequenceNumber: 6349
  [traf] size=140
    [tfhd] size=24 version=0 flags=020022
     - trackID: 66
     - defaultBaseIsMoof: true
     - sampleDescriptionIndex: 1
     - defaultSampleFlags: 02000000 (isLeading=0 dependsOn=2 isDependedOn=0 hasRedundancy=0 padding=0 isNonSync=false degradationPriority=0)
    [tfdt] size=16 version=0 flags=000000
     - baseMediaDecodeTime: 438264959
    [trun] size=92 version=0 flags=000301
     - sampleCount: 9
[mdat] size=3183
[moof] size=164
  [mfhd] size=16 version=0 flags=000000
   - sequenceNumber: 6350
  [traf] size=140
    [tfhd] size=24 version=0 flags=020022
     - trackID: 66
     - defaultBaseIsMoof: true
     - sampleDescriptionIndex: 1
     - defaultSampleFlags: 02000000 (isLeading=0 dependsOn=2 isDependedOn=0 hasRedundancy=0 padding=0 isNonSync=false degradationPriority=0)
    [tfdt] size=16 version=0 flags=000000
     - baseMediaDecodeTime: 438282239
    [trun] size=92 version=0 flags=000301
     - sampleCount: 9
[mdat] size=3208

where the duplicated tags are located in different fragments (moof boxes).

Running GPAC without fragment mode as:

gpac/bin/gcc/gpac \
-i example.ts \
-o gpac_local/playlist.mpd:gpac:dual:gpac:buf=-150:ll_part_hb=-1:seg_sync=no:segdur=1.0:llhls=sf:hlsc=true:cmaf=cmf2:dmode=dynamic:pssh=mv:template=segment:keep_segs=true

that is, without cdur=0.2, eliminates the duplicates. Unfortunately this is not an option for us as we want to use GPAC in low latency mode.

[rbouqueau]

Thank for reporting and documenting. I plan to have a look at id3 soon as we spotted some potential regression related to our recent commits as well.