[FFmpeg-devel] [PATCH] mov: fix DTS calculation for samples with negative stts duration

[Andreas Cadhalpun]

On 28.05.2015 04:29, Michael Niedermayer wrote:
> On Thu, May 28, 2015 at 12:11:00AM +0200, Andreas Cadhalpun wrote:
>> A negative sample duration is invalid according to the spec, but there
>> are samples that use it for the DTS calculation, e.g.:
>> http://files.1f0.de/samples/mp4-negative-stts-problem.mp4
>>
>> These currently get out of A/V sync.
>>
>> Also change the logging type to AV_LOG_WARNING, because decoding the
>> sample can continue.
>>
>> Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun at googlemail.com>
>> ---
>>  libavformat/mov.c | 20 ++++++++++++++------
>>  1 file changed, 14 insertions(+), 6 deletions(-)
>>
>> diff --git a/libavformat/mov.c b/libavformat/mov.c
>> index 51cdd21..730f097 100644
>> --- a/libavformat/mov.c
>> +++ b/libavformat/mov.c
>> @@ -2234,12 +2234,6 @@ static int mov_read_stts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
>>          sample_count=avio_rb32(pb);
>>          sample_duration = avio_rb32(pb);
>>  
>> -        /* sample_duration < 0 is invalid based on the spec */
>> -        if (sample_duration < 0) {
>> -            av_log(c->fc, AV_LOG_ERROR, "Invalid SampleDelta %d in STTS, at %d st:%d\n",
>> -                   sample_duration, i, c->fc->nb_streams-1);
>> -            sample_duration = 1;
>> -        }
>>          if (sample_count < 0) {
>>              av_log(c->fc, AV_LOG_ERROR, "Invalid sample_count=%d\n", sample_count);
>>              return AVERROR_INVALIDDATA;
>> @@ -2439,6 +2433,7 @@ static void mov_build_index(MOVContext *mov, AVStream *st)
>>          unsigned int distance = 0;
>>          unsigned int rap_group_index = 0;
>>          unsigned int rap_group_sample = 0;
>> +        int64_t dts_correction = 0;
>>          int rap_group_present = sc->rap_group_count && sc->rap_group;
>>          int key_off = (sc->keyframe_count && sc->keyframes[0] > 0) || (sc->stps_count && sc->stps_data[0] > 0);
>>  
>> @@ -2522,6 +2517,19 @@ static void mov_build_index(MOVContext *mov, AVStream *st)
>>  
>>                  current_offset += sample_size;
>>                  stream_size += sample_size;
>> +                current_dts += dts_correction;
> 
> i think this could depending on the next stts value end up making
> DTS non monotone which seems wrong/unintended

Indeed. Fixed patch attached.

Best regards,
Andreas

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-mov-fix-DTS-calculation-for-samples-with-negative-st.patch
Type: text/x-diff
Size: 3506 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20150528/9089bda5/attachment.bin>