[FFmpeg-devel] [PATCH] avformat/mlpdec: fix time_base for packet timestamps

[Andreas Rheinhardt]

Paul B Mahol:
> On Sun, Sep 5, 2021 at 6:12 PM Andreas Rheinhardt <
> andreas.rheinhardt at outlook.com> wrote:
> 
>> Paul B Mahol:
>>> Signed-off-by: Paul B Mahol <onemda at gmail.com>
>>> ---
>>>  libavformat/mlpdec.c | 38 +++++++++++++++++++++++++++++++++++---
>>>  1 file changed, 35 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/libavformat/mlpdec.c b/libavformat/mlpdec.c
>>> index 8f0aabb510..f13d0fac8c 100644
>>> --- a/libavformat/mlpdec.c
>>> +++ b/libavformat/mlpdec.c
>>> @@ -22,8 +22,11 @@
>>>   */
>>>
>>>  #include "avformat.h"
>>> +#include "avio_internal.h"
>>> +#include "internal.h"
>>>  #include "rawdec.h"
>>>  #include "libavutil/intreadwrite.h"
>>> +#include "libavcodec/mlp_parse.h"
>>>
>>>  static int av_always_inline mlp_thd_probe(const AVProbeData *p,
>> uint32_t sync)
>>>  {
>>> @@ -50,6 +53,36 @@ static int av_always_inline mlp_thd_probe(const
>> AVProbeData *p, uint32_t sync)
>>>      return 0;
>>>  }
>>>
>>> +static int mlp_read_header(AVFormatContext *s)
>>> +{
>>> +    int ret = ff_raw_audio_read_header(s);
>>> +
>>> +    if (ret < 0)
>>> +        return ret;
>>> +
>>> +    ret = ffio_ensure_seekback(s->pb, 10);
>>> +    if (ret == 0) {
>>> +        int sample_rate, type;
>>> +
>>> +        avio_skip(s->pb, 7);
>>> +        type = avio_r8(s->pb);
>>> +
>>> +        switch (type) {
>>> +        case 0xba:
>>> +            sample_rate = mlp_samplerate(avio_r8(s->pb) >> 4);
>>> +            break;
>>> +        case 0xbb:
>>> +            sample_rate = mlp_samplerate((avio_rb16(s->pb) >> 4) & 0xF);
>>> +            break;
>>> +        }
>>> +
>>> +        avpriv_set_pts_info(s->streams[0], 64, 1, sample_rate);
>>> +        avio_seek(s->pb, -9 - (type == 0xbb), SEEK_CUR);
>>
>> This presumes that one of the two cases of the switch will be taken.
>> I think it would be easier if you just read the first 10 bytes into a
>> stack buffer and parsed from it instead. Then you can always seek back
>> by 10 bytes.
>>
>> (I always thought that thd can have very high samplerates (up to
>> 192kHz). Am I wrong?)
>>
> 
> Yes, but lowest is 44100 Hz.
> 
Seems like I completely misunderstood mlp_samplerate(). I forgot the
shifting.

- Andreas