[FFmpeg-devel] MPEG TS encoding problem on change from stereo to mono, pay for fix

[Andreas Unterhuber]

Works for me in the current configuration. Made a test with two files - every file with min 2 changes from mono to stereo. 

Am 22.11.2010 um 10:27 schrieb Stefano Sabatini:

> On date Monday 2010-11-22 00:12:36 +0100, Reimar D?ffinger encoded:
>> On Sun, Nov 21, 2010 at 11:11:30PM +0100, Stefano Sabatini wrote:
>>> On date Friday 2010-11-19 23:05:34 +0100, Reimar D?ffinger encoded:
>>>> On Fri, Nov 19, 2010 at 03:26:57PM +0100, Stefano Sabatini wrote:
>>>>> diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
>>>>> index bff9477..22e3615 100644
>>>>> --- a/libavcodec/avcodec.h
>>>>> +++ b/libavcodec/avcodec.h
>>>>> @@ -3076,6 +3076,14 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
>>>>>                                         int filter_length, int log2_phase_count,
>>>>>                                         int linear, double cutoff);
>>>>> 
>>>>> +int av_audio_resample_reinit(ReSampleContext **ctx,
>>>>> +                             int output_channels, int input_channels,
>>>>> +                             int output_rate, int input_rate,
>>>>> +                             enum AVSampleFormat sample_fmt_out,
>>>>> +                             enum AVSampleFormat sample_fmt_in,
>>>>> +                             int filter_length, int log2_phase_count,
>>>>> +                             int linear, double cutoff);
>>>> 
>>>> I'm not convinced this should be public API, since it only
>>>> checks for changes to the first arguments but not filter_length
>>>> etc. I'd consider it too confusing/hackish to make it public...
>>> 
>>> The problem is that those values cannot be checked in the
>>> ReSampleContext, as are defined in AVResampleContext, which is opaque
>>> in resample.c where the function is defined.
>>> 
>>> If we want to quickly fix the problem we could simply state in the
>>> docs that those values are supposed to be the same, or make
>>> AVResampleContext definition visible in resample.c, which doesn't look
>>> a good idea.
>> 
>> Or we could (for now) leave it completely up to the application to
>> keep track of which values they created the context with.
>> Would also be simpler in this case, since within FFmpeg we don't
>> have to check if any of the output stuff changed since that is
>> always the same.
> 
> Makes sense, patch updated.
> -- 
> FFmpeg = Fancy and Fabulous Mystic Puritan Exxagerate Game
> <0001-Fix-encoding-when-the-input-audio-format-rate-channe.patch>_______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel