[FFmpeg-devel] [PATCH] ffmpeg: always init output stream before reaping filters

[Marton Balint]

On Sat, 7 Oct 2017, Hendrik Leppkes wrote:

> On Wed, Oct 4, 2017 at 12:48 AM, Marton Balint <cus at passwd.hu> wrote:
>> Otherwise the frame size of the codec is not set in the buffersink.
>>
>> Fixes ticket #6603 and the following simpler case:
>>
>> ffmpeg -c aac -filter_complex "sine=d=0.1,asetnsamples=1025" out.aac
>>
>> Signed-off-by: Marton Balint <cus at passwd.hu>
>> ---
>>  fftools/ffmpeg.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>>
>> diff --git a/fftools/ffmpeg.c b/fftools/ffmpeg.c
>> index 1d248bc269..5be8788ea8 100644
>> --- a/fftools/ffmpeg.c
>> +++ b/fftools/ffmpeg.c
>> @@ -4528,6 +4528,15 @@ static int transcode_step(void)
>>      }
>>
>>      if (ost->filter && ost->filter->graph->graph) {
>> +        if (!ost->initialized) {
>> +            char error[1024] = {0};
>> +            ret = init_output_stream(ost, error, sizeof(error));
>> +            if (ret < 0) {
>> +                av_log(NULL, AV_LOG_ERROR, "Error initializing output stream %d:%d -- %s\n",
>> +                       ost->file_index, ost->index, error);
>> +                exit_program(1);
>> +            }
>> +        }
>>          if ((ret = transcode_from_filter(ost->filter->graph, &ist)) < 0)
>>              return ret;
>>          if (!ist)
>
> Doesn't this basically void all the work that was done to get proper
> output frame information before initializing the output, so its
> created with the proper info at all times?

To be frank, I am not entirely sure. Fate passes, so this can't be that 
bad. The original patch that caused the regression is 
af1761f7b5b1b72197dc40934953b775c2d951cc which in the commit message only 
mentions the delayed initialization of filters, this patch keeps that.

Regards,
Marton