[FFmpeg-devel] [PATCH 03/25] avfilter/af_headphone: Check for the existence of samples

[Andreas Rheinhardt]

Nicolas George:
> Andreas Rheinhardt (12020-09-08):
>> Not providing any samples makes no sense at all. And if no samples
>> were provided for one of the HRIR streams, one would either run into
>> an av_assert1 in ff_inlink_consume_samples() or into a segfault in
>> take_samples() in avfilter.c.
>>
>> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at gmail.com>
>> ---
>>  libavfilter/af_headphone.c | 8 +++++++-
>>  1 file changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/libavfilter/af_headphone.c b/libavfilter/af_headphone.c
>> index 1024ff57b1..f488e0e28d 100644
>> --- a/libavfilter/af_headphone.c
>> +++ b/libavfilter/af_headphone.c
>> @@ -631,8 +631,14 @@ static int activate(AVFilterContext *ctx)
>>              if ((ret = check_ir(ctx->inputs[i], i)) < 0)
>>                  return ret;
>>  
>> -                if (ff_outlink_get_status(ctx->inputs[i]) == AVERROR_EOF)
>> +            if (ff_outlink_get_status(ctx->inputs[i]) == AVERROR_EOF) {
>> +                if (!ff_inlink_queued_samples(ctx->inputs[i])) {
>> +                    av_log(ctx, AV_LOG_ERROR, "No samples provided for "
>> +                           "HRIR stream %d.\n", i - 1);
>> +                    return AVERROR_INVALIDDATA;
>> +                }
>>                      s->in[i].eof = 1;
>> +            }
> 
> No, this is bogus. A filter should not call ff_outlink_get_status() on
> its input, that is breaking the abstraction provided by the API. In
> fact, everything in `git grep 'ff_outlink_get_status.*input'` is a bug.
> 

How should one then check whether an input is finished? Check for
ff_inlink_check_available_samples(ctx->inputs[i],
ff_inlink_queued_samples(ctx->inputs[i]) + 1) being 1?

- Andreas