[FFmpeg-devel] [PATCH 2/2] examples/decode_filter_video: Add loop for draining the filtergraph

[Anton Khirnov]

Quoting Tobias Rapp (2024-03-27 13:46:40)
> On 27/03/2024 13:06, Stefano Sabatini wrote:
> 
> > On date Wednesday 2024-03-27 11:51:31 +0100, Tobias Rapp wrote:
> >> Depending on the filters used the filtergraph can produce trailing data
> >> after feeding it the last input frame. Update the example to include the
> >> necessary loop for draining the filtergrap.
> >>
> >> Signed-off-by: Tobias Rapp <t.rapp at noa-archive.com>
> >> ---
> >>   doc/examples/decode_filter_video.c | 19 +++++++++++++++++++
> >>   1 file changed, 19 insertions(+)
> >>
> >> diff --git a/doc/examples/decode_filter_video.c b/doc/examples/decode_filter_video.c
> >> index 454c192..a57e6df 100644
> >> --- a/doc/examples/decode_filter_video.c
> >> +++ b/doc/examples/decode_filter_video.c
> >> @@ -276,6 +276,25 @@ int main(int argc, char **argv)
> >>           }
> >>           av_packet_unref(packet);
> >>       }
> >> +    if (ret == AVERROR_EOF) {
> >> +        /* signal EOF to the filtergraph */
> >> +        if (av_buffersrc_add_frame_flags(buffersrc_ctx, NULL, 0) < 0) {
> >> +            av_log(NULL, AV_LOG_ERROR, "Error while closing the filtergraph\n");
> >> +            goto end;
> >> +        }
> >> +
> >> +        /* pull remaining frames from the filtergraph */
> >> +        while (1) {
> >> +            ret = av_buffersink_get_frame(buffersink_ctx, filt_frame);
> >> +            if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
> >> +                break;
> > how are we supposed to handle the EAGAIN case? Shouldn't this be a
> > sleep and retry?
> 
> Good suggestion. I could add something like usleep(100) upon EAGAIN.

No, EAGAIN from a filter after EOF on all inputs is a bug.

-- 
Anton Khirnov