On Fri, 30 Sep 2016, Nicolas George wrote:
Le nonidi 9 vendémiaire, an CCXXV, Marton Balint a écrit :Signed-off-by: Marton Balint <c...@passwd.hu> --- libavdevice/openal-dec.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/libavdevice/openal-dec.c b/libavdevice/openal-dec.c index 0647952..8773944 100644 --- a/libavdevice/openal-dec.c +++ b/libavdevice/openal-dec.c @@ -191,6 +191,9 @@ static int read_packet(AVFormatContext* ctx, AVPacket *pkt) alcGetIntegerv(ad->device, ALC_CAPTURE_SAMPLES, (ALCsizei) sizeof(ALCint), &nb_samples); if (error = al_get_error(ad->device, &error_msg)) goto fail; + if (nb_samples == 0) + return AVERROR(EAGAIN); +EAGAIN is only acceptable in non-blocking mode. You may be able to use REDO (I do not remember the exact name), but beware you are not creating a busy-wait loop.
Alsa dev and v4l2 seem to work this way regardless if non blocking mode is set or not, and ffmpeg.c also waits if it encounters AVERROR(EAGAIN) in av_read_frame. So AVERROR(EAGAIN) return value seems to be consistent with existing code, which assumes that if EAGAIN is returned from av_read_frame, wait a bit.
If what you wrote is the way to move forward, then at least send a documentation patch, to establish this av_read_frame return value behaviour.
Regards, Marton _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
