[FFmpeg-devel] [PATCH] ffprobe: port to new decode API

wm4 nfxjfg at googlemail.com
Fri Mar 31 14:28:13 EEST 2017

On Fri, 31 Mar 2017 03:49:16 +0200
Michael Niedermayer <michael at niedermayer.cc> wrote:

> On Thu, Mar 30, 2017 at 05:05:36PM +0200, wm4 wrote:
> > Not sure if it behaves ideally in presence of decoding errors.
> > ---
> > The original version of this patch is actually 1 year old.
> > ---
> >  ffprobe.c | 42 +++++++++++++++++++++++++++++-------------
> >  1 file changed, 29 insertions(+), 13 deletions(-)  
> changes:
> http://samples.mplayerhq.hu/A-codecs/WMA/wav/FLCL_Ending_My.wav
> ./ffprobe -v 0 -show_frames -of compact=nk=1:p=0 ~/tickets/1905/FLCL_Ending_My.wav
> after the patch there are alot of "N/A" in the output where numbers
> where before

This is just what the API returns. It has nothing to do with this
patch. One packet results in multiple frames in this case, and with the
new decode API, the timestamp fields in subsequent frames have the
timestamp fields unset, while the old API duplicates the ones from the
first frame of a packet.

This in turn happens because the "emulation" code in utils.c resets the
timestamps on subsequent frames explicitly:

        avctx->internal->buffer_pkt->pts   = AV_NOPTS_VALUE;
        avctx->internal->buffer_pkt->dts   = AV_NOPTS_VALUE;

So this is very much by design, and I think the result with the new API
is better too. (Old API users probably had to explicitly handle this
case themselves, because nobody wants non-increasing timestamps.)

More information about the ffmpeg-devel mailing list