Quoting Vittorio Giovara (2015-11-08 20:19:05)
> On Sun, Nov 8, 2015 at 11:54 AM, Anton Khirnov <[email protected]> wrote:
> > All encoders set pts and dts properly now (and have been doing that for
> > a while), so there is no good reason to do any timestamp guessing in the
> > muxer.
> > ---
> >  libavformat/avformat.h |  8 ++++++
> >  libavformat/internal.h | 12 ++++++++
> >  libavformat/mux.c      | 77 
> > ++++++++++++++++++++++++++++++++++++++++++++++++--
> >  libavformat/utils.c    | 10 +++++++
> >  libavformat/version.h  |  3 ++
> >  5 files changed, 108 insertions(+), 2 deletions(-)
> >
> > diff --git a/libavformat/avformat.h b/libavformat/avformat.h
> > index 470bbc6..e441486 100644
> > --- a/libavformat/avformat.h
> > +++ b/libavformat/avformat.h
> > @@ -688,6 +688,8 @@ typedef struct AVIndexEntry {
> >   */
> >  #define AV_DISPOSITION_ATTACHED_PIC      0x0400
> >
> > +typedef struct AVStreamInternal AVStreamInternal;
> > +
> >  /**
> >   * Stream structure.
> >   * New fields can be added to the end with minor version bumps.
> > @@ -878,6 +880,12 @@ typedef struct AVStream {
> >                                      support seeking natively. */
> >      int nb_index_entries;
> >      unsigned int index_entries_allocated_size;
> > +
> > +    /**
> > +     * An opaque field for libavformat internal usage.
> > +     * Must not be accessed in any way by callers.
> > +     */
> > +    AVStreamInternal *internal;
> >  } AVStream;
> >
> >  #define AV_PROGRAM_RUNNING 1
> > diff --git a/libavformat/internal.h b/libavformat/internal.h
> > index a65a3b7..7bbf775 100644
> > --- a/libavformat/internal.h
> > +++ b/libavformat/internal.h
> > @@ -90,6 +90,18 @@ struct AVFormatInternal {
> >       * Timebase for the timestamp offset.
> >       */
> >      AVRational offset_timebase;
> > +
> > +#if FF_API_COMPUTE_PKT_FIELDS2
> > +    int missing_ts_warning;
> > +#endif
> > +};
> > +
> > +struct AVStreamInternal {
> > +    /**
> > +     * Set to 1 if the codec allows reordering, so pts can be different
> > +     * from dts.
> > +     */
> > +    int reorder;
> >  };
> 
> Is this new private field necessary? if reordering is the only
> usecase, couldn't we just check the AV_CODEC_PROP_REORDER flag from
> avctx->codec?

There is no avctx->codec available to the muxer.

This is a property of the codec id, so we either have to look up the
corresponding codec descripter at each call, or do what I did.

> 
> > -int av_write_frame(AVFormatContext *s, AVPacket *pkt)
> > +static int prepare_input_packet(AVFormatContext *s, AVPacket *pkt)
> >  {
> >      int ret;
> 
> would it be a problem to move the prepare_input_packet() factorization
> to a separate patch?

There is no factorization, I'm just renaming the check_packet()
function. Since it no longer just checks, but potentially also modifies
the packet, the old name would be misleading.

> 
> > @@ -389,16 +406,70 @@ int av_write_frame(AVFormatContext *s, AVPacket *pkt)
> >      if (ret < 0)
> >          return ret;
> >
> > +#if !FF_API_COMPUTE_PKT_FIELDS2
> > +    /* sanitize the timestamps */
> > +    if (!(s->oformat->flags & AVFMT_NOTIMESTAMPS)) {
> > +        AVStream *st = s->streams[pkt->stream_index];
> > +
> > +        /* when there is no reordering (so dts is equal to pts), but
> > +         * only one of them is set, set the other as well */
> > +        if (!st->internal->reorder) {
> > +            if (pkt->pts == AV_NOPTS_VALUE && pkt->dts != AV_NOPTS_VALUE)
> > +                pkt->pts = pkt->dts;
> > +            if (pkt->dts == AV_NOPTS_VALUE && pkt->pts != AV_NOPTS_VALUE)
> > +                pkt->dts = pkt->pts;
> > +        }
> > +
> > +        /* check that the timestamps are set */
> > +        if (pkt->pts == AV_NOPTS_VALUE || pkt->dts == AV_NOPTS_VALUE) {
> > +            av_log(s, AV_LOG_ERROR,
> > +                   "Timestamps are unset in a packet for stream %d\n", 
> > st->index);
> > +            return AVERROR(EINVAL);
> > +        }
> > +
> > +        /* check that the dts are increasing (or at least non-decreasing,
> > +         * if the format allows it */
> > +        if (st->cur_dts != AV_NOPTS_VALUE &&
> > +            ((!(s->oformat->flags & AVFMT_TS_NONSTRICT) && st->cur_dts >= 
> > pkt->dts) ||
> > +             st->cur_dts > pkt->dts)) {
> > +            av_log(s, AV_LOG_ERROR,
> > +                   "Application provided invalid, non monotonically 
> > increasing "
> > +                   "dts to muxer in stream %d: %" PRId64 " >= %" PRId64 
> > "\n",
> > +                   st->index, st->cur_dts, pkt->dts);
> > +            return AVERROR(EINVAL);
> > +        }
> > +
> > +        if (pkt->pts < pkt->dts) {
> > +            av_log(s, AV_LOG_ERROR, "pts %" PRId64 " < dts %" PRId64 " in 
> > stream %d\n",
> > +                   pkt->pts, pkt->dts, st->index);
> > +            return AVERROR(EINVAL);
> > +        }
> > +    }
> > +#endif
> > +
> > +    return 0;
> > +}
> > +
> > +int av_write_frame(AVFormatContext *s, AVPacket *pkt)
> > +{
> > +    int ret;
> > +
> > +    ret = prepare_input_packet(s, pkt);
> > +    if (ret < 0)
> > +        return ret;
> > +
> >      if (!pkt) {
> >          if (s->oformat->flags & AVFMT_ALLOW_FLUSH)
> >              return s->oformat->write_packet(s, pkt);
> >          return 1;
> >      }
> >
> > +#if FF_API_COMPUTE_PKT_FIELDS2
> >      ret = compute_pkt_fields2(s, s->streams[pkt->stream_index], pkt);
> >
> >      if (ret < 0 && !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
> >          return ret;
> > +#endif
> >
> >      ret = write_packet(s, pkt);
> >
> > @@ -552,7 +623,7 @@ int av_interleaved_write_frame(AVFormatContext *s, 
> > AVPacket *pkt)
> >  {
> >      int ret, flush = 0;
> >
> > -    ret = check_packet(s, pkt);
> > +    ret = prepare_input_packet(s, pkt);
> >      if (ret < 0)
> >          goto fail;
> >
> > @@ -561,8 +632,10 @@ int av_interleaved_write_frame(AVFormatContext *s, 
> > AVPacket *pkt)
> >
> >          av_log(s, AV_LOG_TRACE, "av_interleaved_write_frame size:%d dts:%" 
> > PRId64 " pts:%" PRId64 "\n",
> >                  pkt->size, pkt->dts, pkt->pts);
> > +#if FF_API_COMPUTE_PKT_FIELDS2
> >          if ((ret = compute_pkt_fields2(s, st, pkt)) < 0 && 
> > !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
> >              goto fail;
> > +#endif
> >
> >          if (pkt->dts == AV_NOPTS_VALUE && !(s->oformat->flags & 
> > AVFMT_NOTIMESTAMPS)) {
> >              ret = AVERROR(EINVAL);
> > diff --git a/libavformat/utils.c b/libavformat/utils.c
> > index 734f83f..bb17b4a 100644
> > --- a/libavformat/utils.c
> > +++ b/libavformat/utils.c
> > @@ -2460,6 +2460,8 @@ static void free_stream(AVStream **pst)
> >      if (st->attached_pic.data)
> >          av_packet_unref(&st->attached_pic);
> >
> > +    av_freep(&st->internal);
> > +
> >      av_dict_free(&st->metadata);
> >      av_freep(&st->probe_data.buf);
> >      av_free(st->index_entries);
> > @@ -2551,6 +2553,11 @@ AVStream *avformat_new_stream(AVFormatContext *s, 
> > const AVCodec *c)
> >          av_free(st);
> >          return NULL;
> >      }
> > +
> > +    st->internal = av_mallocz(sizeof(*st->internal));
> > +    if (!st->internal)
> > +        goto fail;
> > +
> >      if (s->iformat) {
> >          /* no default bitrate if decoding */
> >          st->codec->bit_rate = 0;
> > @@ -2583,6 +2590,9 @@ AVStream *avformat_new_stream(AVFormatContext *s, 
> > const AVCodec *c)
> >
> >      s->streams[s->nb_streams++] = st;
> >      return st;
> > +fail:
> > +    free_stream(&st);
> > +    return NULL;
> >  }
> >
> >  AVProgram *av_new_program(AVFormatContext *ac, int id)
> > diff --git a/libavformat/version.h b/libavformat/version.h
> > index d74968a..d004a55 100644
> > --- a/libavformat/version.h
> > +++ b/libavformat/version.h
> > @@ -60,5 +60,8 @@
> >  #ifndef FF_API_LAVF_FMT_RAWPICTURE
> >  #define FF_API_LAVF_FMT_RAWPICTURE      (LIBAVFORMAT_VERSION_MAJOR < 58)
> >  #endif
> > +#ifndef FF_API_COMPUTE_PKT_FIELDS2
> > +#define FF_API_COMPUTE_PKT_FIELDS2      (LIBAVFORMAT_VERSION_MAJOR < 58)
> > +#endif
> 
> I might be missing something, but aren't FF_API defines only for
> public fields? This is mostly private so you could change the code
> right away couldn't you?

I could, but callers that do not set the timestamps properly would
break. Since we supported that until now, it makes more sense to
deprecate it first.

-- 
Anton Khirnov
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to