[FFmpeg-devel] [PATCH] Wrong duration in TS container (Ticket #1836)

[Michael Niedermayer]

Hi

On Sun, Nov 11, 2012 at 04:19:53PM +0900, Hee Suk Jung wrote:
> 
> From 25afab03b00a822421d07da2330483809f179506 Mon Sep 17 00:00:00 2001
> From: Heesuk Jung <heesuk.jung at lge.com>
> Date: Sat, 10 Nov 2012 23:02:21 -0800
> Subject: [PATCH] Wrong duration in TS container (Ticket #1836)
> 
> Sometimes TS container file has unreasonbly big PTS in last frame.
> In this case, we use previous last PTS value to decide stream dueation.
> 
> Fix 2nd promlematic file cas in Ticket #1836
> link : https://docs.google.com/open?id=0B6r7ZfWFIypCNFVuWHowMFBBN0E
> ---
>  libavformat/utils.c |   16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/libavformat/utils.c b/libavformat/utils.c
> index 0240b0c..6d953ed 100644
> --- a/libavformat/utils.c
> +++ b/libavformat/utils.c
> @@ -2157,6 +2157,7 @@ static void estimate_timings_from_bit_rate(AVFormatContext *ic)
> 
>  #define DURATION_MAX_READ_SIZE 250000LL
>  #define DURATION_MAX_RETRY 4
> +#define DURATION_MULTIPLE_LIMIT 8
> 
>  /* only usable for MPEG-PS streams */
>  static void estimate_timings_from_pts(AVFormatContext *ic, int64_t old_offset)
> @@ -2166,6 +2167,7 @@ static void estimate_timings_from_pts(AVFormatContext *ic, int64_t old_offset)
>      int read_size, i, ret;
>      int64_t end_time;
>      int64_t filesize, offset, duration;
> +    int64_t ref_duration, pre_duration;
>      int retry=0;
> 
>      /* flush packet queue */
> @@ -2186,6 +2188,7 @@ static void estimate_timings_from_pts(AVFormatContext *ic, int64_t old_offset)
>      /* XXX: may need to support wrapping */
>      filesize = ic->pb ? avio_size(ic->pb) : 0;
>      end_time = AV_NOPTS_VALUE;
> +    i = 0;
>      do{
>          offset = filesize - (DURATION_MAX_READ_SIZE<<retry);
>          if (offset < 0)
> @@ -2215,8 +2218,19 @@ static void estimate_timings_from_pts(AVFormatContext *ic, int64_t old_offset)
>                  if (duration < 0)
>                      duration += 1LL<<st->pts_wrap_bits;
>                  if (duration > 0) {
> -                    if (st->duration == AV_NOPTS_VALUE || st->duration < duration)
> +                    if (st->duration == AV_NOPTS_VALUE || st->duration < duration) {
> +                        pre_duration = st->duration;
>                          st->duration = duration;
> +                        i++;
> +                        if (i == 1) {
> +                            ref_duration = st->duration;
> +                        } else {
> +                            if (st->duration > ref_duration*DURATION_MULTIPLE_LIMIT) {
> +                                st->duration = pre_duration;

st->duration is in st->time_base units, 2 Streams can use different
time bases.
ref_duration*DURATION_MULTIPLE_LIMIT might theretically overflow,
better to divide or shift the other side of the comparission

setting ref_duration to 0 or AV_NOPTS_VALUE should avoid the need for
i


[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Breaking DRM is a little like attempting to break through a door even
though the window is wide open and the only thing in the house is a bunch
of things you dont want and which you would get tomorrow for free anyway
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121115/b265cb6b/attachment.asc>