[vlc-devel] [PATCH] AVFormat Demux : Reset PCR for streams initialized later on.

[Jai Menon]

On Sun, Jul 25, 2010 at 1:49 PM, Jai Menon <jmenon86 at gmail.com> wrote:
> On Sun, Jul 25, 2010 at 1:59 AM, Laurent Aimar <fenrir at elivagar.org> wrote:
>> Hi,
>>
>> On Sun, Jul 25, 2010 at 01:28:19AM +0530, Jai Menon wrote:
>>> On Sat, Jul 24, 2010 at 10:36 AM, Jai Menon <jmenon86 at gmail.com> wrote:
>>> > On Sat, Jul 24, 2010 at 9:50 AM, Jai Menon <jmenon86 at gmail.com> wrote:
>>> >> A situation where this is desirable is when the video stream is
>>> >> initialized after the audio stream. In this case the PCR must be
>>> >> reset based on the video stream packet timestamp to avoid dropping
>>> >> frames.
>>> >> ---
>>> >>  modules/demux/avformat/demux.c |    3 ++-
>>> >>  1 files changed, 2 insertions(+), 1 deletions(-)
>>> >>
>>> >> diff --git a/modules/demux/avformat/demux.c b/modules/demux/avformat/demux.c
>>> >> index a8042b8..86aeb5d 100644
>>> >> --- a/modules/demux/avformat/demux.c
>>> >> +++ b/modules/demux/avformat/demux.c
>>> >> @@ -588,7 +588,8 @@ static int Demux( demux_t *p_demux )
>>> >>  #endif
>>> >>
>>> >>     if( p_frame->i_dts > VLC_TS_INVALID  &&
>>> >> -        ( pkt.stream_index == p_sys->i_pcr_tk || p_sys->i_pcr_tk < 0 ) )
>>> >> +        ( pkt.stream_index == p_sys->i_pcr_tk || p_sys->i_pcr_tk < 0 ||
>>> >> +          p_frame->i_dts < p_sys->i_pcr ) )
>>> >>     {
>>> >>         p_sys->i_pcr_tk = pkt.stream_index;
>>> >>         p_sys->i_pcr = p_frame->i_dts;
>>> >> --
>>> >> 1.7.1.1
>>> >
>>> > I'm still considering if there's a better solution to this, so please
>>> > disregard for now.
>>>
>>> After thinking some more, I think maybe something like attached would
>>> be worth considering. Essentially, the reference is set after we get
>>> atleast a few video and audio packets.
>>  I may have misunderstood  your patch, but you must call ES_OUT_SET_PCR
>> before sending the first packets. Otherwise they will be rejected.
>
> would you be okay with updating the pcr after receiving a few packets
> instead? this way nothing is lost.

Here's what I have right now in my tree. If there aren't any obvious
issues I'm missing, maybe this could go in the 1.1.2 point release?

-- 
Jai Menon
-------------- next part --------------
A non-text attachment was scrubbed...
Name: avformat-set-correct-pcr.diff
Type: text/x-patch
Size: 2797 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20100727/428da928/attachment.bin>