[FFmpeg-devel] [PATCH 1/5] Move timestamp correction code from ffplay to cmdutils

[Alexander Strange]

On Sep 28, 2010, at 6:28 AM, Stefano Sabatini wrote:

> On date Tuesday 2010-09-28 10:02:11 +0200, Jean-Daniel Dupas encoded: 
>> Le 28 sept. 2010 ? 04:53, Alexander Strange a ?crit :
> [...]
>>>>> cmdutils.c |   27 +++++++++++++++++++++++++++
>>>>> cmdutils.h |   24 ++++++++++++++++++++++++
>>>>> ffplay.c   |   35 ++++++++++++-----------------------
>>>>> 3 files changed, 63 insertions(+), 23 deletions(-)
>>>>> f917f2e1bf6e7633b7030e37b08d24c20f9fd655  0001-Move-timestamp-correction-code-from-ffplay-to-cmduti.patch
>>>>> From 86434ec1c52877f3d1b1a98319ee0b56f551a442 Mon Sep 17 00:00:00 2001
>>>>> From: Alexander Strange <astrange at ithinksw.com>
>>>>> Date: Sat, 26 Jun 2010 21:21:29 -0700
>>>>> Subject: [PATCH 1/3] Move timestamp correction code from ffplay to cmdutils
>>>> 
>>>> ok if tested
>>> 
>>> Applied.
>> 
>> I know I'm late about this one, but wouldn't it be preferable to
>> move this code in libavcore instead ? This code should be usable by
>> all FFmpeg client, not only the built-in command lines utility isn't
>> it ?
> 
> Yes, libavformat or libavcodec looks a more appropriate place where to
> move the code.
> 
> Regards.

I didn't want to add more functions to the API if they weren't necessary.
This function is really a patch over bad values coming from lavf/not matching the decoder well, so improving the initial values might remove the need for it.

Even if not, I think clients should see a good-enough-for-playback timestamp value come out of ffmpeg libraries by default without having to call more functions, so lavc could call this and put it in AVFrame. Then it would still be an internal library function, so it could move again out of here at some point.