[FFmpeg-devel] FW: pts/dts generation and index for mpegts (vob)

[Baptiste Coudurier]

Hi,

On 6/16/2009 1:26 PM, Erik Van Grunderbeeck wrote:
>>   
>> I'm not quite sure if we are talking about the same thing. However, I 
>> (and many others) very definitely have an interest of getting seeking of 
>> MPEG-TS working correctly (which most probably will involve generating 
>> index entries for seeking in the process, if you meant that), and this 
>> on PTS. Unfortunately, I don't have any time for implementing it now.
> 
>> Support for chapter/menu/etc. is for me personally uninteresting, but 
>> others may feel differently.
> 
>> Regards,
> 
>> Ivan
> 
> Yes, its one of the things included I changed by now. Involves looking at
> cell/vts level and looking at pts discontinuities. Unfortunately some of it
> will mean some changes to the api (both at seek level and at extending the
> index generation API for ege clearing). 
> 
> I could probably introduce these bit by bit, but it doesn't make any sense
> to propose all that if the original concept is not accepted.

This feature is definitely welcome.

> To make things easier; Ill summarize them:
> 
> 1) Introduce a new reader (with a new namespace in case of DVD) depending on
> libdvdread (in which I fixed a few things I will forward to them).

By reader you mean demuxer or protocol to speak in libavformat terms.

> 2a) Extend the mpeg ts reader to allow commands to be send to the API
> through a new stream type. That stream type includes "commands" to be
> executed.

Will PS be supported ?
Is STREAM_TYPE_COMMAND needed if you use STREAM_TYPE_DATA and
CODEC_ID_DVDNAVDATA ?

> 2b) On ffmpeg/ffplay level implement a few things (ege stream reset, a
> simple command handling, subtitle color tables)
> 
> 3) on the index generation allow for reset of the index (in case of dvd
> multi vob read) 



> 4) Introduce a new seek flag to allow seek on chapter level, and fix the
> end-generation for duration calculation.

We are reworking seeking API currently during GSOC, any ideas are
welcome. Maybe an AVSEEK_FLAG_CHAPTER ?

> 5) Fix a few other things.
> 
> I feel using libdvdread is a good way forward. Its rather stable and mature,
> and works well on all os's. Furthermore, one simply needs a vm parser to
> have angle and all other features, including pts/dts discontinuities
> working. Rewriting another one is simply silly IMHO.

Well, why not, if it's clean and simple.
Will you work with current PS demuxer ? I think you can add a dvd
special demuxer but it must reuse code in PS demuxer when it can.

> Note that I have full dvd reading now implemented already (including angle
> handling, substream and pts/dts changes) and it works well. I have a few
> things to debug yet (including for some reason that dvd substream subtitle
> handling gets bogus when I switch streams).

That's great, thanks for your work.

[...]

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org