[FFmpeg-devel] [PATCH] Parse H.264 is ASF.

[Alex Converse]

On Tue, May 25, 2010 at 6:42 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Tue, May 25, 2010 at 02:39:38AM -0400, Alex Converse wrote:
>> On Sat, May 22, 2010 at 3:26 AM, Reimar D?ffinger
>> <Reimar.Doeffinger at gmx.de> wrote:
>> > On Fri, May 21, 2010 at 10:52:07PM -0400, Alex Converse wrote:
>> >> On Thu, May 20, 2010 at 2:08 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
>> >> > On Wed, May 19, 2010 at 04:36:16PM -0400, Alex Converse wrote:
>> >> >> My H.264 in asf samples[1] seem to put SPS and PPS in its own packet
>> >> >> which creates all kinds of problems when remuxing (notably duplicate
>> >> >> timestamps and an empty avcC). The attached patch turns on the parser
>> >> >> to combine the packets. All my H.264 in asf samples are from the same
>> >> >> source so if people have other samples to verify this on, let me know
>> >> >> (I didn't find anything on samples.mphq.hu)
>> >> >>
>> >> >>
>> >> >> [1] /incoming/h264_aac_zeropad.asf
>> >> >>
>> >> >> Regards,
>> >> >> Alex Converse
>> >> >
>> >> >> ?asfdec.c | ? ?2 ++
>> >> >> ?1 file changed, 2 insertions(+)
>> >> >> cb107db59fdfb193268b15dfd1f5519a0ba1195f ?0003-Parse-H.264-in-ASF-because-SPS-PPS-lives-in-its-own-.patch
>> >> >> From ecd789ea2122615205dcd2250715aae1463ef0b4 Mon Sep 17 00:00:00 2001
>> >> >> From: Alex Converse <alex.converse at gmail.com>
>> >> >> Date: Wed, 19 May 2010 15:57:37 -0400
>> >> >> Subject: [PATCH 3/3] Parse H.264 in ASF because SPS+PPS lives in its own packet.
>> >> >
>> >> > microsoft doesnt fail to come up with new stupidities.
>> >> > are there more than some pps/sps at the begin?
>> >> > if yes your patch is ok
>> >> > otherwise we maybe could try to disable parsing once we are past the pps/sps,
>> >> > would make the code slightly faster
>> >> >
>> >>
>> >> To risk sounding like a total goon, where would be a good place to
>> >> turn the parser off with some sort of reliable synchronization? Would
>> >> it be best to add some sort of PARSE_FLAG_ONCE and do it in the parser
>> >> or would there still be giant parser overhead?
>> >
>> > Is there a problem with just (mis-)using AVSTREAM_PARSE_HEADERS?
>> > I suspect it won't work, but maybe to keep it analogous a AVSTREAM_PARSE_EXTRACT_HEADERS?
>>
>> AVSTREAM_PARSE_EXTRACT_HEADERS seems to sound like the the description
>> of AVSTREAM_PARSE_FULL.
>>
>> > And IIRC there should be some parser already that switches itself of (by directly
>> > returning the input data), but I couldn't find it right now.
>> >
>>
>> I can't seem to find it either.
>>
>> Does attached seem fair?
>
> looks ok to me but please document that this has only been implemented in h264
>

Fixed and applied