[Ffmpeg-devel-irc] ffmpeg-devel.log.20131009

[burek]

[00:07] <llogan> ffmpeg still compiles fine on a p2
[00:54] <cone-824> ffmpeg.git 03Lukasz Marek 07master:60136345e680: lavd/pulse: move common code to separate file
[00:54] <cone-824> ffmpeg.git 03Michael Niedermayer 07master:885739f3b4ca: ffserver: strip odd chars from html error messages before sending them back
[04:17] <cone-412> ffmpeg.git 03Timothy Gu 07master:df08fd5e1efa: doc/default.css + doc/t2h.init: New styles
[04:17] <cone-412> ffmpeg.git 03Lou Logan 07master:f115e83c81bf: MAINTAINERS: add Communication section
[09:43] <cone-239> ffmpeg.git 03Carl Eugen Hoyos 07master:f4e85afd99ed: Add pulse_audio_common.h to skipheaders.
[09:43] <cone-239> ffmpeg.git 03Carl Eugen Hoyos 07master:3aabfaa37b24: Fix make checkheaders for vp9 header files.
[11:18] <cone-239> ffmpeg.git 03Luca Barbato 07master:b2148faca9e9: shorten: Extend fixed_coeffs to properly support pred_order 0
[11:18] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:7f22df3a4945: Merge remote-tracking branch 'qatar/master'
[11:43] <ubitux> oh no :(
[11:52] <Daemon404> durandal_1707, ping
[12:03] <durandal_1707> Daemon404, pong
[12:03] <Daemon404> durandal_1707, which environments are having issues with the new strip patch
[12:04] <Daemon404> i know OS X warns, but not much else
[12:05] <durandal_1707> there is ticket on trac and mail on list
[12:05] <durandal_1707> http://forum.doom9.org/showthread.php?t=168963
[12:05] <Daemon404> i feel like my brain cells were lost by reading that post
[12:06] <Daemon404> i do not see any mail on -devel
[12:06] <Daemon404> and i am not subscribe to the user list
[12:07] <durandal_1707>  strip: unrecognized option: -wN
[12:07] <Daemon404> so yes a warning
[12:07] <durandal_1707> Usage: strip [-AnuSXx] [-] [-d filename] [-s filename] [-R filename] [-o output] file [...]
[12:07] <Daemon404> when binutils is too old
[12:07] <Daemon404> i.e. binutils on bsd is ancient
[12:07] <durandal_1707> not here
[12:08] <durandal_1707> not my problem
[12:08] <Daemon404> i dont know if it's work adding a configure check for binutils version just to get rid of some harmless warning
[12:08] <Daemon404> y binutils for 2-3 years ago has -wN at least
[12:08] <Daemon404> my*
[12:08] <durandal_1707> https://lists.ffmpeg.org/pipermail/ffmpeg-user/2013-October/017849.html
[12:09] <Daemon404> oh it errors
[12:09] <Daemon404> but continues
[12:10] <Daemon404> pengvado, feel free to chime in (Since it's your patch)
[12:13] <pengvado> no comment
[12:14] <Daemon404> lol ok.
[12:14] <pengvado> afaict you've correctly diagnosed the problem
[12:14] <pengvado> not that I've tried to reproduce it
[12:15] <Daemon404> i guess ill look into 'solutions'
[12:45] <superware> I'm getting "[rtp @ 01e708a0] Unable to receive RTP payload type 96 without an SDP file describing it", is there NO WAY to play h264 rtp over udp without a physical SDP file? even directly with the ffmpeg lib?
[13:01] <superware> anyone? :|
[13:05] <ubitux>          h += y;
[13:05] <ubitux> -        y = 0;
[13:05] <ubitux> +        y  = 0;
[13:05] <ubitux>      }
[13:05] <ubitux> lol
[13:05] <ubitux> come on..
[13:06] <superware> heh
[13:07] <durandal_1707> ubitux: why you do not like that?
[13:07] <ubitux> http://pastie.org/8389053 i will feel so much better after such diff, at last i can breath, what a relief
[13:08] <superware> ubitux: do you happen to know whether it's possible to play a h264-rtp stream directly from url/mrl, without an SDP file?
[13:08] <ubitux> durandal_1707: because it's kind of stupid
[13:08] <ubitux> superware: no
[13:09] <durandal_1707> ubitux: i suppose you copy pasted this from fork?
[13:09] <ubitux> yet another random k&r patch yes
[13:09] <ubitux> i don't get it honestly
[13:10] <durandal_1707> but how to show your boss that you are doing something useful?
[13:14] <ubitux>      }
[13:14] <ubitux> -    *got_frame      = 1;
[13:14] <ubitux> +    *got_frame = 1;
[13:14] <ubitux>      if ((ret = av_frame_ref(data, &c->pic)) < 0)
[13:14] <ubitux> lol
[13:14] <ubitux> sounds like a reliable script
[13:16] <ubitux> my bad on that one ^
[13:22] <cone-239> ffmpeg.git 03Clément BSsch 07master:56db86a0b2bf: doc/filters: remove extra spacing in vid.stab section.
[13:25] <ubitux> so
[13:25] <ubitux> i'm going to fix the examples
[13:25] <ubitux> what method should i prefer?
[13:25] <ubitux> old school or ref counted?
[13:25] <durandal_1707> "fix"?
[13:25] <ubitux> yeah, they crash and leak
[13:26] <durandal_1707> ah, i was thinking about k&r
[13:27] <iive> the funny thing is that in the above y__=0; the extra space actually indented so the equal sign is under the equal sign on the previous line, aka vertical alignment.
[13:27] <iive> oh, my bad....
[13:27] <iive> that was the intended change.
[13:28] <iive> but yeh... k&r is the second name of diego.
[13:29] <durandal_1707> ubitux: why you reconsider old school?
[13:29] <ubitux> because that's the default
[13:30] <ubitux> ref counted requires an explicit user request
[13:30] <ubitux> aka av_dict_set "refcounted_frames" ...
[13:30] <ubitux> also, in a lot of situation, you don't need ref counting
[13:31] <ubitux> i'm tempted to rename my example submission from a while ago to something like refcounted_decode_video.c 
[13:36] <ubitux> btw
[13:37] <ubitux> in lavfi/src_movie.c (and likely in other demuxing code), we have sth like ret = avcodec_decodec...(); pkt->data += ret; pkt->size -= ret;
[13:37] <ubitux> is this necessary for audio only?
[13:38] <ubitux> it seems to be done for anything !video actually
[13:38] <ubitux> but that source only supports a/v anyway
[13:40] <durandal_1707> that is needed for some audio codecs
[13:41] <ubitux> is this documented in some doxy?
[13:42] <ubitux> ah yeah, in the decode audio
[13:43] <ubitux> i wonder if that couldn't be used for dvb subtitles btw
[13:43] <ubitux> instead of the ugly nb hack in ffmpeg.c
[13:48] <ubitux> so much changes required..
[13:52] <durandal_1707> its needed for shorten, which have ugly design
[13:52] <durandal_1707> and some adpcm codecs iirc
[13:52] <ubitux> ape maybe?
[13:53] <durandal_1707> its because you get several frames in single packet
[13:53] <durandal_1707> nope, ape just have extremly big frames
[13:54] <durandal_1707> and because of that it used CODEC_CAP_SUBFRAMES
[13:57] <iive> ubitux: i thought this is how all decoders are supposed to work, even video ones.
[13:59] <ubitux> no
[14:00] <ubitux> it seems the ret is "fixed" for video
[14:00] <ubitux> like in case the decode video doesn't return the size of the packet
[14:00] <ubitux> see libavfilter/src_movie.c
[14:05] <iive> in what sense fixed?
[14:05] <ubitux>     if (!ret || st->st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
[14:05] <ubitux>         ret = pkt->size;
[14:05] <ubitux>     pkt->data += ret;
[14:05] <ubitux>     pkt->size -= ret;
[14:06] <ubitux> aka "if video, eat all the packet"
[14:06] <iive> but for what purpose?!
[14:07] <iive> does ret hold any other meaning?
[14:07] <ubitux> i guess making sure it works with decoders who don't return the appropriate size
[14:07] <iive> i still can not get any sense out of this...
[14:08] <iive> what are these decoders and why they don't return the appropriate size?
[14:08] <ubitux> i don't know
[14:08] <iive> i can imagine that some packets are stuffed or padded, but the decoder should be able to skip that anyway...
[14:09] <ubitux> ee97982408c83766a3dcb235a673761da629cf99
[14:09] <ubitux> https://ffmpeg.org/trac/ffmpeg/ticket/2556
[14:09] <ubitux> oh it's a bug i opened
[14:18] <cone-239> ffmpeg.git 03Derek Buitenhuis 07master:9e3946aadaef: build: Require yasm 0.8.0 or higher
[14:18] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:4ce82e6fa94c: Merge commit '9e3946aadaef6f7d7ef15c31b5ff1e73d0ba79a1'
[14:24] <durandal_1707> saste: what about interleave patch? and mergeplanes?
[14:24] <durandal_1707> is anybody writing any filter?
[14:36] <cone-239> ffmpeg.git 03Henrik Gramner 07master:ad7d7d4f6a13: x86inc: Remove .rodata kludges
[14:36] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:31d0d355608f: Merge remote-tracking branch 'qatar/master'
[14:55] <cone-239> ffmpeg.git 03Paul B Mahol 07master:19fda0a1bcb1: avcodec/vmnc: use av_freep()
[15:34] <michaelni> BBB, ff_emu_edge_vfix21_sse() contains SSE2 instructions it seems (https://trac.ffmpeg.org/ticket/3017)
[15:48] <cone-239> ffmpeg.git 03Paul B Mahol 07master:8c7bc7d3e50f: avcodec/exr: use avpriv_report_missing_feature()
[15:56] <durandal_1707> wouldn't pts code in decimate be just like one in fps?
[16:40] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:8780f7fbfaac: avfilter/ff_filter_frame_needs_framing: fix handling of AV_NOPTS_VALUE
[16:40] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:0f322feffdf2: avfilter/buffersink: try to fix AV_NOPTS_VALUE
[20:08] <cone-239> ffmpeg.git 03Carl Eugen Hoyos 07master:3aa576513b31: lavd/xv.c: Include XShm.h before Xvlib.h.
[22:14] <llogan> mpeg-2 video long gop. what is the typical gop size?
[22:40] <cone-239> ffmpeg.git 03Michael Niedermayer 07master:83fc6c822b06: avformat/wavdec: Dont trust the fact chunk for PCM
[00:00] --- Thu Oct 10 2013