[Ffmpeg-devel-irc] ffmpeg-devel.log.20131018

[burek]

[08:57] <Neel> hello
[14:12] <durandal_1707> avio_seek cant use SEEK_END?
[15:18] <cone-893> ffmpeg.git 03Paul B Mahol 07master:a6881765358b: avformat/flvdec: use avpriv_request_sample()
[15:49] <cone-893> ffmpeg.git 03Carl Eugen Hoyos 07master:5b9f39860d83: Support 4k AVC-Intra in mov.
[16:53] <cone-893> ffmpeg.git 03Carl Eugen Hoyos 07master:10c240a08593: avcodec: Add little-endian G726 decoder
[16:53] <cone-893> ffmpeg.git 03Carl Eugen Hoyos 07master:640c70ddc2d9: avformat: Use G726LE decoder for Sun AU files
[17:20] <Compn> the great g726 patch has been applied! 
[17:20] <Compn> :P
[17:20] <Compn> quite the bikeshed on that one
[17:27] <durandal_1707> great?
[17:28] Action: Compn trolls durandal_1707
[17:39] <iive> so, what was the final color?
[17:43] <durandal_1707> it is not about color
[17:45] <Compn> iive : it was about what endianness it should be
[17:46] <Compn> and the shed would look better or worse if the wrong one was used
[17:46] <Compn> the old codecs are crazy :\
[17:47] <{V}> sheds have endianness? no wonder I'm having trouble storing my bike
[17:47] <Daemon404> you merely need an endian-agnostic bike
[17:49] <ubitux> i guess sox has a different (default?) endianess
[17:49] <ubitux> ?
[17:49] <iive> ok, front wheel first or back wheel first?
[17:49] <iive> ;)
[17:50] <iive> maybe I should take a look of the thread. But generally the output should be in the native format of the codec. Meaning, not doing additional conversion.
[17:51] <durandal_1707> some encoders just changed order
[17:52] <durandal_1707> imagine someone do this with flac....
[17:52] <iive> well, there won't be a way to detect it, would it?
[17:52] <Daemon404> you could use a heuristic pretty well
[17:53] <Daemon404> but it would be slow
[17:53] <Daemon404> and error prone
[17:53] <iive> it is not the job of the decoder to make guesses.
[17:54] <iive> is g726  looseless?
[17:54] <Daemon404> if the format itself has no endianness, it's 'wrong' of us output it as "BE" or "LE"
[17:55] <iive> you will have to output it, eventually.
[18:14] <iive> ok, so g726 is mac containers is BE and in other containers is in LE format. carl proposes a patch that creates a new decoder that outputs in a different endianes. Or to be exact, same decoder just label the output data as different format.
[18:15] <iive> it is always better to use a reliable method than making wild guesses.
[18:15] <iive> as of should it be different fourcc, codec_id or flag in extra-data... it's just a matter of preference.
[18:16] <Compn> the bikeshed, it continues!
[18:25] <iive> of course... it must be in blue.
[18:37] <michaelni> ubitux, the following 2 subtitle patches seem to have no reply, dunno if they are still relevant but if they are they should get some kind of agree/disagree/obsolete/... reply "0808 17:03 Nicolas George  (5.4K) 
[18:50] <wm4> michaelni: that's nicolas' charset non-sense, I hope it won't ever make it
[18:50] <wm4> not in this form
[18:50] <durandal_1707> reply with: rejected
[18:50] <wm4> good intentions, horrible outcome
[18:51] <Compn> wm4 : did you look at #libav-devel ban list? :P
[18:51] <wm4> aha
[18:51] <wm4> fun
[18:51] <durandal_1707> who is on that list?
[18:51] <wm4> mini
[18:52] <wm4> the ban hostmask doesn't even match anymore
[18:52] <Compn> wm4 : do you know if libav is working on subs? i saw elenril ask about it
[18:52] <wm4> no, libav gives a shit about subs and it shows
[18:53] <Compn> missing a word in there 
[18:53] <wm4> what word
[18:53] <Compn> they give a shit, or they dont give a shit ?
[18:54] <michaelni> I think someone should talk with nicolas about the subtitle/charset stuff, i mean maybe not ML and maybe not email but not sure
[18:54] <durandal_1707> in rl?
[18:55] <Compn> what country is he in ?
[18:55] <ubitux> fr
[18:55] <wm4> oh right
[18:55] <wm4> +don't
[18:55] <ubitux> michaelni: i've not that much clue on what should be done for charset honestly
[18:55] <Compn> michaelni : your idea to automatically guess charset ?
[18:55] <wm4> it needs utf-16 support
[18:56] <ubitux> yes right
[18:56] <Compn> only seen one utf16 sample?
[18:56] <wm4> the subtitle parsers can read any encoding just fine, and the application can detect and convert it
[18:56] <wm4> except utf-16, because that is not an ascii superset
[18:56] <Compn> so whats missing? just a way to select the charset? or the entire iconv conversion ?
[18:56] <wm4> but nicolas wants to add some naive super broken and enforced charset detection
[18:57] <wm4> (and conversion)
[18:57] <Compn> wm4 : well thats because ffmpeg is both a program and a set of libraries. the 'controlling application' in this case would be ffmpeg
[18:57] <Compn> so that feature would be useful for ffmpeg (the application)
[18:58] <durandal_1707> who and why put stuff on our banlist?
[18:58] <wm4> another case where ffmpeg (the application) causes sub-par stuff to be added to the libraries
[18:58] <michaelni> Compn, ubitux i dunno subtitle stuff that well either, so iam not ideal for choosing a design
[18:58] <Compn> durandal_1707 : one guy was spammer, dunno bout the other bans
[18:58] <wm4> lol I'm probably the only one who uses the subtitle stuff (and not from ffmpeg.c/ffplay.c)
[18:59] <Compn> durandal_1707 : can check irc logs, might bei n there
[18:59] <Compn> yes i dont think vlc or mplayer use subtitle stuff yet
[19:00] <Compn> should ask j-b what , if anything, he wants from lavf subtitle
[19:00] <Compn> wm4 : probably you are the only user so far :D
[19:00] <wm4> feels bad
[19:00] <ubitux> :)
[19:00] <ubitux> that's cool wm4!
[19:00] <ubitux> :)
[19:00] <j-b> well parsing subtitles is easy. Rendering is very very hard
[19:00] <ubitux> you helped making vobsub demuxer better ;)
[19:01] <Compn>  j-b :want subtitle render in video filter ?
[19:01] <wm4> j-b: libass ftw
[19:01] <j-b> well, libass is a bit slow, and the fontcache, man!
[19:01] <wm4> it also has a nice library name, providing for endless fun puns
[19:01] <wm4> it's slow?
[19:02] <wm4> but yeah... fontconfig
[19:03] <durandal_1707> 75$ for specification
[19:04] <Compn> j-b : or do you want some other way to render subs ?
[19:04] <durandal_1707> yes via vlc
[19:04] <Compn> wm4 : how do you render subs ?
[19:04] <Compn> i mean
[19:04] <wm4> Compn: libass
[19:04] <Compn> do you render subs without libass ?
[19:05] <wm4> no
[19:05] <Compn> like mplayer has its old renderer...
[19:05] <Compn> ok
[19:05] <wm4> the old renderer was horrible
[19:05] <wm4> like everything in mplayer, really
[19:05] <Compn> it was a lot of hacks to deal with crappy 2001 hardware
[19:05] <Compn> voodoo cards
[19:05] <Compn> you werent around for the 3dfx cards wm4, you dont know how bad video was back then
[19:05] <j-b> Compn: the issue is that it needs harfbuzz+fribidi+freetype+font-lookups to properly render subs
[19:06] <j-b> Compn: and the only ones who do have that are gtk, qt or webbrowsers
[19:06] <Compn> fribidi :\
[19:07] <durandal_1707> saste: write ripple filter?
[19:08] <wm4> there was some talk (and even code) to use native APIs in libass (on windows/osx), but, it's probably well possible, but it went nowhere
[19:10] <Compn> so we should just use libass until an ffmpeg wheel can be reinvented ?
[19:10] <Compn> since it has fribidi and freetype and fontcache
[19:13] <wm4> ffmpeg already has vf_ass, but of course it doesn't work the way you'd expect (you have to provide an external subtitle file to vf_ass directly)
[19:14] <durandal_1707> and there is no subtitle filtering
[19:15] <ubitux> and there is not really anyone doing relevant work on sub :(
[19:16] <durandal_1707> or anything else
[19:16] <Compn> wm4 : do you want it to automatically use embedded subs ?
[19:16] <Compn> i mean, that seems like a simple addition
[19:16] <Compn> we can shoot for that :P
[19:17] <ubitux> durandal_1707: i'm in a learning stage, i'll be back to productive mode later
[19:17] <wm4> bikeshedding, dude
[19:17] <Compn> j-b : lavfi in vlc, when ? :)
[19:17] <j-b> Compn: when there is a need?
[19:17] <durandal_1707> vlc have own filters
[19:18] <j-b> MOAR filters is always better
[19:18] <Compn> j-b : yadif? :P
[19:19] <j-b> Compn: already in.
[19:19] <Compn> ok :)
[19:19] <j-b> Compn: we have adjust, gradfun, yadif, some grain adder, hqdn3d, IVTC (and so many deinterlacer)
[19:20] <j-b> I don't know what people need in addition to that
[19:20] <ubitux> mandelbrot
[19:21] <j-b> ?
[19:21] <ubitux> j-b: ffplay -f lavfi mandelbrot
[19:21] <wm4> maybe vf_pp7?
[19:22] <j-b> ubitux: useless, it seems :)
[19:22] <Compn> wm4 : someone ported that ?
[19:22] <ubitux> j-b: you're always so boring :(
[19:22] <wm4> Compn: it's already ported! it works with vf_mp!
[19:22] <Compn> ehe
[19:22] <j-b> ubitux: right... Did you ever launch VLC lately?
[19:22] <j-b> wm4: what is pp7 ?
[19:22] <ubitux> yes
[19:23] <j-b> ubitux: 2.1.0 ?
[19:23] <ubitux> but only through cvlc
[19:23] <wm4> j-b: some crap nobody wants to port
[19:23] <wm4> so the (very unclean) libmpcodecs wrapper stays in lavfi
[19:23] <ubitux> j-b: yes that's my current version
[19:23] <j-b> cvlc --video-filter puzzle 
[19:23] <j-b> cvlc --video-filter puzzle  <input>
[19:24] <ubitux> :D
[19:24] <j-b> use your mouse know :)
[19:24] <wm4> useful... should be ported to lavfi ASAP
[19:24] <ubitux> oh shit it's interactive
[19:24] <j-b> ubitux: you can use the rotation version too
[19:24] <ubitux> it's extremely slow though
[19:24] <j-b> so that a click or mouse rotates +/-90
[19:25] <j-b> ubitux: slow? no shit!
[19:25] <ubitux> please optimize this
[19:25] <j-b> riite
[19:25] <ubitux> that will be the only reason i'll use vlc now
[19:25] <j-b> well, VLC 2.2 has vhs effect
[19:25] <Compn> j-b : wow thats a fun feature :)
[19:25] <Compn> video puzzle
[19:26] <j-b> Compn: the hard way is that one:
[19:26] <j-b> cvlc --video-filter puzzle -V caca <input>
[19:26] <Compn> caca :D
[19:26] <j-b> it actually worsk
[19:26] <j-b> but interactivity is limited
[19:27] <durandal_1707> what deinterlacers?
[19:27] <j-b> ubitux: 2.2 has vhs and oldmovies effects. You said boooring?
[19:27] <wm4> why does cvlc not close when closing it with the window manager's close button?
[19:27] <ubitux> j-b: :D
[19:27] <j-b> wm4: because of a bug on VLC xv output, IIRc
[19:28] <j-b> durandal_1707: that was a question for me?
[19:28] <Compn> wm4 : so no downstream wants to use ffmpeg subtitle options.
[19:28] <Compn> wm4 : so , we focus on ffmpeg subtitle for ffmpeg application only.
[19:28] <wm4> the default video driver doesn't seem to implement the required X protocol (so that the application closes itself on request) while caca simply ignores it
[19:28] <ubitux> oh i can reshuffle when i'm done
[19:29] <Compn> caca is laggy in win32 :)
[19:29] <Compn> vhs video effects? i love vhs...
[19:30] <j-b> ubitux: note that there are 4 modes
[19:30] <j-b> ubitux: 0 (jigsaw puzzle), 1 (sliding puzzle), 2 (swap puzzle), 3 (exchange puzzle)
[19:30] <wm4> I certainly find it interesting that video filters can capture mouse input
[19:31] <ubitux> j-b: how do you set it?
[19:31] <wm4> what if several things want input?
[19:31] <Compn> wm4 : isnt it like dvd menu ?
[19:31] <j-b> --puzzle-mode <i>
[19:31] <j-b> wm4: --video-filter zoom is interactive too
[19:32] <j-b> maybe it is called magnify now, though
[19:32] <wm4> it's called magnify
[19:33] <Compn> thats neat
[19:33] <Compn> people were asking how to make real pan and scan versions of films
[19:34] <durandal_1707> you mean zoom & pan?
[19:34] <j-b> swscale ftw
[19:35] <Compn> durandal_1707 : i dont know what they call it in other countries, here its always p&s , make a 4:3 film from a 16:9 film by cropping and panning
[19:35] <j-b> ubitux: and for more useless thing, vlc --video-filter psychedelic
[19:35] <durandal_1707> what that does?
[19:36] <j-b> I cannot explain
[19:36] <durandal_1707> Compn: that can be done with lavfi
[19:36] <j-b> It's a LSD simulator
[19:36] <wm4> lavfi still has a long way to go
[19:36] <ubitux> mmh
[19:36] <durandal_1707> wm4: meh meh
[19:36] <ubitux> i'm disappointed j-b 
[19:36] <j-b> ubitux: wut?
[19:36] <wm4> durandal_1707: does lavfi have an lsd simulator?
[19:36] <durandal_1707> geq
[19:36] <cone-893> ffmpeg.git 03Carl Eugen Hoyos 07master:6e65e3483706: lavf/riff: Fix a comment.
[19:36] <ubitux> j-b: it doesn't messed up the sound
[19:36] <j-b> ubitux: lol
[19:37] <ubitux> j-b: also, i think you can do the same with lavfi actually
[19:37] <j-b> ubitux: tbh, the biggest mess with all this, is to decide what are the ones that are important
[19:37] <j-b> ubitux: of course.
[19:37] <ubitux> with hue and an eval
[19:37] <ubitux> and for the moving one, should be doable as well
[19:37] <j-b> is a grain filter more or less important than a denoiser?
[19:38] <Compn> j-b : i get usermails to webmaster at mphq talking about getting viruses from downloading mplayer on trojan sites. what do you tell vlc users who get same problem ? :(
[19:38] <j-b> what denoiser should be the most important?
[19:38] <j-b> hqdn3d, gaussianblur, or temporal blur?
[19:38] <ubitux> you're gonna make Daemon404 angry
[19:39] <j-b> Compn: You don't imagine how long the answer to this question is going to be.
[19:39] <Compn> j-b : dont you have automated form email reply yet? ::)
[19:39] <Compn> i could just copy / replace vlc with 'mplayer' and send. :)
[19:40] <Daemon404> gaussian blur is considered a denoiser? LOL
[19:40] <j-b> no, because it is not there fault
[19:40] <j-b> ubitux: it was done a bit on purpose,  :)
[19:40] <Compn> its the wild west of the internet fault
[19:40] <Compn> but i like the wild west so , on it goes
[20:12] <durandal_1707> phosphor deinterlacer?
[20:16] <durandal_1707> there are bunch of nice (toy) filters to port from vlc, instead of libmpcodecs crap
[20:16] <cbsrobot> durandal_1707: you could add a mda filter - if lsd is too boring
[20:19] <durandal_1707> cbsrobot: mda?
[20:19] <cbsrobot> durandal_1707: http://en.wikipedia.org/wiki/3,4-Methylenedioxyamphetamine
[20:19] <durandal_1707> that is not cheap enough
[20:20] Action: cbsrobot always reads durandal_1707's email as "one" "mda" at gmail
[20:20] <cbsrobot> never thought it could mean something else ...
[20:27] <durandal_1707> i have bunch of filters not applied because of bikeshed
[21:27] <cone-893> ffmpeg.git 03Carl Eugen Hoyos 07master:9860fb445a90: Fix decoding of G.726 in AIFF.
[21:54] <ubitux> http://xiphmont.livejournal.com/61334.html #daala, part4
[21:55] <durandal11707> and? anything?
[23:15] <cone-893> ffmpeg.git 03Stefano Sabatini 07master:8efed2993733: MAINTAINERS: add my name to a few entries
[23:41] <cone-893> ffmpeg.git 03Lukasz Marek 07master:1421ee263742: lavd/fbdev: pass proper memory map length
[23:41] <cone-893> ffmpeg.git 03Lukasz Marek 07master:fb74c7e0446f: lavd/pulse_audio_dec: apply cosmetic changes
[23:41] <cone-893> ffmpeg.git 03Stefano Sabatini 07master:f6b56b1f26a1: lavd/fbdev: use av_str2err() macro for printing error messages
[00:00] --- Sat Oct 19 2013