[Ffmpeg-devel-irc] ffmpeg-devel.log.20141004

burek burek021 at gmail.com
Sun Oct 5 02:05:02 CEST 2014


[00:00] <Compn> llogan : k i'll take over mailing list moderation
[00:01] <akira4> Okay got it.
[00:01] <llogan> Compn: thanks
[00:01] Action: Compn pretends like hes sent stickers already
[00:01] Action: Compn scrambles looking for envelopes
[00:02] <Compn> busy week...
[00:03] <hawken> Heh, I like it
[00:04] <hawken> btw, my attempts at MVC have halted, I don't know enough about this stuff and Britz haven't replied yet either
[00:04] <llogan> Compn: no hurry.
[00:04] <hawken> I think it's a better idea to make it from scratch, to be honest, because his licensing is unclear and his code is very messy to say the least. Give me a PM if you want my cleanup :P
[00:07] <Compn> let the trolling commence :P
[00:15] <Compn> i forgot who that is on the trac
[00:15] <Compn> Daemon404 : is it you ?
[00:15] <Compn> or wm4 ?
[00:15] <wm4> me
[00:15] <Compn> ehe
[00:15] <Compn> but seriously, that request to make it known that one developer does not speak for the project was made by libav 
[00:15] <Compn> in response to carls' comments
[00:16] <Compn> from 3 years ago apparently.
[00:16] <wm4> wut
[00:17] <Compn> nothing important.
[00:17] <nevcairiel> what makes some people think that spawning yet another place for discussions is suddenly going to change the world
[00:20] <hawken> Well I'm not against it :3
[00:20] <wm4> nobody is
[00:21] <nevcairiel> there are those that seem to disagree on principle
[00:21] <hawken> meh I don't have a single line of code in either project
[00:21] <hawken> :P
[00:23] <alpa_astero> it seems like a "close, WONTFIX" ticket, frankly
[00:24] <Compn> alpa_astero : no one here is opposed to project unification
[00:24] <Compn> at least as far as all of the devels i've asked about it
[00:24] <wm4> it's still WONTFIX
[00:24] <wm4> or shall we reply "alright we'll get to work ASAP"
[00:25] <Compn> that would be a hilarious response, sure
[00:25] <alpa_astero> Compn: fair; most of the key libav developers are, though, and the projects genuinely do some things differently
[00:25] <wm4> because obviously nobody tried before and these are all completely new and fresh ideas
[00:25] <Compn> alpa_astero : yes, there are policy differences in the projects and a few devels who refuse to work with each other. 
[00:25] <Compn> or refuse to work on servers adminned by certain people :P
[00:26] <alpa_astero> Compn: I'm far too aware of that, alas
[00:32] <hawken> Wow, that's kinda silly :P
[00:33] Action: ubitux is tempted to close the ticket with "ffmpeg already merges libav"
[00:33] <alpa_astero> ubitux: that would be oddly fitting
[00:34] Action: hawken chuckles
[00:34] <hawken> please don't do that :3
[00:35] <ubitux> i can't, there is also a "reconciliate" request
[00:36] <Compn> ubitux : remove the word 'neutral' from your blog post :D
[00:36] <Compn> ehe
[00:36] <Compn> ah maybe enough trolling already
[00:36] <alpa_astero> yeah, trolling doesn't help
[00:37] <alpa_astero> there's too much already
[01:14] <cone-492> ffmpeg.git 03Michael Niedermayer 07master:075a165d2715: avcodec/svq3: Dont memcpy AVFrame
[01:14] <cone-492> ffmpeg.git 03Thomas Volkert 07master:1c4c78ee4073: sdp: add support for H.261
[01:50] <kierank> wm4: are you gjdfgh
[01:50] <wm4> yes
[01:50] <kierank> lol
[01:51] <wm4> duh maybe I should just re-register
[01:51] <wm4> OTOH I don't think anybody cares
[03:21] <jamrial> is the mxf sample in http://samples.mplayerhq.hu/jpeg2000/ supposed to not decode with ffmpeg git?
[03:21] <jamrial> or is it a regression?
[03:32] <Compn> jamrial : one sec, trying old versions
[03:33] <jamrial> just tried 2.4 and 2.3 and both failed
[03:33] <jamrial> i'm compiling 2.2 now
[03:33] <Compn> its possible it never worked
[03:33] <Compn> j2k was broken for a long time
[03:34] <Compn> j2k in mxf is newer thing... that sample is from 2010 , before we even had j2k iirc
[03:34] <jamrial> the errors are both from mxf and j2k, though
[03:34] <jamrial> 2.2 also fails
[03:43] <jamrial> yeah, it never worked
[03:48] <Compn> jamrial : it goes back to the question the other day... what software handles j2k in mxf? :P
[03:48] <Compn> i think someone posted a twitter post with a list of software
[03:48] <Compn> be interesting to see if any of those handle that sample.
[03:49] <jamrial> well, ffmpeg does :p
[03:49] <jamrial> the j2k fate test uses a mxf file
[03:50] <jamrial> which is why i wondered if this sample not working was a regression or not
[04:14] <cone-492> ffmpeg.git 03Michael Niedermayer 07master:8c50704ebf17: avcodec/vorbisdec: Fix off by 1 error in ptns_to_read
[04:15] <Compn> jamrial : thanks for looking into it :)
[04:46] <Compn> aww :p
[04:47] <jamrial> lol compn
[04:47] <hawken> :3
[04:49] <llogan> Compn: please add yourself to MAINTAINERS under "trac troll".
[04:49] <Compn> lol!
[04:50] <Compn> i just find it weird
[04:50] <Compn> when someone closes a bug
[04:50] <Compn> its like everyone stops commenting on it
[04:50] <Compn> closed bugs are still open for comment, why people give up on it :D
[04:50] <llogan> that would be nice
[04:50] <Compn> it would be nice to get rid of bug status all together :P
[04:51] <llogan> maybe i can lock it. then i won't have to get notices about it
[04:51] <Compn> you can remove yourself from cc 
[04:51] <Compn> at least i did a long time ago :)
[04:51] <Compn> which is why i continue to troll 
[04:51] <Compn> because i dont get the notices :)
[04:53] <llogan> anyway, i'll see you at the meeting at my 8 am (I think)
[04:53] <Compn> i have no idea what time that thing is
[04:53] <Compn> :D
[04:53] <Compn> i knew there was something going on tomorrow
[04:53] <Compn> keep forgetting about it
[04:54] <llogan> UTC 16
[04:54] <Compn> then i forget if i'm -4 or -5
[04:54] <Compn> -4 it is.
[04:55] <Compn> so noon.
[04:55] <Compn> dunno if will make it, busy day
[04:55] <llogan> I think you're -5 if you're in central w/ DST
[04:56] <Compn> i'm east though
[04:56] <llogan> oh. i forgot.
[04:56] <Compn> if they taught us time zones in school, i was asleep.
[04:57] <Compn> i mean besides the 4 in the (continental) usa
[04:58] <llogan> they still used mimeograph machines when I was in shitty podunk grade school
[04:58] <llogan> i still remember the chemical smell...
[04:59] <Compn> purple copies
[04:59] <Compn> i remember those
[04:59] <Compn> vaguely
[05:02] <Compn> http://clairenoden.files.wordpress.com/2013/01/images.jpg
[05:18] <Compn> https://www.facebook.com/ffmpegback
[05:18] <Compn> my eyes, the goggles do nothing! 
[05:21] <jamrial> 16 likes
[05:21] <jamrial> heh
[05:29] <cone-492> ffmpeg.git 03Michael Niedermayer 07master:c3d7f00ee3e0: avformat/mpegts: Check desc_len / get8() return code
[05:29] <cone-492> ffmpeg.git 03Michael Niedermayer 07master:27f6da292118: avformat/mpegts: use a padded buffer in read_sl_header()
[05:29] <omero> 'nite all :)
[11:04] <pross> there are libav memes :)
[11:06] <pross> that made my day
[12:17] <ffffffff11111111> hi, what is the purpose of "libavutil/softfloat.c"? TIA
[12:18] <nevcairiel> float emulation for hardware with no native floats
[12:24] <ffffffff11111111> nevcairiel: 10x
[12:26] <cone-450> ffmpeg.git 03Josh Allmann 07master:7d97511e13ef: configure: Add -lrt dependency to avutil's pc file.
[12:26] <cone-450> ffmpeg.git 03Michael Niedermayer 07master:dceebe1a60b1: Merge commit '7d97511e13ef30cd99bae0e36400de055b72918d'
[12:28] <cone-450> ffmpeg.git 03Clément BSsch 07master:138902dfb60f: avformat/assdec: make sure pos is initialized
[12:45] <cehoyos> Hi everybody!
[12:47] <cehoyos> hawken: I tried to reach Britz yesterday by phone, but I forgot that yesterday was public holiday in Germany...
[12:48] <cehoyos> Did you also try Jochen.Britz at dfki.de ?
[12:50] <Compn> hey its our official spokesperson :p
[12:51] <ubitux> hey cehoyos 
[12:51] <ubitux> have you had time to look at trying to propose something for x264/pkg-config?
[12:51] <cehoyos> http://fate.ffmpeg.org/report.cgi?time=20141004023642&slot=x86_64-archlinux-gcc-ubsan
[12:51] <cehoyos> Did anybody else try to reproduce this?
[12:51] <ubitux> (btw, there is the meeting in h-5 right?)
[12:51] <cehoyos> I see the same errors, but the variables in question are never negative when I test them with (if variable < 0)...
[12:52] <cehoyos> There is a meeting, I don't know when;-)
[12:52] <ubitux> 16 UTC today afaict
[12:52] <cehoyos> ubitux: Not yet, I'll be in hospital next week, not sure if this will mean more time or less time...
[12:53] <ubitux> erm. ok.
[12:53] <cehoyos> Imo, the only sane solution is a new function pkg_config_check2() that first checks for pkg_config and then tries the old way
[12:54] <cehoyos> That is if you don't agree with Benoit that using pkg-config as fallback is ok ;-))
[12:54] <ubitux> afaict benoit- understood when i told him why it was broken to use pkg-config as a fallback
[12:55] <cehoyos> More urgent is current x264: Should we just update the necessary version (and be "incompatible" with avconv) or should we add an #ifdef to libx264.c ?
[12:55] <cehoyos> ...udpate in configure
[12:55] <ubitux> also, you probably want to be strict when pkg-config fails to find the package
[12:55] <cehoyos> No
[12:55] <cehoyos> Definitely not, this breaks hell loose on all installations I check
[12:56] <ubitux> it's because it's a change of behavior, and that's expected
[12:56] <ubitux> it also would fail with proper error message to correct the usage
[12:57] <cehoyos> If the change means that existing installation (needlessly!) fail suddenly, we shouldn't change behaviour.
[12:57] <cehoyos> But it is ok if you need pkg-config, I still believe it can be fixed.
[12:57] <cehoyos> You can choose the error message, I will not interfere.
[12:58] <cehoyos> (It is ok if it says "there is no way you install without pkg-config", it already says that now)
[12:58] <cehoyos> No opinions about current libx264?
[12:58] <cehoyos> Should we just update the version in configure or add a workaround for old versions?
[13:00] <ubitux> your fallback will not work as you expect it to
[13:00] <cehoyos> In which situation?
[13:00] <ubitux> in the situation where the user has pkg-config
[13:00] <cehoyos> I already wrote that if a user has a working x264 installation with pkg-config, he cannot choose another one, that is ok
[13:01] <ubitux> and still, current scripts will specify a --extra-{c,ld}flags that won't be honored
[13:01] <cehoyos> Only if pkg-config doesn't find anything usable the old variant is tested
[13:01] <ubitux> and thus might detect another x264
[13:02] <cehoyos> Yes, I understand: This is the compromise I am willing to accept if you insist on pkg-config
[13:02] <cehoyos> (I believe I already said so mulitiple times.)
[13:03] <ubitux> so you are fine with the hack being a pain to maintain in the configure, being even more broken than previously, and having unexpected behaviour in the current scripts?
[13:03] <ubitux> (i have no opinion on the ifdefery, i don't know how recent x264 14x is)
[13:04] <arwa> In writing_filters.doc, there is a command --> "make -j<whatever> ffmpeg" ..... This command is giving me errors (Undefined reference to many things)....Can someone help me out with this?
[13:04] <cehoyos> I am not fine with breaking all current scripts for an alternative that users generally cannot know about. If you insist on allowing this althernative, i will have to accept that it breaks some corner-cases. It will still work for the typical cross-compiling situation.
[13:04] <J_Darnley> arwa: have you run configure?
[13:05] <cehoyos> I can test the issue here;-))
[13:06] <cehoyos> error: x264_param_t has no member named i_avcintra_class
[13:08] <cehoyos> The needed version is 142
[13:08] <cehoyos> It is from January 10th
[13:08] <arwa> yes sir
[13:10] <arwa> I am actually confused about where do I need to run that command line?? In the ffmpeg directory itself right?
[13:17] <cehoyos> Patch sent.
[13:18] <wm4> arwa: yes
[13:18] <wm4> arwa: or the place where you told configure to build
[13:19] <wm4> arwa: depends on what build instructions you followed (they should be clear about make too)
[13:20] <cehoyos> michaelni: I (very) slightly prefer my patch but yours fixes compilation here.
[13:26] <arwa> @wm4 I executed that statement where I configured it. But its giving me errors(Undefined reference to many things). :(
[13:28] <wm4> like what?
[13:39] <arwa> 'avfilter_graph_create_filter' , 'avfilter_graph_free', 'avfilter_graph_alloc',  'avfilter_graph_config', 'av_dup_packet',etc,etc.
[13:51] <wm4> arwa: yeah that sounds bad
[13:51] <wm4> I'm not sure what in your build could have failed or broken... maybe try on a new checkout
[13:52] <arwa> okay...as in install it from scratch?
[14:06] <wm4> arwa: yeah... as it is, it's hard to tell what's wrong on your side, other than "it shouldn't happen"
[14:09] <arwa> okay...i will try doing it. Thanks :D
[14:23] <Compn> cehoyos : is mplayer compilation broken with git ffmpeg ? or just users being dumb again ?
[14:23] <Compn> :)
[14:26] <wm4> mplayer should fix its build system
[14:26] <wm4> this may have worked years ago, but it can't afford this crap anymore
[15:02] <michaelni> Compn, mplayer build works here
[15:03] <michaelni> whats the issue ?
[15:04] <cone-450> ffmpeg.git 03Michael Niedermayer 07master:9734a7a1de30: avcodec/h264: Check mode before considering mixed mode intra prediction
[15:59] <arwa> @wm4 I tried installing it again and now when i am running that command, my laptop is freezing . What is the problem?
[15:59] <wm4> compiling something freezes your laptop?
[16:00] <wm4> maybe it overheated
[16:01] <arwa> yes :/ but my laptop is not getting overheated also
[16:10] <ubitux> hey all
[16:12] <ubitux> hima is another potential opw student looking for the subtitles task; i gave the task of the spruce subtitles demuxer to akira4, but i would like a suggestion for another format to give as a qualification task to hima 
[16:12] <ubitux> does anyone have a suggestion for it?
[16:13] <wm4> hm, libxmp and vapoursynth bridges would be nice to have and probably simple, but I'm not sure if these are good qualification tasks (probably not)
[16:13] <ubitux> it's for subtitles
[16:13] <wm4> but surely there are dozens of trivial formats around that could be implemented
[16:13] <wm4> ah, specifically?
[16:13] <ubitux> yes
[16:14] <wm4> I think I remember some lyrics format
[16:14] <wm4> but not its name
[16:14] <ubitux> mmh
[16:15] <wm4> [00:23.03]Baby, I promise you this
[16:15] <wm4> [00:26.97][02:20.71]I'll send you all my love
[16:15] <wm4> lrc
[16:15] <wm4> http://en.wikipedia.org/wiki/LRC_(file_format)
[16:15] <wm4> and yes, a user wanted it once
[16:17] <ubitux> it looks easy enough for a qualification task
[16:17] <ubitux> wm4: is suppose the extra [<time>] is a non official extension for timer ending?
[16:18] <wm4> I have no idea
[16:19] <wm4> research for the qualifier?
[16:25] <ubitux> ah maybe subtitlesedit and stuff can provide some more
[16:26] <ubitux> oh sbv looks very much alike srt
[16:28] <ubitux> > .SBV is the file format outputted by YouTube's Automatic Timing feature which creates automatically timed captions based on a transcript. 
[16:28] <ubitux> heh
[16:28] <ubitux> is this still a thing?
[16:28] <wm4> hm, supporting youtube subtitle formats would also be nice
[16:28] <wm4> they have two last I checked
[16:28] <ubitux> is it xml based?
[16:29] <wm4> one of them yes
[16:29] <wm4> or maybe they switched to webm hopefully?
[16:29] <ubitux> webvtt?
[16:29] <ubitux> probably, but we already have such demuxer
[16:31] <ubitux> well maybe i'll just go for LRC
[16:33] <ubitux> https://github.com/roowe/lyrics yeah ok, let's say it's useful
[16:33] <ubitux> hima: so, here it is, this is your qualification task
[16:34] <ubitux> supporting the lrc format
[16:34] <ubitux> i provided the irc log i had with akira4, so for any additional question, feel free to ask (publicly here is prefered)
[16:35] <ubitux> the tricky part is probably the duration of the event
[16:35] <ubitux> looking at the .lrc files, it seems they handle this by adding an empty event
[16:36] <ubitux> but that shouldn't be really hard to handle
[16:36] <hima> i am going through the logs you sent trying to understand first 
[16:37] <hima> i will soon bug with my doubts :P
[16:37] <hima> *bug you 
[16:37] <ubitux> sure
[17:33] <saste> hi all
[17:33] <saste> anybody going to attend the ffmeeting?
[17:35] <kierank> i might watch
[17:37] <ubitux> saste: i'll be there
[17:37] <hawken> cehoyos: No, I only sent to gomajobri at googlemail.com. I'll forward it to the address you provided.
[17:40] <hawken> Forwarded my mail :3
[17:40] <saste> please join the #ffmpeg-meeting channel if you want to attend
[17:41] <hawken> Autoreply from that address: "Out of office until 2014/10/13"
[17:41] <hawken> saste: is it okay for me to just join and log it so I can read it later?
[17:41] <saste> hawken, sure
[17:41] <hawken> ^_^
[17:41] <saste> hawken, the log is usually published on the ML
[17:42] <hawken> Ah so not really any need then
[17:42] <hawken> I'm just curious :3
[17:44] <cone-450> ffmpeg.git 03Carl Eugen Hoyos 07master:aaf65bc2a5cb: lavc/libx264: Fix compilation with libx264 version < 142.
[17:44] <cehoyos> compn, michaelni: Compilation of both FFmpeg and MPlayer with old libx264 installed should be fixed now.
[17:52] <cehoyos> Could anybody comment on "Fix fate-ffprobe with --target-path"? Our iPhone works atm and this is supposed to fix the last remaining tests...
[17:56] <cone-450> ffmpeg.git 03Michael Niedermayer 07master:df8394c693d3: avcodec/hevc: fix chroma transform_add size
[18:05] <hima> ubitux: i built the source code and just did run the ffprobe tool with an srt file :)
[18:06] <ubitux> hima: cool :)
[18:06] <hima> should i go through the file srtdec.c to understand how it functions?
[18:08] <ubitux> so, did you read the complete log?
[18:09] <ubitux> i suggest you look at libavformat/vplayerdec.c actually
[18:15] <ubitux> it's very simple
[18:15] <ubitux> you also probably want to get all the subtitles samples from FATE, see https://ffmpeg.org/fate.html
[18:15] <hima> half way through the log 
[18:15] <hima> umm trying this ffmpeg -i in.srt -c copy out.srt. i get an error command not found?
[18:15] <ubitux> ./ffmpeg in the source directory
[18:15] <ubitux> after compilation
[18:15] <hima> okay worked 
[18:15] <hima> in ffmpeg -i Her.srt -c copy out.srt and ffmpeg -Her.srt -c copy out.mkv.  What difference should i observe in the output?
[18:17] <ubitux> out.srt will be a text readable srt file
[18:23] <ubitux> and out.mkv is a matroska (binary) format file that can contain also video and audio streams
[18:23] <ubitux> (as well as all kind of other things such as fonts etc)
[18:23] <ubitux> the .srt will mux the packets with timestamps as strings
[18:23] <ubitux> and .mkv will have packets mux with timestamps in binary, typically
[18:23] <hima> okkay :) yeah i saw that 
[18:23] <hima> so we use ffmpeg -i in.srt -c copy out.srt to send it to the muxer 
[18:23] <hima> and use the ffprobe tool for demuxer?
[18:23] <hima> am i right?
[18:23] <ubitux> the ffmpeg command will call the demuxer to get packets out of in.srt, and will use the srt muxer to put them back into out.srt
[18:23] <ubitux> ffprobe will only use the demuxer, so you can see what's going out of it
[18:23] <hima> ohh okay now i understand 
[18:23] <hima> using the ffprobe we actually saw how thw packets are made! 
[18:23] <hima> right?
[18:23] <ubitux> yep :)
[18:23] <ubitux> it's useful for debugging
[18:24] <hima> ohh cool :D
[18:28] <Compn> i back
[18:28] <Compn> where is meeting ?
[18:28] <hawken> #ffmpeg-meeting
[18:29] <hawken> Compn: Want some backlog in a PM?
[18:29] <Compn> sure
[18:30] <hima> ubitux: please explain h264 packets and ogg packets. These are just formats of packets right?
[18:31] <hawken> Compn: Woops
[18:31] <rcombs> hawken: try a private gist :3
[18:31] <hawken> Heh
[18:32] <ubitux> hima: h264 are just normal packets, where AVPacket.data is a data blob only the h264 decoder (in libavcodec) can understand
[18:32] <ubitux> ogg packets means "packets outputed by the ogg demuxer", it's not very specific
[18:33] <hima> ohh okkay. Meaning this are just packets output of various kind of decoder and demuxers. right?
[18:36] <ubitux> the "ogg packets" could be for different codecs, you're generally talking about "vorbis packets" (a Xiph audio codec) or "theora packets" (anoth video codec you can find muxed in .ogg)
[18:36] <ubitux> but the ogg container can have various other codecs
[18:36] <hima> okay :)
[18:38] <hima> 23:57:14 <@ubitux> the decoded frames box is where you plug your filters typically (to alter the image itself)
[18:38] <hima> 23:57:27 <@ubitux> ok, so this is how the audio and video works
[18:38] <hima> 23:57:43 <@ubitux> subtitles are a PITA so it's a bit different
[18:38] <hima> what does " subtitles are PITA" mean 
[18:38] <hima> ?
[18:38] <ubitux> http://www.urbandictionary.com/define.php?term=pita :P
[18:38] <rcombs> "pain in the ass"
[18:39] <hima> LOL :P ok sorry :P 
[18:44] <llogan> ubitux: maybe webvtt tooptip thumbnails can be a task? unless it is a non-spec thing made by jw player.
[18:49] <ubitux> llogan: it's too specific
[18:50] <ubitux> i prefer the demuxer exercise, because it helps understanding the whole chain
[19:08] <ubitux> hima: btw, when you have a moment, you should add yourself on http://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12#Subtitles just like it's done in other sections (look for "Interested applicants")
[19:08] <ubitux> that way we can track our candidates easily, and it will provide insight for applicants as well
[19:12] <hima> okay i will 
[19:12] <ubitux> thank you
[19:15] <hima> ubitux: can you please explain my qualification task in a lil detail. I think i will understand better now 
[19:16] <ubitux> yes, you have to implement a demuxer for the .lrc lyrics format
[19:16] <ubitux> here you can find a bunch of examples of such files: https://github.com/roowe/lyrics
[19:16] <ubitux> and you can find some information about the format itself at http://en.wikipedia.org/wiki/LRC_(file_format)
[19:17] <Compn> we have example demuxers in the code and in our docs as well
[19:17] <ubitux> hima: so basically, you create a libavformat/lrcdec.c and write something similar to libavformat/vplayerdec.c (for example)
[19:19] <hima> ohh okay got it . I will go through the code and ask my doubts :)
[19:50] <ubitux> cehoyos: why do you drop the tag?
[19:55] <durandal_1707> he is just evil
[20:02] <ubitux> heh i'm looking at the lrc samples of the repository
[20:02] <ubitux> looking for nasty samples
[20:02] <ubitux> i can find some funny ones
[20:02] <ubitux> https://github.com/roowe/lyrics/blob/master/%E5%91%A8%E5%8D%8E%E5%81%A5-%E5%85%B6%E5%AE%9E%E4%B8%8D%E6%83%B3%E8%B5%B0.lrc
[20:02] <ubitux> this one for example...
[20:03] <ubitux> i have no idea wtf is going on in this one
[20:03] <ubitux> it seems a few others have such layout
[20:04] <ubitux> the end of that one as well: https://github.com/roowe/lyrics/blob/master/%E6%A2%81%E9%9D%99%E8%8C%B9-C%27est%20La%20Vie.lrc
[20:07] <ubitux> broken timestamps @ https://github.com/roowe/lyrics/blob/master/%E8%BF%AA%E5%85%8B%E7%89%9B%E4%BB%94-%E7%88%B1%E5%A6%82%E6%BD%AE%E6%B0%B4.lrc
[20:09] <ubitux> not sure how lrc this is but.. https://github.com/roowe/lyrics/blob/master/%E7%8E%8B%E5%8A%9B%E5%AE%8F-%E4%BD%A0%E6%98%AF%E6%88%91%E5%BF%83%E5%86%85%E7%9A%84%E4%B8%80%E9%A6%96%E6%AD%8C.lrc
[20:10] <ubitux> oh i see about the multi timestamp thing
[20:11] <ubitux> it's probably for repeating at different time the chorus or something
[20:23] <ubitux> https://github.com/roowe/lyrics/blob/master/Carly%20Rae%20Jepsen-Call%20Me%20Maybe.lrc and another one with trailing crap
[20:25] <ubitux> should probably end up in extradata
[20:30] Action: saste runs
[20:31] <ubitux> akira4: i forgot to tell you, when you have a moment you should add yourself on http://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12#Subtitles just like it's done in other sections (look for "Interested applicants")
[20:31] <ubitux> that way we can track our candidates easily, and it will provide insight for applicants as well
[20:31] <akira4> cool. I'll do that :)
[20:32] <ubitux> ah, and btw, there is another applicant for subtitles
[20:32] <akira4> yeah?
[20:32] <ubitux> she is working on another demuxer, but i wanted you to be aware of it
[20:33] <akira4> sure.
[20:37] <akira4> oh yeah ubitux here is the link to the diff :  http://pastebin.com/u3YGpHHY
[20:38] <akira4> this one gives the subtitles in srt format but there are still problems with it
[20:38] <ubitux> akira4: you should use git commit -a --amend
[20:38] <ubitux> so your diff is merged with the previous one
[20:38] <ubitux> and you can send the full diff again
[20:39] <akira4> okay. 
[20:39] <ubitux> what's the problem anyway?
[20:40] <akira4> well the first is that the subtitles dont go to the next line in case of a '|'
[20:40] <akira4> that should be the decoder's part right?
[20:41] <ubitux> yep
[20:41] <ubitux> let me look again how that's handle
[20:42] <akira4> cool and the other problem is with the probe function
[20:42] <ubitux> yeah right, you'll have to add a decoder for this
[20:42] <ubitux> but it's kind of trivial
[20:42] <ubitux> look at libavcodec/textdec.c
[20:42] <akira4> Hmm. alright
[20:42] <ubitux> but it can be done later
[20:42] <ubitux> when the demuxer is done
[20:43] <akira4> and regarding the probe function
[20:43] <akira4> the score is very low
[20:43] <akira4> so it says that there is chance of misdetection
[20:43] <ubitux> it probably matches only the extension
[20:43] <ubitux> you should keep the &c
[20:43] <ubitux> and make sure a character is read after the timestamps
[20:44] <ubitux> because right now the sscanf can not work
[20:44] <akira4> I see.
[20:44] <wm4> apropos scores, they're pretty random, but the docs say that AVPROBE_SCORE_MAX/4 + 1 is the recommended "safe" minimum level
[20:45] <wm4> so if a prober reaches this score, and there are no other formats with a higher score, this format will be chosen
[20:45] <ubitux> akira4: a pattern with a '*' is not counted, so you need to make sure all the previous %*d are actually matched by adding a character at the end which you "force" match (by adding it to the counter of matches)
[20:45] <ubitux> that is at least one of the solution
[20:46] <akira4> yeah. I had thought of the counting solution. I'll try that then.
[20:49] <akira4> ubitux,  the link to the diff, http://pastebin.com/A06s2BeC
[20:50] <ubitux> let me know when it works so i can do a proper review
[20:50] <ubitux> i'll guide you for the next part (adding a codec to deal with the | typically)
[20:51] <akira4> okay
[21:04] <akira4> ubitux,  the AV_PROBE_SCORE_MAX that is returned in the probe function is what exactly?
[21:06] <ubitux> it's a score of confidence
[21:06] <ubitux> an arbitrary high value
[21:07] <akira4> okay. so I tried adding the %c at the end of the sscanf line but no luck
[21:08] <ubitux> what does your sscanf looks like?
[21:09] <akira4> sscanf(ptr,%2d:%2d:%2d:%2d , %2d:%2d:%2d:%2d , %c)
[21:09] <akira4> ah the quotes got messed up 
[21:10] <ubitux> do you assign the %c to a character?
[21:10] <ubitux> you're expected what value from sscanf?
[21:10] <ubitux> expecting*
[21:11] <akira4> yeah here is the link. http://pastebin.com/vPN527wd.
[21:11] <ubitux> ok
[21:11] <ubitux> what's the first line of your file?
[21:11] <ubitux> (the .stl file)
[21:12] <akira4> the first line is a command. starts with a $
[21:12] <ubitux> then it can't work
[21:13] <ubitux> p->buf will always be the beginning of the file
[21:13] <ubitux> the buffer will just get larger and larger eventually
[21:14] <akira4> so does skipping lines with '$'  make it better?
[21:14] <ubitux> yes
[21:14] <ubitux> and probably the empty lines as well
[21:15] <akira4> okay so I'll add the skipping part and try again.
[21:15] <ubitux> you can also skip the comment lines
[21:16] <akira4> cool.
[21:16] <ubitux> if you expect something that is not: an empty line, a line starting with a '$' and having a = on the same line, a line starting with a comment, the timestamp line, you can return 0
[21:16] <ubitux> otherwise, you go to the next line (unless it's the timestamp line in which case you can return the max score)
[21:18] <akira4> where do I find the functions for manipulating the buffer? I kinda get lost in the header files :-/
[21:19] <ubitux> what kind of functions are you looking for?
[21:20] <ubitux> you're probably just looking for ff_subtitles_next_line()
[21:21] <akira4> oh yeah. Its quite similar to the srtdec.c 's probe function
[21:21] <cone-44> ffmpeg.git 03Mickaël Raulet 07master:4ba6371a83c9: x86/hevc: get rid off packusdw for ssse3 compatibility
[21:22] <ubitux> note that srtdec supports the utf-16, which is not your case
[21:22] <ubitux> the get_pts function in srtdec works on normal utf-8 though
[21:23] <akira4> hmm. what exactly is the difference in handling utf-16 and utf-8?
[21:25] <ubitux> utf-16 is problematic because characters are on 16-bit instead of 8
[21:25] <ubitux> so... you generally have a 0-character every 2 bytes :p
[21:25] <ubitux> you can't use simple string manipulation on them
[21:26] <ubitux> you don't need to support that anyway
[21:27] <iive> hum... you don't convert it to utf8 before processing?
[21:27] <ubitux> iive: only with the formats we found utf-16
[21:27] <ubitux> srt is one of them, sami as well
[21:28] <ubitux> it needs to be handled in the probe function "manually"
[21:28] <ubitux> with have helper for those, which converts to utf-8 and allow normal text processing
[21:28] <iive> aha, in the probe :)
[21:36] <wm4> hm I should update my patch for bom-less utf-16 probing
[21:44] <akira4> offtopic , but do I need to send a mail to get added in the Interested Applicants section?
[21:44] <ubitux> no, you just register on the wiki
[21:44] <ubitux> and edit the page yourself
[21:45] <akira4> okay. Thanks :)
[21:47] <cone-44> ffmpeg.git 03Janne Grunau 07master:36f3aec3630f: fate-vc1_ilaced_twomv: use -flags +bitexact
[21:47] <cone-44> ffmpeg.git 03Michael Niedermayer 07master:578da1f0dd1b: Merge commit '36f3aec3630f27df64f4ff2b52a1c9ced760eb52'
[22:26] <cone-44> ffmpeg.git 03Michael Niedermayer 07master:550f3e9df341: avcodec/on2avc: Check number of channels
[22:49] <akira4> ubitux, how do I match the '\r' or '$' or '\n' in the buffer that is being read? using FFTextReader?
[22:50] <ubitux> FFTextReader is for supporting utf-16
[22:50] <ubitux> are you sure you want to do that?
[22:50] <ubitux> you can obviously, but it will require you to adjust the read_header function accordingly
[22:50] <ubitux> (that's not exactly difficult though)
[22:50] <akira4> I'd rather try with utf-8 first
[22:51] <ubitux> then don't use that api
[22:51] <ubitux> just do ptr += ff_subtitles_next_line and look at *ptr
[22:51] <akira4> hmm. Okay got it.
[23:01] <wm4> "theres a group wanting ffmpeg back in debian :)"
[23:01] <wm4> lol
[23:02] <BtbN> Called people with common sense?
[23:04] <J_Darnley> people with comon sense don't use facebook
[23:04] <Compn> did you see the meme images they made ?
[23:05] Action: wm4 agrees with J_Darnley 
[23:15] <akira4> ubitux,  I made the changes. Here is the link to the diff : http://pastebin.com/zHnAYhsb
[23:16] <ubitux> *ptr+1  are you sure of this?
[23:16] <ubitux> check if it works when you have a comment line
[23:17] <akira4> okay.
[23:20] <cone-44> ffmpeg.git 03Diego Biurrun 07release/2.4:9eb442cca23b: Update default FATE URL for release/11
[23:20] <cone-44> ffmpeg.git 03Rémi Denis-Courmont 07release/2.4:09c848855a7b: h264: Always invoke the get_format() callback
[23:20] <cone-44> ffmpeg.git 03Rémi Denis-Courmont 07release/2.4:ce0972ecddef: mpeg12: Always invoke the get_format() callback
[23:20] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:4f4f78ea48d8: Merge commit '9eb442cca23bb8d33b3e6d37b8c0e3afff04c37a' into release/2.4
[23:20] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:91ff803a748c: Merge commit '09c848855a7bf11a9ab4a69971cced260b17817d' into release/2.4
[23:20] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:6101187c5501: Merge commit 'ce0972ecddef3eee3c03e393108e5728821e3e37' into release/2.4
[23:28] <cone-44> ffmpeg.git 03Luca Barbato 07release/2.4:570cefb02b4f: hevc: Initialize mergecand_list to 0
[23:28] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:ff24824a7215: h264: reset ret to avoid propagating minor failures
[23:28] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:63660277fd40: Merge commit '570cefb02b4f149c4263677e1a3eea216ee0ec8c' into release/2.4
[23:28] <cone-44> ffmpeg.git 03Michael Niedermayer 07release/2.4:3b364ac18a42: Merge commit 'ff24824a721576195c3b4a711e3ee2af900de795' into release/2.4
[23:45] <akira4> yeah that doesnt work with the comments.
[23:50] <akira4> ubitux,  changed it : http://pastebin.com/KxM3SvNT
[23:50] <ubitux> this code won't work that way
[23:51] <ubitux> typically if you have $foo=bar\n//bla\n$bar=baz
[23:52] <akira4> so basically check if the line contains any of those characters?
[23:53] <akira4> instead of just checking if it starts with it
[00:00] --- Sun Oct  5 2014


More information about the Ffmpeg-devel-irc mailing list