[Ffmpeg-devel-irc] ffmpeg-devel.log.20141012

[burek]

[00:00] <ubitux> hum, wasn't the drawtext api thing removed?
[00:15] <akira4> ubitux, [text @ 0x2d15e40] Invalid UTF-8 in decoded subtitles text; maybe missing -sub_charenc option error is given when wrong input is read from the subtitle file right?
[00:17] <ubitux> what's your input?
[00:18] <ubitux> you probably need to do just as it's suggested
[00:18] <akira4> reading an stl file without whitespaces
[00:18] <ubitux> i mean, what's your stl file?
[00:19] <akira4> http://www.eso.org/~lchriste/trans/eyes/subtitles/soundtrack.stl
[00:20] <ubitux> try with -sub_charenc cp1252
[00:20] <ubitux> maybe
[00:20] <ubitux> (before the -i)
[00:21] <akira4> now it gives [text @ 0x25c66c0] Unable to open iconv context with input character encoding "cp152"
[00:22] <akira4> character encoding problem basically?
[00:22] <ubitux> yeah probably
[00:23] <ubitux> try to remove the lines with the special characters for now
[00:23] <akira4> hmm..I wanted to ask
[00:23] <ubitux> iconv -f cp1252 -t utf-8 soundtrack.stl > soundtrack2.stl
[00:23] <akira4> if I use sscanf with whitespaces then if there are none it doesnt try to match them
[00:23] <ubitux> doesn't this work?
[00:24] <ubitux> what do you mean?
[00:24] <ubitux> ah, you mean trying to match " , " vs ","
[00:24] <akira4> yup
[00:24] <akira4> its optional in sscanf to match " "
[00:24] <ubitux> show me your sscanf again
[00:25] <akira4> (sscanf(*buf, "%2d:%2d:%2d:%2d , %2d:%2d:%2d:%2d , %n",&hh1, &mm1, &ss1, &ms1, &hh2, &mm2, &ss2, &ms2, &len)
[00:25] <ubitux> mmmh... try replacing the space with %*1[ ] maybe
[00:26] <ubitux> not exactly sure about the pattern, but you might have a trick with sscanf
[00:27] <akira4> hmm...I read a little about this on stackoverflow http://stackoverflow.com/questions/26319809/taking-formatted-input-using-sscanf-with-optional-whitespaces
[00:27] <ubitux> otherwise you can just do 2 sscanf
[00:28] <ubitux> that's not ideal but well
[00:28] <ubitux> you may try %*[^,] as well
[00:28] <ubitux> but that works only before the ,
[00:28] <ubitux> %*[ ] might actually works
[00:29] <akira4> hmm..it might..I tried %*c
[00:29] <akira4> before and after the ,
[00:32] <ubitux> it doesn't seem to allow no character
[00:32] <ubitux> mmh
[00:32] <ubitux> actually, with the space it works here
[00:34] <ubitux> http://pastie.org/pastes/9640785/text
[00:35] <akira4> hmm...so then the space in the sscanf format shouldn't be a problem
[00:35] <ubitux> in my case it doesn't seem to cause any trouble
[00:35] <ubitux> test by yourself
[00:35] <ubitux> i have to go, see you later
[00:36] <akira4> sure. Thanks :)
[00:36] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:0db1f2c2c78d: avcodec/mjpegdec: sanity check bits
[01:04] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:bd239c9a2e15: lavf: Don't drop both pts and dts if timestamps are invalid
[01:04] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:bec077a1ab1f: Merge commit 'bd239c9a2e151c0aeeb653a571f130aae7320894'
[01:45] <jamrial> michaelni: is it possible to rename the vp90-2-vp90-2-trac3849.webm sample in vp9-test-vectors/ to vp90-2-trac3849.webm?
[01:45] <jamrial> i ask because the name is kinda redundant as is. it hasn't been used in any fate test so far so renaming it shouldn't break anything
[01:58] <michaelni> jamrial, renamed
[02:26] <jamrial> michaelni: thanks
[05:13] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:20df02680c94: avcodec/error_resilience: avoid pointer arithmetic with NULL
[08:16] <arwa> hey i was looking into some codes written by Hyllian in openGL...they have various structures like float4,half etc....do we have similar kind in ffmpeg also?
[10:36] <arwa> do we already have reverse antialiasing filter?
[10:37] <ubitux> arwa: GPUs work with vectors, C doesn't
[10:38] <ubitux> basically when you see op on float4 it means it's actually doing the op on 4 different float
[10:39] <arwa> okay...so, i will have to work with an array of float..!!
[10:41] <ubitux> probably, or just N different float variables
[10:43] <arwa> Okay, I need a reverse antialiasing filter...do we already have it implemented?
[10:45] <ubitux> not exactly sure what this is about
[10:50] <arwa> I also dont know exactly....but in xbr filter they are using this filter also
[10:51] <arwa> it kind of reverses the anti-aliasing effects....to get sharp edges..!!
[10:51] <arwa> So, should I first try to implement this filter and then jump to xbr?
[10:52] <ubitux> arwa: i would say you should try to make a simpler nearest scaler first
[10:52] <ubitux> then move on with the algorithm itself
[10:53] <ubitux> i don't know what the AA effect is defined in xbr
[10:54] <arwa> simple nearest as in ??? just copy the pixel rows and columns?
[10:54] <ubitux> right
[10:54] <arwa> okay
[10:54] <ubitux> so at least you have a filter that scale
[10:54] <arwa> i will proceed with that
[10:54] <ubitux> and you can then focus on the xbr itself
[10:54] <arwa> okay cool :D
[11:18] <arwa> what does av_rescale do?
[11:21] <ubitux> a*b/c
[11:23] <arwa> ohh..okay
[11:23] <ubitux> look at the documentation above the prototypes
[11:23] <ubitux> it's explained
[11:39] <arwa> where is the documentation? I am not able to get it..!
[11:40] <wm4> arwa: the comments in the header file, I think
[11:46] <arwa> its not there :/
[11:48] <arwa> and what does sws_scale do?
[11:50] <ubitux> it's the main function of libswscale
[11:50] <ubitux> "libswscale is a library performing highly optimized image scaling and color space/pixel format conversion operations."
[11:50] <ubitux> it's part of the FFmpeg framework
[11:52] <wm4> arwa: http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavutil/mathematics.h;h=ac94488729fe0478b0b540ccb74fb637a796ba91;hb=HEAD#l86
[12:11] <anshul_mahe> Do we have anything yet to make segment of subtitle
[12:12] <ubitux> -ss and -t doesn't work?
[12:12] <ubitux> -ss or -itoffset maybe depending on the usage
[12:25] <anshul_mahe> I know for sure there is no implementation of it in hls segment.
[12:25] <anshul_mahe> so where should I start implementing subtitle segment
[12:25] <anshul_mahe> I want to create hls segment
[13:42] <wm4> what does ffmpeg.c to let the user add chapters?
[13:42] <wm4> shouldn't there be some text format of some sort?
[13:42] <wm4> that contains the metadata?
[14:43] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:28b829b8bb47: avcodec/eatqi: use av_freep(), do not leave stale pointers in memory
[14:46] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:acea53fea0b8: avcodec/eatgv: use av_freep(), do not leave stale pointers in memory
[18:19] <wm4> so uh how many opw slots are there?
[18:19] <wm4> I thought it was 1
[18:48] <Compnn> wm4 : i also thought it was 1
[20:12] <ubitux> jamrial: gonna push the drawtext api ifdefery?
[20:12] <ubitux> +removal
[20:13] <jamrial> i didn't because i assumed we were going to reintroduce it
[20:19] <ubitux> jamrial: no no, go for it
[20:19] <ubitux> i added the necessary vars to the timeline
[20:19] <wm4> the sinus curve approach to API design?
[20:20] <ubitux> :)
[20:20] <jamrial> ah ok
[20:29] <michaelni> wm4, Compnn, based on your logic we always had 0 slots for gsoc.
[20:29] <ubitux> why are we still accepting opw students?
[20:30] <ubitux> there are less than 10 days
[20:30] <ubitux> remaining
[20:30] <michaelni> what!?
[20:30] <ubitux> aren't we in october?
[20:30] <ubitux> i thought we had until 21 oct. or something
[20:30] <wm4> I thought it was July!
[20:30] <michaelni> AFAIK applicants have until oct 22 to submit an application
[20:31] <wm4> but as it is, only 1 applicant will be accepted, right?
[20:31] <wm4> or are more possible?
[20:31] <michaelni> after that we have until 12 nov to rate applicationa i belive
[20:32] <michaelni> and then depending on  that rating and how many donations pinned to ffmpeg and how many donations without any pinning and the rating of all other projects we will get some slots if i undestand correctly
[20:32] <michaelni> so we dont know how many slots we will have
[20:32] <ubitux> ah, ok my bad
[20:34] <michaelni> so maybe we will get 1 slot, maybe we will get 3, also there are some companies ive asked about if they are interrested in donating and didnt receive an awnser yet so even the number of donations pinned to ffmpeg isnt 100% certain just 1
[20:35] <michaelni> it most likely will be 1 doation pinned to ffmpeg in this round though
[20:36] <michaelni> due to there being not that much time left ...
[20:37] <michaelni> also students can submit multiple applications, so just because a student submits one for ffmpeg doesnt mean she didnt also submit one for other projects
[20:40] <ubitux> what is "an application" exactly?
[20:40] <ubitux> the project for the student?
[20:43] <michaelni> yes
[20:43] <ubitux> alright
[20:45] <michaelni> and ATM theres just 1 applicant who submitted an application and IIUC she submited also applications to other organizations than ffmpeg ...
[20:47] <michaelni> so we dont really have too many applications ATM
[20:48] <michaelni> and in gsoc IIRC we generally had like 30+ applications and got like 8 or so slots
[20:55] <cone-922> ffmpeg.git 03James Almer 07master:59d8050df132: avfilter: remove obsolete FF_API_DRAWTEXT_OLD_TIMELINE cruft
[21:11] <michaelni> also, all opw mentors probably want to subscribe to https://mail.gnome.org/mailman/listinfo/opw-announce-list
[21:19] <Compnn> michaelni : ok ,  thanks for the update :)
[21:20] <Compnn> i didnt know it was based on applications recieved.
[21:37] <michaelni> Compnn, the number of slots, i dont know what it will b based on, but i guess its the opw admins who have the last word on which org gets how many slots
[21:38] <Compnn> no problem :)
[21:39] <llogan> Compn 'n michaelni: thanks for keeping up with the ML. I'm back now.
[21:42] <michaelni> also IIUC, from what ive just read on the opw announce list, orgs generally will not get more than one slot from general / unpinned funding in this round, so it would be 1 or 2 slots for us unless we  find more funding
[22:28] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:1d3c87d6cdc9: postprocess: add YUV 4:4:0 support
[22:28] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:9a460db63be7: avfilter/vf_pp: add yuv 4:4:0 support
[22:28] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:941aaa39e8cd: postproc/postprocess: fix quant store for fq mode
[22:34] <cone-922> ffmpeg.git 03Martin Storsjö 07master:8bef43388132: smoothstreamingenc: Simplify code by removing a redundant variable
[22:35] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:5afe1233c591: Merge commit '8bef43388132b53f59a6e90add18900a3bb4cc60'
[22:43] <Compn> michaelni : ah its unfortunate we have more people than slots. a common problem also with gsoc days :\
[22:47] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:72f801619a1a: movenc: Adjust edit lists to trim out parts of tracks with negative pts
[22:49] <rcombs> there are probably worse problems to have :)
[22:49] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:031285dedf66: Merge commit '72f801619a1ae91969fee9a7d72519422433c998'
[22:50] <michaelni> cone-922, you are lagging, these 2 commits where pushed at the same time not 2 minutes appart
[22:53] <wm4> it has to rate limit itself due to your working speed
[22:59] <cone-922> ffmpeg.git 03Martin Storsjö 07master:95ee4e2ce774: movenc: Add some comments explaining subtle details in writing the edit lists
[23:01] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:687cc836ea01: Merge commit '95ee4e2ce774e0339632d067161596bf3dadfc72'
[23:18] <J_Darnley> Perhaps we can retain the OPW applicants that are genuinely interested in contributing here regardless of whether they get any money.
[23:19] <J_Darnley> The ones working on subs seemed quite... enthusiastic
[23:29] <akira4> ubitux, I tried testing with the sscanf with whitespaces and it  gives a proper output. The stl file had some special characters which it had trouble decoding and was giving a error because of that.
[23:30] <akira4> so then the special characters would be dealt with by the decoder right?
[23:49] <ubitux> akira4: the encoding this? normally it's handled by -sub_charenc
[23:50] <ubitux> unless you meant '|'?
[23:50] <ubitux> in which case yes, it's to be handled by the decoder
[23:51] <akira4> hmm. Alright.
[00:00] --- Mon Oct 13 2014