[Ffmpeg-devel-irc] ffmpeg-devel.log.20141011

[burek]

[00:10] <ubitux> the ass text output of zvbi is kind of... broken.
[00:12] <ubitux> it seems to produce a lot of unknown glyph
[00:13] <ubitux> and is actually considering all texts at the same level
[00:13] <ubitux> but unfortunately, it's used for all kind of annotations
[00:14] <ubitux> http://ubitux.fr/pub/pics/_zvbi-shit-happens.jpg
[00:15] <ubitux> i have when i'm trying to make sure my code is not going to break something and the thing i'm not trying to break actually already is
[00:15] <ubitux> i hate*
[00:16] <ubitux> i guess it's time to sleep
[00:16] <ubitux> gn
[01:00] <cone-984> ffmpeg.git 03Luca Barbato 07master:1d4579e38eca: tcp: Add AVOption support
[01:00] <cone-984> ffmpeg.git 03Michael Niedermayer 07master:ae17a0c25873: Merge commit '1d4579e38ecad578c86516fad2837a273d11b320'
[01:17] <thardin> ShadowJK: 
[01:17] <thardin> ups
[03:56] <cone-984> ffmpeg.git 03Michael Niedermayer 07master:b8a55994f0cb: avcodec/mpeg12dec: print aspect_ratio_info in debug output
[05:15] <cone-984> ffmpeg.git 03Michael Niedermayer 07master:fba894615d69: swscale: support internal scaler cascades
[06:57] <jamrial> libavutil/lzo-test has a couple issues
[06:57] <jamrial> guess nobody noticed/cared till now since it's not run by fate
[07:02] <shresh> hi
[07:02] <shresh> i am shreshtha. I wish to apply for a project under OPW. 
[07:03] <shresh> previously I had a talk with ubitux regarding the same, but now i am undable to contact him again
[07:04] <shresh> can anyone please help me?
[07:05] <jamrial> stick around or come back later if none of the OPW mentors is online. it's early morning for most right now
[07:05] <shresh> ok..thnaks! :)
[09:11] <cehoyos> Hi!
[09:11] <cehoyos> jermy: Did you provide a sample?
[09:12] <cehoyos> Just leave a link here on irc, the channel is logged.
[09:42] <cehoyos> pross: You wrote that the time of the irc meeting was bad for you: What would be an acceptable range for you?
[10:58] <ubitux> shresh: i'm here now
[10:58] <ubitux> what's your question?
[12:27] <cone-940> ffmpeg.git 03Ruoyu 07master:c44d8d9552f9: avutil/lzo: fix resource leak
[13:29] <ubitux> ok i think i fixed the ass outputs
[13:29] <ubitux> does anyone oppose to have full output file in subtitles tests instead of md5?
[13:30] <ubitux> assuming that they are small text files and it's easier to track changes
[13:30] <wm4> why would anyone be opposed
[13:31] <anshul_mahe> I am transcoding ccaption subtitle to ass format, is there any way I can give multiple start and end timed rectangle in single decode call
[13:32] <ubitux> wm4: i don't know :)
[13:32] <ubitux> anshul_mahe: please don't use multiple rectangles
[13:32] <ubitux> just translate everything into a new dialogue each time
[13:33] <ubitux> anshul_mahe: what's the use case exactly?
[13:33] <ubitux> anshul_mahe: are you writing something like libzvbi?
[13:34] <anshul_mahe> I am writing closed caption decoder, I have decoded closed caption but want to save in AVPacket
[13:34] <anshul_mahe> sry in AVSubtitle back when done
[13:35] <anshul_mahe> I am taking refrence of libavcodec/srtdec.c and there it is saved to ass format
[13:36] <ubitux> yes, but please don't add more than one rectangle at once
[13:37] <ubitux> (or more Dialogue)
[13:37] <anshul_mahe> In my scenerio I get multiple timed subtitle in single call how to handle that
[13:37] <ubitux> probably by not consuming the whole packet
[13:39] <anshul_mahe> Its like after decoding that packet I get to know that I have multiple timed packet
[13:40] <anshul_mahe> Should i fake hat, I have still not use the packet
[13:40] <ubitux> make the decode function return 0 until every text event has been written?
[13:40] <anshul_mahe> *that
[13:41] <anshul_mahe> ok
[13:43] <ubitux> i'm not exactly sure if we have such situation already though
[13:44] <ubitux> i need to introduce a avcodec_decode_subtitle3 for text sub like this :(
[13:44] <ubitux> (and for the sub next gen ©)
[14:00] <wm4> ubitux: well, for matroska-style ASS output and AVFrame usage you need that anyway?
[14:01] <ubitux> except that no decoder was actually exploiting that feature
[14:12] <ubitux> http://ubitux.fr/pub/pics/_git-word-diff.png grrr
[14:12] <ubitux> git seems to split word on spaces (obviously)
[14:12] <ubitux> so it includes the first fields even if they don't change :(
[14:12] <ubitux> i need a --char-diff
[14:17] <ubitux> it seems there is a regex mode, cool.
[14:19] <ubitux> yay, git diff --color-words=.
[14:21] <ubitux> http://ubitux.fr/pub/pics/_git-word-diff-yay.png much better.
[14:32] <J_Darnley> ubitux: thanks!  I'll have to remember that
[15:12] <jermy> cehoyos: The file I have is customer content, so it's probably ok for a couple of people to have a look at, but I wouldn't want to put it publicly available. However, I'll see if I can get something benign like colour bars that's encoded with the same software & settings
[15:36] <cehoyos> jermy: Then upload it to our ftp server with a txt file explaining that the file is not meant to be downloadable.
[16:09] <jermy> cehoyos: Ugh. Content is in a Avid-style MXF Op-Atom wrapper (which really should give some extra information about max_b_frames, but doesn't), so it doesn't appear particularly easy to cut this down. I'll have another go on Monday at extracting something sensible
[16:25] <ubitux> wm4: https://github.com/ubitux/FFmpeg/compare/ass-fixes
[16:25] <ubitux> it should output proper ass files, but tests very welcome
[16:27] <ubitux> now i really need to mux a mkv with all kind of broken streams in it
[16:32] <wm4> ass_split.c
[16:32] <wm4> these puns hurt
[16:32] <ubitux> haha
[16:33] <wm4> note that libass and vsfilter have their own heuristics how to distinguish SSA and ASS
[16:33] <wm4> did you check with these?
[16:34] <wm4> vsfilter doesn't parse the Format line at all
[16:34] <ubitux> nope
[16:35] <wm4> well except for that your patches seem to look ok, didn't take that close of a look though
[16:38] <ubitux> i need 2 .ass with SSA and ASS, and mux them both with and without [Events] header in a mkv and add some remux and transcode tests of them
[16:38] <ubitux> because i'm probably gonna break that stuff sooner or later
[16:38] <ubitux> assuming it currently works
[16:39] <cehoyos> jermy: It is nice if a small sample allows to reproduce the problem but there is no size limit in case cutting the sample is not easily possible.
[17:24] <ubitux> JEEB: fyi, yeah ffmpeg supports the <w>:<h>:flags=<flags> form
[17:25] <JEEB> yeh
[17:26] <ubitux> b89ce54e740500715803c2731bf0008f6fd4bfa8 probably
[17:49] <ubitux> Dialoue: Marked=0,0:03:49.03,0:03:50.79,Style1,Comment,0000,0000,0000,,He's nobody suspicious, so rest assured.
[17:49] <ubitux> Dilogue: Marked=0,0:06:37.29,0:06:37.85,Style1,Comment,0000,0000,0000,,Yo!
[17:50] <ubitux> i like how the typo'ed events end up isolated in the CodecPrivate
[18:48] <ubitux> what's the reasoning behind using diff -b?
[18:48] <ubitux> in fate
[18:52] <ubitux> i'm curious about the effect of removing it
[18:53] <ubitux> afaik we don't use plateform specific line endings
[18:53] <ubitux> but maybe i'm mistaken
[18:57] <cone-940> ffmpeg.git 03Christophe Gisquet 07master:140f535517e9: proresenc_ks: use chroma sampling in autoselection
[19:20] <anshul_mahe> for text file it is the obvious option, I dont have much knowledge of how fate work, but fate is comparing between
[19:20] <anshul_mahe> what fate is comparing there.
[19:22] <ubitux> i added a workaround
[19:22] <ubitux> i'll submit a patchset soon
[19:22] <ubitux> just need to craft a broken mkv
[19:22] <ubitux> :p
[19:57] <jamrial> ubitux, BBB: vp90-2-vp90-2-trac3849.webm was uploaded to the fate suit but i don't think the actual test was ever made
[19:58] <ubitux> it wasn't, afaik
[20:07] <rcombs> do `sync_stream_index` and `sync_file_index` in `StreamMap` have any purpose?
[21:30] <BBB> jamrial: hm.. ok, can you enable it? I dont remember what that was
[21:34] <jamrial> commit 14e3025518
[21:34] <jamrial> and i guess. would a normal framemd5 test like the rest of the vp9 ones suffice?
[22:00] <jamrial> looks like it does. i'll send a patch later
[22:32] <cone-940> ffmpeg.git 03Simon Thelen 07master:8c4ca4aa5aa9: ffmpeg_opt: abort if -to <= -ss
[22:56] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:c712d240b168: avcodec/eamad: use av_freep(), do not leave stale pointers in memory
[22:56] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:3521c70dabb2: avcodec/dirac_parser: use av_freep(), do not leave stale pointers in memory
[22:56] <cone-940> ffmpeg.git 03Michael Niedermayer 07master:e24e5986a6f5: avcodec/dct: use av_freep(), do not leave stale pointers in memory
[23:35] <cone-940> ffmpeg.git 03Karl Kiniger 07master:903156aa8a35: vf_drawtext: add missing clear of pointers after av_expr_free()
[00:00] --- Sun Oct 12 2014