[Ffmpeg-devel-irc] ffmpeg-devel.log.20170726

[burek]

[01:18:09 CEST] <cone-695> ffmpeg 03Muhammad Faiz 07master:c8305079dae1: avcodec/rdft: reorder calculation
[03:37:02 CEST] <tdjones> atomnuker: Do you have a sample with excessive clicking with the vorbis psy patch? I want to make sure I've fixed what you are hearing before I resubmit.
[03:38:28 CEST] <atomnuker> yep, sure
[03:41:27 CEST] <tdjones> I think it may be due to scaling the audio based on framesize before the mdct, I have a clip-avoidance patch that may help prevent it.
[03:41:34 CEST] <atomnuker> tdjones: original: https://pars.ee/temp/sample_click.wav encoded with -q:a 0.9 https://pars.ee/temp/test_click.ogg
[03:42:54 CEST] <tdjones> Thank you, I'll try to fix this by morning
[03:43:36 CEST] <atomnuker> take your time, if it requires tuning it'll take a while
[11:34:24 CEST] <jcelerier> if I want to get the length of an audio file (of most common formats : mp3, ogg, flac, wav...) is there a way to do it without parsing the whole damn thing ? 
[11:56:02 CEST] <jcelerier> seems not...
[11:56:11 CEST] <jcelerier> ah sorry, wrong channel
[12:54:34 CEST] <kierank> hmm where is wm4 when you need him
[12:57:18 CEST] <JEEB> on holidays
[13:00:59 CEST] <Shiz> since when does he rest
[14:35:50 CEST] <Compn> jcelerier : unfortunately, headers lie, files are incomplete, and bitrates are variable.
[16:34:23 CEST] <cone-864> ffmpeg 03James Almer 07master:4cebf0fc456c: avutil/frame: add ICC Profile to av_frame_side_data_name
[16:34:24 CEST] <cone-864> ffmpeg 03James Almer 07master:5e4e9afaa127: ffprobe: add support for ICC Profile frame side data
[16:34:25 CEST] <cone-864> ffmpeg 03James Almer 07master:c220fe008c5a: avcodec/webp: add support for ICCP chunks
[17:05:13 CEST] <feliwir> hey, my custom io context is crashing, can someone take a look?
[17:05:18 CEST] <feliwir> line 65: https://gist.github.com/feliwir/951b883167eda0c4ab2964e2f9ef3adf
[17:07:58 CEST] <cone-864> ffmpeg 03Michael Niedermayer 07master:74c1c22d7f0d: avcodec/hevc_ps: fix integer overflow in log2_parallel_merge_level_minus2
[17:07:59 CEST] <cone-864> ffmpeg 03Michael Niedermayer 07master:1b0060031950: avcodec/jpeg2000dec: Fix division by zero in jp2_find_codestream()
[18:10:48 CEST] <J_Darnley> atomnuker: kierank wants me to ask you if you have any more thoughts about why this transform is rubbish
[18:10:54 CEST] <J_Darnley> http://imgur.com/a/4wWX6
[18:11:32 CEST] <J_Darnley> The three images show the 5_3 transform with increasing wavelet depth, from 1 to 3
[18:12:00 CEST] Action: durandal_1707 coding in pure asm
[18:12:09 CEST] Action: J_Darnley wishes he was
[18:12:55 CEST] <J_Darnley> He thinks it might be something about offsets.  Perhaps I'm doing the transform on incorrectly positioned slices.
[18:13:09 CEST] <J_Darnley> I don't think so because haar works perfectly.
[18:13:39 CEST] <J_Darnley> but then haar can be used as a block transform.
[18:15:15 CEST] <J_Darnley> The other transforms require neighbouring data, which we have in ffmpeg because I am doing them in the whole plane.
[19:16:46 CEST] <J_Darnley> kierank and I may have just identified the issue.  The transform function copies out of the source buffer.
[19:17:09 CEST] <J_Darnley> Now I just need to reimplemt that transform.  Shit!
[19:59:34 CEST] <cone-864> ffmpeg 03Clément BSsch 07master:2ba20d799ad2: lavc/htmlsubtitles: simplify 1-char tags case insensitive test
[20:17:57 CEST] <atomnuker> J_Darnley: looks like a stride issue somewhere
[20:21:39 CEST] <J_Darnley> nah.
[20:22:15 CEST] <J_Darnley> it is (probably) the "additional precision" copying at the beginning of 5_3 and 9_7
[20:23:07 CEST] <J_Darnley> they copy the samples within the slice boundaries out of the plane meaning we don't have the neighbor samples available
[22:17:47 CEST] <cone-864> ffmpeg 03foo86 07master:56277f6d1905: avcodec/dolby_e: add 'f' suffixes to floating point literals
[22:17:48 CEST] <cone-864> ffmpeg 03foo86 07master:0689cc247555: avcodec/dolby_e: add AV_CODEC_CAP_CHANNEL_CONF capability
[22:17:49 CEST] <cone-864> ffmpeg 03foo86 07master:5e715b583dab: avcodec/dolby_e: fix potentially undefined pointer arithmetic
[22:17:50 CEST] <cone-864> ffmpeg 03foo86 07master:6029b8a6bbc8: avformat/s337m: fix potentially undefined pointer arithmetic
[00:00:00 CEST] --- Thu Jul 27 2017