[Ffmpeg-devel-irc] ffmpeg-devel.log.20151212

[burek]

[00:47:21 CET] <michaelni> same as in the other functions, the "depth" of the VLC table, that is how many levels of indirection are in it at max
[01:18:55 CET] <cone-281> ffmpeg 03Ganesh Ajjanagadde 07master:b4f1636a4d77: lavc: typo fix cliping -> clipping, saftey -> safety
[09:13:33 CET] <durandal_1707> ubitux: selectivecolor is in wrong section in docs
[13:05:18 CET] <durandal_1707> michaelni, everyone: ok to push sofalizer?
[13:33:46 CET] <michaelni> durandal_1707, probably yes
[13:55:44 CET] <cone-528> ffmpeg 03Michael Niedermayer 07master:b78885a3c5bc: avcodec/aacsbr: Split the env_facs table
[14:02:28 CET] <cone-528> ffmpeg 03Clément BSsch 07master:09a8b6691c77: doc: fix selectivecolor filter position
[14:35:21 CET] <cone-528> ffmpeg 03zjh8890 07master:c18176bd551b: avcodec/aarch64/neon.S: Update neon.s for transpose_4x4H
[15:23:19 CET] <kierank> anyone familiar with the vlc bitstream reader?
[15:23:20 CET] <kierank> michaelni: ?
[15:23:25 CET] <kierank> still can't get it to read properly
[16:02:48 CET] <cone-528> ffmpeg 03Jean Delvare 07master:ec11cfdf75d8: avfilter/vf_delogo: Use AVPixFmtDescriptor.nb_components
[16:04:41 CET] <michaelni> kierank, did you try to define TRACE ? maybe that shows something that points to where the problem is, that assumes you use vlc functions which support trace
[17:57:44 CET] <kierank> michaelni: thanks, TRACE doesn't match hex editor at all :(
[17:58:40 CET] <jamrial> kierank: icc doesn't seem to like some of your dirac changes: http://fate.ffmpeg.org/report.cgi?time=20151211004622&slot=x86_64-linux-gnu-icc-2013_sp1.0.080
[17:59:40 CET] <jamrial> "dirac_get_arith_uint overflow"
[17:59:49 CET] <kierank> jamrial: fixed in the next commit
[17:59:52 CET] <jamrial> probably miscompilation, since older versions work fine
[18:00:00 CET] <jamrial> ah good
[18:06:39 CET] <kierank> hmm no the vlc reader gets it right
[18:06:40 CET] <kierank> hmmm
[18:20:29 CET] <kierank> michaelni: sorry for more questions but the bits read are correct but the RL value is nonsense, any idea how to debug that
[18:22:30 CET] <kierank> the bits aren't a correct VLC though
[18:22:55 CET] <michaelni> maybe av_log() the values as the init code reads them and see if that matches what they should be
[18:23:34 CET] <kierank> my VLCs are quite large though
[18:23:39 CET] <kierank> 27 bits maximum
[18:24:09 CET] <michaelni> what values are used for depth and bits or what they where called ?
[18:24:11 CET] <cone-528> ffmpeg 03Carl Eugen Hoyos 07master:7c1c453ce9f0: lavf/matroskadec: Set codec_tag also for audio codecs.
[18:24:40 CET] <kierank> michaelni: that's the init: http://pastie.org/private/vdr7pyok5vqg0rhf3lh4pg
[18:24:58 CET] <kierank> http://pastie.org/private/si7nd6pn5egcg3y0iglgw
[18:25:01 CET] <kierank> here is how it's read
[18:26:20 CET] <kierank> GET_RL_VLC with TRACE does  int show  = SHOW_UBITS(name, gb, 24);
[18:26:23 CET] <kierank> doesn't look like enough
[18:27:57 CET] <michaelni> iam not sure GET_RL_VLC supports depth = 3
[18:28:08 CET] <michaelni> only 1 and 2 maybe
[18:29:39 CET] <michaelni> its probably relatively simple to add support for 3 just see how GET_VLC does it
[18:29:48 CET] <kierank> I get a segfault with a larger VLC_BITS
[18:30:54 CET] <michaelni> what value of VLC_BITS segfaults ?
[18:31:03 CET] <kierank> 14
[18:32:03 CET] <kierank> hmm looks like assert isn't working
[18:32:18 CET] <kierank> that was why
[18:34:27 CET] <michaelni> also need_update for GET_RL_VLC() probably needs to be 1 for 14 bits
[18:35:47 CET] <kierank> still gets the read wrong
[18:36:26 CET] <kierank> TRACE shows: 11000011                  195  8   0/-4  rlv
[18:36:28 CET] <kierank> that's not valid
[18:45:51 CET] <BtbN> I lost track of that nvenc discussion. So I basicaly add that exception to the header in nvenc.c, and everything is fine?
[18:52:12 CET] <JEEB> yeah, that thread kind of went woo-woo
[18:52:30 CET] <JEEB> I think it would have to be a license change on the whole piece of software, no?
[18:52:39 CET] <JEEB> if one goes that way
[18:52:58 CET] <JEEB> just like applications have exceptions to GPL for openssl
[19:30:08 CET] <cone-533> ffmpeg 03Michael Niedermayer 07master:625b582d5a91: avcodec/aacsbr_template: Add Check to read_sbr_envelope()
[20:18:21 CET] <cone-533> ffmpeg 03Vicente Olivert Riera 07master:1dcd96f72f77: mips: remove --mips64r6 configure option
[20:18:22 CET] <cone-533> ffmpeg 03Vicente Olivert Riera 07master:88d66f691a4c: mips: add i6400 core optimizations
[20:18:23 CET] <cone-533> ffmpeg 03Vicente Olivert Riera 07master:69e80d6ce42c: mips: i6400 core does have an FPU, so not disable it
[20:54:47 CET] <JEEB> durandal_1707: out of interest, why doesn't this use zimg for the YCbCr->RGB conversion? a bit older version of FFmpeg, but still... https://kuroko.fushizen.eu/samples/410p_with_zimg.log
[21:02:38 CET] <Daemon404> because packed rgb is not a valid output for zscale and it chooses teh shortest path
[21:02:41 CET] <Daemon404> probably.
[21:02:43 CET] <Daemon404> a guess.
[21:02:58 CET] <Daemon404> you may be able to force it
[21:03:38 CET] <durandal_1707> yes, use gbrp, no rgb24
[21:04:17 CET] <JEEB> ah, ok
[21:04:34 CET] <durandal_1707> I'm waiting for simd optimised packed rgb from zimg
[21:07:31 CET] <JEEB> interesting message from swscaler with gbrp
[21:07:46 CET] <JEEB> [Parsed_zscale_0 @ 0x39c0d20] w:16 h:16 fmt:yuv410p sar:1/1 -> w:16 h:16 fmt:gbrp sar:1/1
[21:07:49 CET] <JEEB> [swscaler @ 0x39cec80] Forcing full internal H chroma due to input having non subsampled chroma
[21:08:04 CET] <JEEB> next being the auto-inserted scaler
[21:08:13 CET] <JEEB> which is from gbrp to rgb24
[21:08:26 CET] <JEEB> although I guess that is correct, although that shouldn't have separate chroma planes
[21:08:46 CET] <nevcairiel> unless there is a specialized converter, it always goes through yuv
[21:09:53 CET] <cone-533> ffmpeg 03Paul B Mahol 07master:0a19538bcf40: avfilter: add SOFAlizer audio filter
[21:11:36 CET] <durandal_1707> JEEB: there should use unscaled version
[21:12:00 CET] <durandal_1707> its just shuffling bytes
[21:12:12 CET] <JEEB> yeah, that's what it should be doing :)
[21:12:38 CET] <JEEB> would have been surprised if it was actually doing RGB->YCbCr->RGB for gbrp->rgb24
[21:13:16 CET] <durandal_1707> I mean output should be same when doing gbrp->rgb24 and vice versa
[21:20:21 CET] <JEEB> oh, trying with avfilter input seems to assert in zimg :)
[21:20:30 CET] <JEEB> ffmpeg -f lavfi -i "color=color=white:size=16x16" -frames:v 1 -vf "zscale,format=yuv410p" 410p_zimg.png
[21:21:07 CET] <JEEB>  Assertion `(!(src->plane[0].stride) || (src->plane[0].stride) % zimg::ALIGNMENT == 0) && ("buffer stride not aligned")' failed.
[21:33:24 CET] <cone-533> ffmpeg 03Paul B Mahol 07master:d62e979b8ccb: avfilter/af_sofalizer: display some info when loading sofa file
[21:35:12 CET] <durandal_1707> JEEB: that's lavfi nonsense
[21:58:22 CET] <cone-533> ffmpeg 03Paul B Mahol 07master:7dc469f3423e: doc/general: add missing @tab for ADPCM VIMA
[23:24:28 CET] <cone-533> ffmpeg 03Lvqier 07master:93d336fb076a: avformat/segment: Fix memory leak of cur_entry.filename
[00:00:00 CET] --- Sun Dec 13 2015