[Ffmpeg-devel-irc] ffmpeg-devel.log.20131024

[burek]

[00:11] <Compn> burek : http://www-staging-uk.webex.com/fileadmin/GLOBAL_PDF/seminars/recording/
[00:17] <Daemon404> webex is just h264 SVC
[00:50] <michaelni> BBB, maybe http://fate.ffmpeg.org/history.cgi?slot=x86_32-debian-kfreebsd-gcc-4.4-cpuflags-sse
[00:59] <Zeranoe> Is there any way to disable usage of Thread local storage in a FFmpeg build? Or do all instances of __thread need to be removed or something?
[01:14] <michaelni> "git grep __thread" has no matches 
[01:16] <Zeranoe> oh... so it must be in the compile then?
[01:20] Action: michaelni has no idea where the thread local comes from
[01:21] <michaelni> do you know some line number or something else that could help pinpoint where its from ?
[01:23] <Zeranoe> This is everything I've been given: http://paste.debian.net/61013/
[01:26] <iive> tls... that trace doesn't look like ffmpeg thing.
[01:27] <Zeranoe> I wonder if I need to recompile my toolchain with --disable-tls...
[01:39] <BBB> michaelni: hm... ok, I guess the problem is that it's a box that can do more ;)
[01:39] <BBB> anyway
[01:39] <BBB> I tried to help the guy I think it's fixed now, I'll send a new patch if it works
[01:39] <BBB> ubitux: pong
[01:50] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:2f31b73a11d6: avcodec/mpegaudio_tablegen: fix build with hardcoded tables
[02:25] <BBB> does anyone have a katmai (intel pentium III) lying around?
[02:26] <llogan> maybe someone on the mailing list will have one
[02:27] <BBB> can you ask for me? it'd be really helpful for me
[02:27] <BBB> (and yes it's ffmpeg-related :) )
[02:27] <BBB> it's actually somewhat tangentially related to a bug report on ffmpeg-users
[02:29] <llogan> i can if you're unable to atm. what are your requirements and how should they contact you?
[02:33] <BBB> reply-to rsbultje at gmail.com, requirement is a intel pentium III (katmai) with some method of being able to put software on it (i.e. either a compiler and assembler, or a network connection to another computer that has one), and a terminal so they can a terminal application
[02:35] <BBB> ideally they are capable of compiling ffmpeg on it, apply a patch and run a subset of fate to show it doesn't break
[02:35] <BBB> but I can write a simpler test if that's not possible
[02:36] <llogan> ...and just to be soure i was referring to ffmpeg-devel mailing list.
[02:36] <llogan> *sure
[02:38] <BBB> oh I can write that myself, I thought you meant ffmpeg-users
[02:38] <BBB> (wider audience, no?)
[02:39] <llogan> heh. i was wondering... 1347 (-devel) vs 1802.
[02:39] <BBB> oh
[02:39] <BBB> hm...
[02:39] <BBB> ok
[02:39] <BBB> then I'll ask on -devel
[02:39] <llogan> write it on -devel and i'll spam -user with it
[02:39] <BBB> ok ty
[02:42] <llogan> and if we get nothing i can make a post on the web site
[02:44] <llogan> libav-user is 659 and 194 for cvslog for the curious
[03:02] <BBB> llogan: I'll send tomorrow, don't wait for it - waiting for some more results
[03:18] <BBB> ubitux: michaelni: patch sent that I think properly fixes it. it's not ideal since it'll slow down a little more on sse-only cpus, but I don't think anyone cares at this point
[03:19] <BBB> now back to simd again
[03:19] <BBB> (vp9 simd)
[07:27] <ubitux> BBB: yeah sorry, i had a question about simd but i solved it
[07:27] <ubitux> i fixed the rounding, and i'm getting closer to the final result
[08:54] <cone-897> ffmpeg.git 03Stefano Sabatini 07master:ff9d11f95deb: lsws: use av_pix_fmt_get_chroma_sub_sample()
[08:54] <cone-897> ffmpeg.git 03Lukasz Marek 07master:759388aa717a: lavd: add fbdev output device
[08:54] <cone-897> ffmpeg.git 03Lukasz Marek 07master:c720d0d9df8e: lavd: rename fbdev.c into fbdev_dec.c
[08:54] <cone-897> ffmpeg.git 03Lukasz Marek 07master:a547c496e99e: lavd/fbdev: move common code to separate file
[09:45] <cone-897> ffmpeg.git 03Paul B Mahol 07master:426abbb632b0: avfilter/libmpcodecs/vf_eq: remove useless storage class specifier in empty declaration
[09:48] <durandal_1707> michaelni: flvdec have unused flv_data_packet() function, merge failure?
[09:48] <ubitux> iirc it's wanted
[09:48] <ubitux> wi
[09:48] <ubitux> i've heard it's broken or untestable
[10:03] <durandal_1707> why libavcodec/mpegaudio_tablegen.c overrride CONFIG_HARDCODED_TABLES?
[10:29] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:049169535843: avdevice/fbdev_enc: avoid declaring variables in for()
[10:42] <cone-897> ffmpeg.git 03Daniel Kang 07master:0e730494160d: avfilter: x86: Port gradfun filter optimizations to yasm
[10:42] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:1ea28ffc4dbc: Merge commit '0e730494160d973400aed8d2addd1f58a0ec883e'
[10:53] <cone-897> ffmpeg.git 03Vittorio Giovara 07master:fb13fe8342fa: golomb: reduce scope of a few variables
[10:53] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:caac12bd4984: Merge remote-tracking branch 'qatar/master'
[10:53] <ubitux> "In this section maybe denoise_row[ 8] and desnoise_row[ 9] could be
[10:53] <ubitux> aligned at the = instead of the [] ?"
[10:54] <ubitux> lol
[10:54] <ubitux> sounds important
[11:01] <michaelni> BBB, where did you sent that patch ? i dont see it on any list ?
[11:07] <durandal_1707> ubitux: ?
[11:09] <ubitux> durandal_1707: cosmetics discussion
[11:11] <durandal_1707> ubitux: real programmers do that every day
[11:22] <saste> durandal_1707, and finally they do nothing else
[11:33] <cone-897> ffmpeg.git 03Timothy Gu 07master:0be74cf0f5b1: doc/encoders: add libvorbis doc
[11:34] <cone-897> ffmpeg.git 03Timothy Gu 07master:6879fe426593: doc/codecs: update skip_alpha documentation
[11:34] <cone-897> ffmpeg.git 03Timothy Gu 07master:683d9cb1f282: lavc/libmp3lame: fix alignment and capitalization in AVOptions
[11:34] <cone-897> ffmpeg.git 03Timothy Gu 07master:07e7bc9cbda1: lavc/libmp3lame: add comments about CBR/VBR modes
[11:50] <saste> durandal_1707, [Parsed_displace_2 @ 0x3416de0] [framesync @ 0x3416ec0] Buffer queue overflow, dropping. Last message repeated 877 times
[11:51] <durandal_1707> so?
[11:52] <durandal_1707> 'full uncut console output missing'
[11:54] <cone-897> ffmpeg.git 03Ingo Brückl 07master:94d707af3c8d: build: remove pointless condition
[11:54] <cone-897> ffmpeg.git 03Derek Buitenhuis 07master:5dfdffeb6fe9: tree-test: Don't return restricted exit codes
[11:54] <cone-897> ffmpeg.git 03Derek Buitenhuis 07master:cbbd0ac764b4: base64-test: Remove posibility of returning restrcted exit codes
[11:54] <cone-897> ffmpeg.git 03Derek Buitenhuis 07master:f80ca1cb002c: dct-test: Remove possibility of returning a restricted exit code
[11:54] <cone-897> ffmpeg.git 03Derek Buitenhuis 07master:b2fdf3fc35b6: fft-test: Remove possibility of returning restricted exit code
[11:55] <saste> ffmpeg -i matrixbench_mpeg2.mpg  -f lavfi -i nullsrc=s=hd720,geq='r=128+50*sin(2*PI*X/800):g=128+50*sin(2*PI*X/800):b=128+50*sin(2*PI*X/800)' -f lavfi -i nullsrc=s=hd720,lutrgb=128:128:128 -lavfi '[0]scale=hd720,setsar=1[t0];[t0][2][1]displace' -y out.mkv -nostats
[11:55] <saste> durandal_1707, ^^
[11:55] <saste> if I use a synthetic video it doesn't happen
[11:57] <durandal_1707> and it happens for first/last/all frames?
[11:58] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/0.10:8096691a681b: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/0.11:5df35d9a21d5: avformat/wavdec: Dont trust the fact chunk for PCM
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/0.11:87f719a03f23: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/0.11:43af9624fe11: avformat/mov: force parsing of headers if stts is absent
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/0.9:1b3803e8c11f: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.0:1cfcc9ac8ec6: avformat/mov: force parsing of headers if stts is absent
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.0:e3b2b884fc33: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.0:3c142c196d41: avformat/matroskadec: only set r_frame_rate if the value is within reasonable limits
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.0:89b5f5f494a2: avformat/wavdec: Dont trust the fact chunk for PCM
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.1:de32de679bcc: avformat/wavdec: Dont trust the fact chunk for PCM
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.1:8ffdcd04c481: avformat/matroskadec: only set r_frame_rate if the value is within reasonable limits
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.1:6636dd551fda: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.1:802c4f523186: avformat/mov: force parsing of headers if stts is absent
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.2:6559bb893f5f: avformat/mov: force parsing of headers if stts is absent
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.2:eff0bf7defdf: avformat/matroskadec: only set r_frame_rate if the value is within reasonable limits
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.2:970109deaf47: avformat/wavdec: Dont trust the fact chunk for PCM
[11:59] <cone-897> ffmpeg.git 03mrlika 07release/1.2:311e58e478c1: lavd/v4l2: do not fail when VIDIOC_ENUMSTD returns EINVAL without a valid match
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/1.2:1fe734f4d342: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03mrlika 07release/2.0:f581e25a6905: lavd/v4l2: do not fail when VIDIOC_ENUMSTD returns EINVAL without a valid match
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/2.0:59431fc84187: avcodec/h264_refs: modify key frame detection heuristic to detect more cases
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/2.0:736851264b79: avformat/wavdec: Dont trust the fact chunk for PCM
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/2.0:6de6d9e2d3ed: avformat/matroskadec: only set r_frame_rate if the value is within reasonable limits
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/2.0:7da810e68b76: avformat/gifdec: make GIF_APP_EXT_LABEL parsing more robust
[11:59] <cone-897> ffmpeg.git 03Michael Niedermayer 07release/2.0:fb1fb462e598: avformat/mov: force parsing of headers if stts is absent
[12:08] <saste> BC@;$†‡ä => best username ever
[12:18] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:a826efb55ad8: avfilter/x86/vf_gradfun_init: fix const and related warnings
[12:20] <durandal_1707> michaelni: but i pushed fix to ml
[12:24] <cone-897> ffmpeg.git 03Carl Eugen Hoyos 07master:f60c3a7c1b1e: lavd/x11grab: Support AV_PIX_FMT_PAL8.
[12:24] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:e7811e48be26: Merge remote-tracking branch 'cehoyos/master'
[12:25] <durandal_1707> saste: that appears to be framesync bug
[12:25] <michaelni> durandal_1707, oops didnt see your patch
[12:26] <durandal_1707> you do not read mails at all
[12:27] <durandal_1707> perhaps your todo list is empty?
[12:27] <durandal_1707> ;/
[12:28] <michaelni> i had seen the warnings after the merge and fixed them before looking at the ML
[12:29] <durandal_1707> yes, somehow we started doing same things (not first time), this did not happened before....
[12:34] <durandal_1707> saste: it does not happen if you generate maps once, and reuse them later
[12:35] <durandal_1707> it is also faster
[12:54] <BBB> michaelni: f.. typo in git send-email command
[12:55] <BBB> michaelni: re-sent
[13:07] <durandal_1707> saste: so you have any more comments for displace filter?
[13:11] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:210afae0ba65: avformat/utils: Check size argument in ff_add_index_entry()
[13:19] <durandal_1707> what about following options for mergeplanes: 4 ints which are flags and each bit maps to single plane
[13:19] <durandal_1707> 4 ints for up to 4 inputs
[13:20] <durandal_1707> i hate '1+2|0|1|1' syntax
[13:24] <saste> durandal_1707, did you manage to fix framesync or what?
[13:25] <saste> from my test displace is unusable if you have non synthetic video
[13:25] <durandal_1707> saste: it works if you save geq stuff into file
[13:26] <durandal_1707> anyway the example you use have static maps
[13:26] <durandal_1707> and there is no point in recomputing same image over and over again
[13:28] <saste> durandal_1707, the example I use comes from the manual page
[13:30] <durandal_1707> so what?
[13:31] <saste> durandal_1707, it should work out of the box
[13:31] <saste> if not you should specify the conditions under which it works
[13:32] <saste> ideally we should fix what it needs to be fixed/extended
[13:46] <durandal_1707> saste: so it should be fixed now?
[14:10] <durandal_1707> saste: it is because geq is slow
[14:11] <durandal_1707> and lavfi sucks
[14:11] <durandal_1707> instead of slowing down decoding of first input it just continues...
[14:14] <durandal_1707> same apply if filter output multiple frames at once, lavfi will happily cache all of them and eat memory
[14:15] <wm4> jit'ed geq?
[14:16] <durandal_1707> geq slowness is not problem
[14:39] <durandal_1707> yes, trolling success
[14:40] <ubitux> :')
[15:00] <cone-897> ffmpeg.git 03Ronald S. Bultje 07master:1b3a7e1f42c3: avcodec/x86/videodsp: Properly mark sse2 instructions in emulated_edge_mc x86 simd as such.
[15:20] <ubitux> player ubitux enters the game
[15:22] <Compn> the best ones are the '_______ sucks' ones :)
[15:23] <Compn> nice one ubitux :)
[15:23] <ubitux> oh come on wm4
[15:23] <mateo`> well i guess i can join the game and add more trolling, s/vapoursynth/gstreamer/
[15:23] <ubitux> :D
[15:24] <wm4> hm I guess I should finish my vsynth demuxer and submit it
[16:36] <durandal_1707> saste: if i use movie=matrixbench... i do not get queue overflow
[16:37] <saste> everyone complains that lavfi sucks, and so?
[16:38] <durandal_1707> but video pts starts from 0
[16:38] <durandal_1707> so i wonder why movie filter helps
[16:38] <durandal_1707> perhaps its slowness match geq....
[16:40] <saste> first thing, try to improve ffsynch debugging
[16:41] <saste> durandal_1707, what is happening with regards to pts?
[16:41] <durandal_1707> will see, i can reproduce same issue with overlay filter, so i will troll little about it now....
[16:52] <saste> durandal_1707, ah ok, it is explained in the ticket
[16:52] <saste> i'll rename the ticket...
[16:52] <durandal_1707> what is explained?
[16:53] <durandal_1707> if you rename ticket, thus not accepting truth, i will rm -rf libavfilter/libmpcodecs
[16:56] <saste> bounty??
[16:57] <saste> who added the bounty?
[16:59] <saste> how do you measure the quality of encoded files?
[16:59] <Compn> psnr ?
[16:59] <saste> is something better available? ssim or something?
[17:00] <durandal_1707> drawtext='acceptable'
[17:03] <durandal_1707> vf_compare from dark side have dssim
[17:05] <durandal_1707> also there is blend=all_mode=phoenix
[17:06] <Compn> ssim should be , yes
[17:06] <Compn> but both are somewhat ... not great indicators
[17:09] <saste> durandal_1707, dssim... was it a patch from lu_zero?
[17:09] <saste> why wasn't it never integrated?
[17:16] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:f7106e00f7f1: avformat/rmdec: fix return code of ff_rm_parse_packet()
[17:16] <cone-897> ffmpeg.git 03Michael Niedermayer 07master:aaf02f6ede55: avformat/rmdec: add some error messages
[17:20] <durandal_1707> saste: how do I know?
[17:21] <durandal_1707> saste: you interested in bounty?
[17:22] <saste> durandal_1707, from other developers?
[17:23] <saste> that's a sort of taboo
[17:23] <saste> also from what i know bounties on trac never worked
[17:28] <Compn> how is that a taboo >
[17:33] <cone-897> ffmpeg.git 03Paul B Mahol 07master:ea9632b7470e: avcodec/flicvideo: use avpriv_request_sample()
[17:43] <durandal_1707> michaelni: do not merge libav fieldorder changes as they are incorrect
[19:55] <saste> burek, ping on docu patch
[19:56] <durandal_1707> saste: you just need to specify what bounty you want, best bidder wins
[19:58] <llogan> michaelni: can you grep for rob.schmidt86 in the mail logs? I would but I'm not a sudoer.
[19:59] <llogan> (and I don't know my password...just the ssh key)
[20:04] <durandal_1707> llogan: password for what?
[20:05] <llogan> avserver. i wanted to see if i was in sudoers, but can't test without a passwd
[20:05] <llogan> someone is having trouble subscribing to libav-user
[20:05] <llogan> exciting. carry on.
[20:06] <durandal_1707> avserver?
[20:06] <llogan> ffmpeg.org
[20:07] <durandal_1707> so you lost you password, and now ask for it via irc?
[20:07] <llogan> not at all.
[20:07] <llogan> i was only issed a login via ssh key and i'm not asking for the passwd via irc.
[20:09] <durandal_1707> so how we should know you are really llogan ?
[20:11] <llogan> i've just added the mod bit to your address in -devel
[20:12] <durandal_1707> what?
[20:12] <llogan> you wanted proof.
[20:12] <llogan> ok...i removed it.
[20:13] <durandal_1707> i didn't notice change
[20:21] <michaelni> mraulet, there are some hevc bugs on trac (https://trac.ffmpeg.org/query?status=new&status=open&status=reopened&keywords=~hevc)
[20:23] <mraulet> ok
[20:25] <mraulet> could you cc me hevc related bugs?
[20:39] <michaelni> not sure i understand but you can add yourself to the CC list of any bug or you could subscribe to https://lists.ffmpeg.org/mailman/listinfo/ffmpeg-trac/ to get all bug changes (which is alot), or i could try to notify you when i see new hevc bugs if thats what you meant but i tend to forget doing such things, carl tends to be better at such things
[20:43] <mraulet> I don't want to get all of them
[20:43] <mraulet> :-)
[20:47] <michaelni> saste, did you see https://trac.ffmpeg.org/ticket/3054 ? (doc uninstall regression since "build: drop unused ALLMANPAGES variable")
[20:54] <saste> michaelni, i'll fix it
[20:54] <michaelni> saste, thx
[20:56] <durandal_1707> michaelni: i do not abuse trac, i want to make critical issues as such
[21:02] <michaelni> see doc/issue_tracker.txt for a description of each priority
[21:03] <durandal_1707> heh, than i will fork
[21:06] <saste> durandal_1707, i found the perfect name for it
[21:06] <saste> yaff
[21:07] <saste> yet another ffmpeg fork
[21:10] <llogan> i prefer ffffffmpeg
[21:11] <saste> https://trac.ffmpeg.org/tags => cool
[21:21] <cone-698> ffmpeg.git 03burek 07master:229042a52d5d: doc/protocols: add description for the RTP protocol
[21:21] <cone-698> ffmpeg.git 03Stefano Sabatini 07master:af1c5388501e: doc/Makefile: fix man pages uninstall path
[21:21] <cone-698> ffmpeg.git 03Ingo Brückl 07master:6c9c636c0022: lavd/fbdev_common: fix compilation with older kernels
[21:29] <saste> ^ trac fight
[21:31] <michaelni> mateo`, can you take a quick look at a 1 line mxf patch "1024 21:28 To FFmpeg devel (0.8K) [FFmpeg-devel] [PATCH] avformat/mxfdec: dont truncate packets" ?
[21:31] <durandal_1707> saste: more like fatchick fight
[21:58] <durandal_1707> query_formats is called before init?
[22:57] <wm4> I'm seeing a ffmpeg muxed file that has no audio frame marked as keyframe
[22:57] <wm4> err, it's a matroska file
[22:59] <wm4> are there even audio codecs where a a matroska "block" isn't on key frame boundaries?
[23:05] <Compn> audio has keyframes ?
[23:06] <wm4> in matroska it does
[23:06] <JEEB> that marker in matroska is a misnomer
[23:06] <JEEB> it means "random access point"
[23:07] <wm4> yeah
[23:07] <wm4> anyway, mplayer2/mpv's demuxer can't handle it if no audio block is marked as keyframe
[23:07] <Daemon404> the bigger question is
[23:07] <wm4> so I'm pretty sure almost any mkv file I looked at had these more or less correctly marked
[23:07] <Daemon404> who the hell muxes mkv with ffmpeg?
[23:07] <Daemon404> and why?
[23:07] <wm4> the file I'm looking at is named gst-hackfest-2013-360p.webm
[23:08] <wm4> | + Muxing application: Lavf54.29.104
[23:08] <wm4> | + Writing application: Lavf54.29.104
[23:08] <Daemon404> gstreamer. got it.
[23:08] <Daemon404> AIDS.
[23:08] <wm4> :)
[23:21] <ubitux> BBB: 4x4 should be done by the end of the week end
[23:21] <ubitux> i'm almost there
[00:00] --- Fri Oct 25 2013