[Ffmpeg-devel-irc] ffmpeg-devel.log.20120524

[burek]

[00:36] <saste> ubitux: assuming i did the get_video_buffer once in the config_props
[00:37] <saste> that would be weird, altough maybe possible at that point (assuming that configuration is backward, from output filters to input filters)
[00:37] <saste> maybe the best thing is just to request the buffer when the buffer is first required
[00:40] <ubitux> since i would unref in uninit, it makes sense to me to instanciate it in init
[00:40] <ubitux> but well, i'll see if that's possible or not
[00:51] <floater_afb60> ffmpeg -f v4l2 -r 1/12 -i "/dev/video0" -f png - | convert - -font Arial -pointsize 10 -draw "gravity SouthEast fill black  text 0,12 ""$(date +"'""%Y-%m-%dT%H:%M:%S%z""' fill white  text 1,11 '""%Y-%m-%dT%H:%M:%S%z""'" )" - | ffmpeg -f image2 -r 30 -i - -vcodec libx264 -preset placebo -vf hqdn3d "./pipelapse.mkv"
[00:51] <floater_afb60> ffmpeg doesn't like me :(
[00:52] <floater_afb60> wanted to stream png files into pipe, timelapse on-the-fly then convert to h264 video stream
[00:52] <floater_afb60> looks like i need a loop ;<
[01:02] <CIA-105> ffmpeg: 03Stefano Sabatini 07master * rf2ee065638 10ffmpeg/ (4 files in 3 dirs): 
[01:02] <CIA-105> ffmpeg: lavu/pixfmt: move AVPALETTE_SIZE and _COUNT to pixfmt.h
[01:02] <CIA-105> ffmpeg: Move AVPALETTE_SIZE and AVPALETTE_COUNT definition from
[01:02] <CIA-105> ffmpeg: libavcodec/avcodec.h to libavutil/pixfmt.h.
[01:02] <CIA-105> ffmpeg: The definition is more useful in libavutil, where it can be shared for
[01:02] <CIA-105> ffmpeg: example by libavfilter and libswscale.
[01:02] <CIA-105> ffmpeg: 03Stefano Sabatini 07master * rfce30d8d73 10ffmpeg/libavfilter/video.c: 
[01:02] <CIA-105> ffmpeg: lavfi: copy palette in start_frame()
[01:02] <CIA-105> ffmpeg: Fix -vf copy with pal8 format.
[01:26] <ubitux> oh it starts to work :)
[01:29] <ubitux> arg, time to sleep.
[01:29] <ubitux> (http://ubitux.fr/pub/pics/_showspectrum-WIP.png \o/)
[01:29] <ubitux> night ppl.
[02:03] <Daemon404> floater_afb60, please do not pm me...
[02:27] <floater_afb60> Daemon404<< it was reply to your question :P
[02:28] <Zeranoe> michaelni: I came up with a patch for the dshow header issue. Is a rename to dshow_capture.h an acceptable name? If so, the issue is resolved with the rename.
[02:28] <Daemon404> floater_afb60, which one
[02:29] <floater_afb60> ... [163212] <Daemon404> ffmpeg with timestamps* ...
[02:30] <floater_afb60> since ffmpeg does not have this feature yet, i used imagick
[02:30] <Daemon404> i think you misunderstand what i mean by timestamps
[02:31] <Daemon404> im referring to a way to set pts/dts
[02:53] <michaelni> Zeranoe, any name you like is fine
[02:57] <Zeranoe> michaelni: 
[02:58] <Zeranoe> michaelni: Can I pastbin the patch to you or would you prefer I submit it to my mailing list?
[03:18] <Zeranoe> michaelni: Here is a pastebin: http://pastebin.com/N9hyiMiH I need to leave. If you need it submitted to the mailing list let me know. I apologize if it isn't formated properly.
[04:50] <floater_afb60>  i am currently wanting to test average RGB color vector distance with threshold level to pre trigger recording 0.5s, what is your opinion?
[05:10] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r77ba8af023 10ffmpeg/libavfilter/asink_anullsink.c: (log message trimmed)
[05:10] <CIA-105> ffmpeg: asink_anullsink: fix copyright
[05:10] <CIA-105> ffmpeg: git blame says authors are:
[05:10] <CIA-105> ffmpeg:  19 S.N. Hemanth Meenakshisundaram
[05:10] <CIA-105> ffmpeg:  2 Reimar Döffinger
[05:10] <CIA-105> ffmpeg: initial commit:
[05:10] <CIA-105> ffmpeg:  commit f0a5543898dad040b8f2d1b44fb6b278c06c3460
[05:10] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r7ef5c80cb4 10ffmpeg/libavfilter/af_anull.c: (log message trimmed)
[05:10] <CIA-105> ffmpeg: af_null: fix copyright
[05:10] <CIA-105> ffmpeg: Authors according to blame are:
[05:10] <CIA-105> ffmpeg:  16 S.N. Hemanth Meenakshisundaram
[05:10] <CIA-105> ffmpeg:  4 Stefano Sabatini
[05:10] <CIA-105> ffmpeg:  3 Anton Khirnov
[05:10] <CIA-105> ffmpeg:  2 Reimar Döffinger
[05:10] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r83995411f7 10ffmpeg/libavfilter/avcodec.c: (log message trimmed)
[05:10] <CIA-105> ffmpeg: lavfi/avcodec.c: fix copyright
[05:10] <CIA-105> ffmpeg: git blame:
[05:10] <CIA-105> ffmpeg:  45 Stefano Sabatini
[05:10] <CIA-105> ffmpeg:  23 Clément BSsch
[05:10] <CIA-105> ffmpeg:  4 Michael Niedermayer
[05:10] <CIA-105> ffmpeg:  3 Robert Nagy
[08:19] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * r001d9d5e93 10ffmpeg/libavcodec/x86/h264_qpel_mmx.c: Fix compilation with --disable-everything.
[08:19] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * r36d0f8c9db 10ffmpeg/configure: Do not disable ffmpeg with --disable-everything.
[10:02] <j-b> mo-ro-ning!
[10:06] <ubitux> that happiness looks suspicious
[11:37] <ubitux> i know i already ask several times but does anyone know the "PPM of the audio"?
[11:37] <ubitux> aka "text" audio format
[11:38] <ubitux> heavy, but easy to generate and read
[11:38] <floater_afb60> BPM ?
[11:39] <ubitux> i don't find anything official about such format
[11:39] <ubitux> any URL?
[12:27] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * rc0532f4bbc 10ffmpeg/configure: Build the aformat filter when ffmpeg is built.
[13:03] <ubitux> floater_afb60: no, nothing? :)
[13:12] <floater_afb60> ubitux<< PPM is Part Per Million to me ;/
[13:13] <ubitux> ah, does netpbm sound better to you? :)
[13:13] <ubitux> https://en.wikipedia.org/wiki/Netpbm_format
[13:13] <floater_afb60> oh thats another thing
[13:13] <ubitux> ppm, portable pixmap fmt
[13:14] <floater_afb60> why not ppf then?
[13:14] <ubitux> anyway, i'm looking for some kind of "standard" text-based amplitudes audio format
[13:14] <ubitux> floater_afb60: "format" is not part of the name
[13:14] <ubitux> Portable PixMap
[13:15] <floater_afb60> then PP is http://en.wikipedia.org/wiki/Polypropylene
[13:15] <floater_afb60> i don't get it what you asking now
[13:16] <floater_afb60> google tells me midi 
[13:16] <floater_afb60> i used XM in 2001
[13:17] <ubitux> it's binary formats, and you need synthesis for them
[13:17] <floater_afb60> FastTracker
[13:17] <ubitux> i'm just looking for a format with amplitudes stored as text
[13:18] <ubitux> with some kind of standard text header to save the sample rate for instance
[13:18] <ubitux> i can invent one myself of course, but i was wondering if there wasn't any already
[13:19] <floater_afb60> ubitux<< ftp://ftp.heanet.ie/disk1/sourceforge/u/project/uf/ufmod/XM%20file%20format%20specification/FastTracker%20II,%20ADPCM%20XM%20and%20Stripped%20XM/XM_file_format.pdf.gz
[13:19] <floater_afb60> here you are, good?
[13:19] <ubitux> xm is the same
[13:19] <ubitux> it's a tracker format
[13:19] <Cervajz> Hello guys :)
[13:19] <ubitux> binary, and requires complex code to decode and do some synthesis
[13:20] <Cervajz> What is the proper PID for the private streams in MPEG-TS please?
[13:23] <Compn> Cervajz : maybe its listed on http://wiki.multimedia.cx
[13:24] <Compn> or in the mpeg specs
[13:26] <floater_afb60> ubitux<< yeah and 600KB gives a 3 minute standard soundtrack
[13:27] <floater_afb60> or 50 kB for an hour chimes melody
[13:27] <floater_afb60> ubitux<< what is your problem with binary? store it in base64 then :P
[13:28] <floater_afb60> binary decoding is no problem in a decent programming language i guess... unless you are creating your "codec" in bash
[13:29] <ubitux> i'm just looking for a *really* easy to generate and parse formats, and human readable
[13:30] <ubitux> parser: srate = fgets(line); while (fgets(line...)) samples[i++] = strtol(line,NULL,10);
[13:30] <ubitux> encoder: printf("%d\n", srate); for (i = 0; i < nb_samples; i++) printf("%d\n", samples[i]);
[13:31] <ubitux> see the point? :)
[13:31] <floater_afb60> yes i see, you want to generate a tone using bash script
[13:31] <floater_afb60> i did things like that too, but it does not have a real purpose
[13:31] <ubitux> it does have a purpose for testing
[13:31] <floater_afb60> ofc i see
[13:32] <ubitux> or experimenting with audio
[13:32] <ubitux> and you might not want to have the format/codec complexity
[13:33] <ubitux> this kind of format exist for images, and it's netpbm
[13:33] <ubitux> i just want to audio equivalent, if any :)
[13:33] <floater_afb60> http://pastebin.com/sshScG5a
[13:34] <floater_afb60> tuned for 1024x768 framebuffer now
[13:34] <floater_afb60> start it with rootshell
[13:38] <floater_afb60> ubitux<< clear;e=eval\ ;p=printf\ ;r=read\ ;x="$e$p\$";u=USER;n=NAME;b=$(stty -g);stty raw -echo;$r-sn1 -p \[$($x$u)@$(${x}HOST$n):~\]$\ ;while $r-n1 a;do $p\\$($p\%03o $(($($p\%d "'"$($p\%c "$a")"")-1)));sleep .$(($RANDOM*2));done< <($p\%q 'Uif!Nbusjy!ibt!zpv!'"$($p\%s "$(grep $($x$u$n) /etc/passwd|awk -F: '{print $5}')"|while $r-n1 y;do $p\%c "$($p\\$($p\%03o $(($($p\%d "'$($p\%c "$y")")+1))))";done;)"'///');stty "$b";$p\\n
[13:39] <ubitux> thx man, all of this is a really great help
[13:55] <floater_afb60> ubitux<< anyway look into wav format
[13:55] <floater_afb60> it will be larger if you store it in text
[13:56] <floater_afb60> given a sample rate, you record the amplitude of the waveform
[13:56] <floater_afb60> 44100 times per second for each channel, and you're done
[13:57] <floater_afb60> that would be the equivalent of your PiXelMap image
[13:57] <floater_afb60> i like the bios font thing btw
[13:58] <floater_afb60> use it for simple text, it is binary encoded, 8 bytes / 64 pixel font.
[13:58] <floater_afb60> and a simple loop can output it
[14:01] <ubitux> yes i know the wav format is the simpler and widely used audio format
[14:01] <ubitux> but still, i can't read it with a simple pager to check the value, or pipe it directly to gnuplot etc
[14:02] <floater_afb60> well you could...
[14:03] <floater_afb60> hexdump -C file | less
[14:06] <ubitux> just like we could hexdump -C image.bmp | less, but the ppm still is a nice format, and is btw used often
[14:07] <floater_afb60> i prefer vectorgraphics
[14:08] <floater_afb60> takes a few bytes to describe a primitive in vectorspace
[14:09] <floater_afb60> and computers happen to have enough processing power to render millions of them every second
[14:09] <ubitux> just like tracker files for audio :p
[14:10] <floater_afb60> yea
[14:10] <floater_afb60> something like that, with instruments and channels
[14:17] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r2f6c5894f4 10ffmpeg/tests/ref/lavfi/pixfmts_copy: 
[14:17] <CIA-105> ffmpeg: fate: update checksum after pal8 changes
[14:17] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:17] <CIA-105> ffmpeg: 03Kyle 07master * r02cc66dd78 10ffmpeg/libavdevice/ (8 files): 
[14:17] <CIA-105> ffmpeg: dshow: rename dshow.h to avoid conflict with system header of equal name
[14:17] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:51] <ubitux> saste: i think i'm going to add the scene detection filter to my todo list, if you don't plan to do it soon :)
[14:52] <ubitux> saste: oh btw, i noticed something strange the other day; drawtext strftime was using my local time :p
[14:52] <ubitux> did you ever expect this?
[17:09] <ubitux> i think i found a problem with the audio filtering
[17:09] <ubitux> from what i can tell, the default audio filtering path appends aformat to the output
[17:09] <ubitux> and aresample/pan/vol to the inputs
[17:10] <ubitux> but with the complex filters
[17:10] <ubitux> we have aformat and async both for outputs
[17:11] <ubitux> well async doesn't work the same as aresample for this
[17:11] Action: ubitux is confused
[18:47] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * rafe4d3bd54 10ffmpeg/ffplay.c: 
[18:47] <CIA-105> ffmpeg: ffplay: set base in dr1 allocation.
[18:47] <CIA-105> ffmpeg: This matches the default non dr1 allocator
[18:47] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:56] <ubitux> anyone to calm down ppl on ffmpeg-user?
[19:56] <ubitux> Compn: are you a moderator?
[20:03] <RobertNagy> what exactly does async do? I never fully understood it
[20:10] <Compn> ubitux : yes
[20:10] <Compn> ubitux : mostly we just let them flame out :P
[20:19] <ubitux> ok :)
[20:20] <Compn> ubitux : if you can, ask the guy if hes on a windows box
[20:20] <Compn> i've seen java exploits that allow worms to send mail thru open webmail accounts
[20:21] <Compn> a few people i know got hit with it 
[20:21] <Compn> steps to reproduce : open yahoo mail, put it background, open other window and browse internet....
[20:21] <Compn> blam. spammed mail
[20:22] <ubitux> too late i delete these mails
[20:40] <Daemon404> michaelni, is it normal for lavf/nut to completely ignore my pkt.dts and pkt.pts?
[20:53] <michaelni> RobertNagy, it syncs audio to some timestamps, that is when sample counts dont match up with the timestamps
[20:53] <michaelni> Daemon404, nut should not ignore pts
[20:54] <Daemon404> im testing like:
[20:54] <Daemon404>         pkt.pts           = curframeinfo->PTS * 2;
[20:54] <Daemon404>         pkt.dts           = curframeinfo->PTS * 2;
[20:54] <Daemon404>         ret = av_interleaved_write_frame(out, &pkt);
[20:55] <Daemon404> yet it always is detected and playsback at the same speed
[20:55] <RobertNagy> michaelni: I still don't quite understand when one would use it?
[20:55] <RobertNagy> or rather, why would you not always use it?
[21:00] <michaelni> timestamps could be noisy and async might make things worse as a result
[21:01] <RobertNagy> ic
[21:01] <michaelni> it also needs a resampler to be run which inevitably reduces quality
[21:01] <michaelni> and causes a tiny slowdown
[21:11] <Daemon404> michaelni, im trying to pipe a vfr file with timestamps via nut
[21:11] <Daemon404> and im finding ffmpeg's timestamp/framrate handlign is COMPLETELY INSANE
[21:14] <RobertNagy> "completely insane" is not very descriptive...
[21:15] <Daemon404> it's a giant pile of hacks ind isnaneness
[21:15] <Daemon404> there are too many things wrogn with it to explain at once
[21:15] <RobertNagy> still not very descriptive
[21:15] <Daemon404> [15:15] <@Daemon404> there are too many things wrogn with it to explain at once
[21:15] <Daemon404> they way i completely ignores pts/dts, and generates its own sometimes based on timebase
[21:16] <Daemon404> it*
[21:16] <Daemon404> is one
[21:17] <Daemon404> ive got the pts and dts set correctly for every frame, but it chooses to ignore them when it gets the resulting file as input
[21:17] <Daemon404> ffprobe prints the proper pts/dts for each packet, but ffmpeg and ffplay could not care less
[21:17] <michaelni> Daemon404, can you upload the file ?
[21:18] <Daemon404> yes
[21:18] <michaelni> iam happy to fix it, but iam not sure i have time now
[21:18] <Daemon404> in a sec
[21:18] <Daemon404> i think i see an issue tho
[21:18] <Daemon404> on my end
[21:19] <RobertNagy> do you even need to set the dts e´when encoding?
[21:19] <RobertNagy> doesn't ffmpeg set the dts?
[21:19] <Daemon404> im piping vfr content
[21:20] <RobertNagy> why not just convert it into cfr? vfr is insane
[21:20] <Daemon404> because ffmpeg is whats supposed to convert it to vfr
[21:20] <Daemon404> and vfr isnt insane if you dont write crap code to handle it
[21:21] <Daemon404> it's just a series of timestamps.
[21:24] <Daemon404> [15:23] <Daemon404>     stream->codec->time_base.den = tb->Den;
[21:24] <Daemon404> [15:23] <Daemon404>     stream->codec->time_base.num = tb->Num;
[21:24] <Daemon404> [15:23] <Daemon404> ^ there are 25000 and 1000 respectively
[21:24] <Daemon404> [15:23] <Daemon404> however
[21:24] <Daemon404> [15:24] <Daemon404>             "time_base": "1/1",
[21:24] <Daemon404> ^ on the resulting file
[21:24] <michaelni> btw you set dts=pts, is this correct for the codec you have choosen ?
[21:25] <Daemon404> it's rawvideo in nut
[21:25] <Daemon404> pts==dts
[21:25] <michaelni> ok
[21:25] <Daemon404> timebase is getting mangled somehow
[21:26] Action: Daemon404 investigates
[21:26] Action: Daemon404 is probably using teh confusing lavf api wrong
[21:27] Action: Daemon404 goes to take a break
[21:28] <michaelni> also once we figure out what was the issue we should make sure we fix docs and code to avoid this from happening to someone else again
[21:31] <j-b> good morning
[21:31] <Compn> how goes it j-b ?
[21:31] Action: Compn likes the 'somethings wrong with ffmpeg' bug reports until mr bug report comes back with 'oh i'm using some custom code and dont know what i'm doing' instead of a proper bug report in the first place
[21:31] Action: Compn looks at Daemon404
[21:32] <j-b> Compn: pretty good, I think. A lot of bugs to fix, but good press coverage and users....
[22:19] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r4e84ab886b 10ffmpeg/MAINTAINERS: 
[22:19] <CIA-105> ffmpeg: MAINTAINERS: update versions
[22:19] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:19] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r2416cd4035 10ffmpeg/Changelog: 
[22:19] <CIA-105> ffmpeg: Changelog: update for 0.11
[22:20] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:20] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r53285ced19 10ffmpeg/doc/ffmpeg-mt-authorship.txt: (log message trimmed)
[22:20] <CIA-105> ffmpeg: doc: remove ffmpeg-mt-authorship.txt from master (for upcoming 0.11)
[22:20] <CIA-105> ffmpeg: This lists git history, which is better kept in the history itself
[22:20] <CIA-105> ffmpeg: that is both ours as this file as well as the actual history of the
[22:20] <CIA-105> ffmpeg: ffmpeg-mt project.
[22:20] <CIA-105> ffmpeg: If someone thinks this is not ok, drop me a mail and ill put it back!
[22:20] <CIA-105> ffmpeg: Also note this file was not carried in our previous release with noone
[22:20] <j-b> michaelni: are you releasing soon?
[22:23] <michaelni> j-b, yes iam trying
[22:23] <michaelni> i wanted to do it today but i think it will be tomorrow
[22:24] <j-b> michaelni: I have 3 security issues reported by MS, where do you want them?
[22:25] <michaelni> email
[22:30] <michaelni> correction, j-b, email unless its too big ...
[22:31] <j-b> michaelni: the first one is in your mail
[22:32] <j-b> michaelni: I hope you have not already fixed those, but it might be possible that you did
[22:32] <Compn> microsoft ?
[22:32] <j-b> michaelni: Microsoft is a bit pushy about those
[22:32] <j-b> Compn: yes.
[22:33] <Compn> one day microsoft will flip a switch and bam! only approved (signed + crypto + md5) applications are allowed :D
[22:34] <Compn> wonder how much microsoft fanfiction there is
[22:34] <j-b> veryy possible
[22:34] <j-b> this is already happening
[22:34] <iive> Compn: at win9 :)
[22:35] <j-b> winRT in win8 has flash that only works on a whitelist of websites
[22:46] <michaelni> j-b, what do i need to do with the files to cause memory corruption ?
[22:46] <michaelni> simply decoding to -f null - doesnt show anything in valgrind
[22:47] <michaelni> i get Invalid number of channels (7)
[22:47] <michaelni> Invalid sample format (null).
[22:47] <michaelni> Error opening filters!
[22:48] <michaelni> ffplay produces:Invalid number of channels (7)
[22:48] <michaelni>  could not open codecs
[22:49] <j-b> ok, so, maybe you fixed it already
[22:49] <j-b> it does crash here.
[22:51] <j-b> 0x00007fffd56064d9 in ff_wma_run_level_decode () from /home/jb/VideoLAN/dev/vlc/build/modules/codec/avcodec/.libs/libavcodec_plugin.so
[22:51] <j-b> http://pastie.org/3962958
[22:52] <michaelni> i tried with 0.10.2 and 0.8.11 neither does anything interresting
[22:53] <j-b> crap
[22:55] <iive> j-b: are you using libav?
[22:56] <j-b> no.
[22:57] <j-b> avformat demux, avcodec codec module and crash in wma_run_level... What can it be...
[22:59] <michaelni> i got a crash when ignoring avformat_find_stream_info() failure but dunno if thats the same
[23:00] <funman> j-b: Y U NO --enable-debug for ffmpeg ?
[23:00] <michaelni> my crash is just a null ptr dereference
[23:01] <j-b> funman: cause this is my distro
[23:01] <j-b> funman: and else, because this is the released VLC...
[23:01] <j-b> funman: else, you can try, there are 3 security issues pending, all in libavcodec module
[23:04] <funman> j-b: ok i can try
[23:08] <funman> j-b: samplees?
[23:08] <j-b> in your mail soon
[23:09] <j-b> funman: you have the 2 first ones
[23:10] <funman> [wmav2 @ 0x7f62a0c1ce60] Invalid number of channels (7)
[23:10] <funman> original plays and repro is rejected
[23:11] <funman> i need a password for the second mail
[23:12] <funman> btw this is with 2416cd403
[23:12] <j-b> funman: vlc 2.1-git?
[23:12] <funman> 1.3.0-git-2607-g669ce09
[23:13] <j-b> crap
[23:13] <funman> why? seems fixed
[23:14] <j-b> because it crashes here with 1.3.0-git-2637
[23:14] <funman> 17:01 <@j-b> funman: cause this is my distro
[23:14] <funman> not ffmpeg HEAD?
[23:15] <j-b> probably not
[23:16] <funman> repro are fuzzed files?
[23:16] <funman> repro.qt crashes in vlc
[23:17] <michaelni> vlc+ffmpeg HEAD ? crash in ffmpeg ?
[23:18] <funman> nope
[23:18] <funman> crash in vlc
[23:18] <j-b> funman: ok
[23:18] <j-b> funman: the last one is in your mail
[23:18] <funman> 1391	            vout_EndWrapper(vout);
[23:18] <funman> (gdb) print vout
[23:18] <funman> Unhandled dwarf expression opcode 0x0
[23:19] <funman> nice one
[23:19] <funman> and my expressions are vertically-challenged. fuck you
[23:19] <j-b> fuck you too :)
[23:21] <ubitux> michaelni: are you sure you want to release 0.11 with the filtering in that state?
[23:21] <michaelni> ubitux, are there regressions ?
[23:21] <funman> oh shit i had built it with clang
[23:22] <funman> (vlc)
[23:22] <michaelni> ubitux, i know cplx+audio doesnt work but thats a new feature it wasnt in the previous release
[23:23] <ubitux> well i'm sure the formats negociation is likely broken
[23:24] <michaelni> ubitux, do you have a testcase that got worse ?
[23:24] <ubitux> lemme check
[23:26] <ubitux> mmh it seems the issues i had has been fixed
[23:26] <michaelni> see, we are in good shape ;)
[23:27] <ubitux> haha maybe, i'm going to check a few things still, just a minute :)
[23:27] <michaelni> no hurry, there wont be a release today
[23:27] <michaelni> plan is tomorrow
[23:28] <sware> is there any way that it could become standard to list all members of a struct rather than using .member = ?
[23:29] <ubitux> michaelni: and you don't want to at least fix the crashes with the complex filters?
[23:31] <michaelni> i want, but i thought you or nicolas wanted to work on it
[23:31] <michaelni> ?
[23:32] <ubitux> by "want to fix" i could mean "wait for a fix"
[23:32] <michaelni> well, theres always something to wait for
[23:33] <michaelni> always a fix or a new features
[23:33] <michaelni> also its just audio or are complex video filters buggy too ?
[23:34] <ubitux> no idea i didn't try
[23:34] <michaelni> and backporting that fix to a 0.11.1 is not hard 
[23:34] <sware> I want to maintain an msvc build, but it's unrealistic to constantly convert structs
[23:34] <ubitux> there might be some issues with the overlay filter which nicolas is rewriting
[23:35] <ubitux> sware: no we don't want to change that
[23:35] <ubitux> we heavily rely on this feature all over the code, it would be a great "regression" to move back to the old c89 way for this
[23:35] <michaelni> sware, you want to write some script that converts standard C to msvc ?
[23:35] <sware> I was thinking of doing it 
[23:36] <michaelni> for you personal use or some public ?
[23:36] <j-b> that would be nice
[23:36] <sware> I spoke with you a while ago
[23:36] <j-b> MSVC or MSVC c++ mode?
[23:36] <sware> under another nick, about porting to msvc, I just got busy with work, I have some time now 
[23:36] <sware> so I am building may 21s snapshot
[23:36] <sware> msvc building as c code 
[23:37] <michaelni> nice
[23:37] <sware> right now i have ARCH_X86 and ARCH_X86_32 turned off because I need to convert between intel and at&t syntax for the inline asm
[23:38] <sware> I have libavutil and libswscale building and I'm half way through libavformat 
[23:38] <sware> the bulk of the work is the structs ;/
[23:40] <michaelni> you should automate the structs convertion somehow ...
[23:40] <sware> ya question is how 
[23:41] <michaelni> there are not so many different structs
[23:41] <sware> there's a bit more to it, because I have to fix structs with unions, and write little funcs for things like rationals and softfloats
[23:42] <michaelni> you could probably hardcode alot in a convert script
[23:42] <michaelni> like replace name by [0] or such
[23:42] <michaelni> i mean no need to parse the struct from the header
[23:42] <sware> it would have to retrieve the struct, and write each member out, then fill in the details with the existing stuff, all wrapped in #ifndef 
[23:44] <sware> well I figure though, if I do a more recent port, and then it will be easy to implement changes from that, so all that does't have to be done again
[23:44] <michaelni> why do you need ifdef ?
[23:44] <ubitux> michaelni: do you mind if i had a #include "avfilter.h" in lavfi/video.h?
[23:44] <sware> a few reasons, I like to see changes in code 
[23:44] <ubitux> (make check is broken)
[23:45] <sware> at least while this is somewhat beta
[23:45] <michaelni> ubitux, include as much public from private as you wish 
[23:46] <sware> I use #ifndef _MSC_VER around the things I change  
[23:46] <cptspiff> weirdos like me use git diff to see changes ;)
[23:47] <sware> they can all be removed at a later point 
[23:49] <CIA-105> ffmpeg: 03Clément BSsch 07master * r781fb46c56 10ffmpeg/libavfilter/video.h: lavfi/video: add avfilter.h include to fix make checkheaders.
[23:49] <sware> the only real issue is the inline asm 
[23:49] <CIA-105> ffmpeg: 03Clément BSsch 07master * rd8e075b1a0 10ffmpeg/ffmpeg.c: ffmpeg: use AUTO_INSERT_FILTER() for -async option.
[23:49] <sware> converting syntax
[23:49] <sware> I need to find a utility to make that faster 
[23:50] <michaelni> forget inline asm
[23:50] <michaelni> its too much work
[23:50] <michaelni> you have yasm based asm anyway
[23:50] <sware> isn't there a speed increase ? 
[23:51] <michaelni> sure there is
[23:51] <sware> I used intrinsics to get libavutil building with an x86 timer.h 
[23:51] <michaelni> but some of this code is tricky and uses gccs inline asm at the borders of gccs abilities
[23:51] <sware> since there's only one line now, I just used __rdtsc(); 
[23:51] <michaelni> i have the fear that the borders of gccs abilities are not a strict subset of msvcs
[23:52] <michaelni> if iam guessing right msvc wont be able to build some of the inline asm after its converted
[23:52] <sware> so ignoring the asm is alright
[23:52] <sware> that saves me a lot of work :)
[23:54] <michaelni> porting the asm to yasm probably makes more sense as in that form some of it could be put in main ffmpeg
[23:56] <sware> are you interested in it when I'm finished ?
[23:57] <sware> I could send the project over 
[23:57] <sware> if someone else wants to see and maybe improve or use it 
[23:59] <sware> I would maintain it if nobody else wanted to
[23:59] <CIA-105> ffmpeg: 03Clément BSsch 07master * r42b2908d7b 10ffmpeg/doc/filters.texi: 
[23:59] <CIA-105> ffmpeg: doc/filters: remove old aformat entry.
[23:59] <CIA-105> ffmpeg: The new description is still available below.
[23:59] <CIA-105> ffmpeg: 03Clément BSsch 07master * rba4aa656ce 10ffmpeg/doc/filters.texi: doc/filters: move "aformat" before "amerge".
[00:00] --- Fri May 25 2012