[Ffmpeg-devel-irc] ffmpeg-devel.log.20120523

[burek]

[02:26] <burek> ok, I guess it works now :)
[10:28] <floater_afb60> hi
[10:28] <floater_afb60> what is this ff_fmt codec_id v4l2_fmt conversion table thingy?
[10:29] <av500> a conversion table between ffmpeg codec IDs and v4l2 formats?
[10:29] <floater_afb60> ok i tell you the coolness
[10:29] <floater_afb60> i would like to make ffmpeg support PJPG
[10:29] <floater_afb60> added a line
[10:30] <floater_afb60> copied last line "JPEG" and modded it to "PJPG"
[10:30] <floater_afb60> since v4l2 has that format.
[10:30] <floater_afb60> wondering if it will work
[10:31] <floater_afb60>     { PIX_FMT_NONE,    CODEC_ID_MJPEG,    V4L2_PIX_FMT_PJPG    },
[10:31] <floater_afb60> PIX_FMT_NONE does not sound very useful
[10:35] <floater_afb60> http://pastebin.com/CsuQd5S7
[10:35] <floater_afb60> cheap webcam from tesco corp PJPG pixel format
[10:36] <av500> adding that line looks OK
[10:38] <floater_afb60> [video4linux2,v4l2 @ 0x2e5b300] Cannot find a proper format for codec_id 0, pix_fmt -1.
[10:38] <floater_afb60> /dev/video1: Input/output error
[10:38] <floater_afb60> unfortunately
[10:40] <floater_afb60> previously it was  codec_id 8
[10:40] <floater_afb60> ;/
[10:41] <floater_afb60> seriously, somebody knows what is ffmpeg doing?
[10:43] <floater_afb60> http://pastebin.com/LWYSj1Av
[11:24] <floater_afb60> currently hacking debug fprintfs into ffmpeg src to be cool
[11:26] <av500> totally cool
[11:33] <floater_afb60> /usr/bin/ffmpeg: error while loading shared libraries: libavdevice.so.53: cannot open shared object file: No such file or directory
[11:34] <floater_afb60> $ whereis libavdevice.so.53
[11:34] <floater_afb60> libavdevice.so: /usr/lib/libavdevice.so.53 /usr/lib/libavdevice.so /usr/lib64/libavdevice.so.52
[11:34] <floater_afb60> ain't that awesome?
[11:34] <floater_afb60> nah ldconfig helped
[11:35] <floater_afb60> ahaha
[11:36] <floater_afb60> this is insane
[11:39] <floater_afb60> http://pastebin.com/KKL3mEA3
[11:39] <floater_afb60> at last, ffmpeg is doing - something - 
[11:41] <floater_afb60> the 480x640 Planar YV12 is a killer
[11:42] <burek> floater_afb60, did you create a new ticket on ffmpeg trac for PJPG as suggested
[11:42] <burek> ?
[11:42] <ubitux> did you check with libv4l?
[11:42] <ubitux> just in case it has a wrapper for this
[11:43] <ubitux> (--enable-libv4l2)
[11:43] <floater_afb60> the other webcam works
[11:43] <ubitux> (does it work with vlc?)
[11:44] <floater_afb60> the video player? i don't have it
[11:44] <ubitux> ok
[11:44] <floater_afb60> mplayer fails though
[11:44] <ubitux> yes but mplayer doesn't use libv4l2
[11:44] <ubitux> that's why i ask :p
[11:44] <floater_afb60> but cheese plays it ok
[11:44] <ubitux> ah?
[11:44] <ubitux> they give it a try.
[11:44] <ubitux> (rebuild ffmpeg with --enable-libv4l2)
[11:44] <floater_afb60> isn't the default enabled?
[11:45] <floater_afb60> the other webcam works withffmpeg.
[11:45] <burek> btw, what is PJPG
[11:45] <burek> did you try to hexdump the output of your webcam
[11:45] <floater_afb60> that is the pixel format of the tesco corp webcam
[11:45] <burek> to see what exactly it is
[11:45] <floater_afb60> and v4l2 supports is
[11:45] <ubitux> floater_afb60: libv4l2 is a lib wrapper to handle old/nasty webcams
[11:45] <floater_afb60> gspca 7302
[11:45] <ubitux> it is mapped on v4l2 api, adding support for crazy cams
[11:46] <floater_afb60> i see, so its not thte same
[11:46] <floater_afb60> what will be the benefit ?
[11:46] <ubitux> just try it, it may work.
[11:47] <ubitux> if not then it might be interesting to look into the pixfmt stuff
[11:47] <burek> whatever I googled for PJPG it always returned something related to single jpeg image only
[11:47] <burek> so I'm wondering of -f image2pipe might help you with that
[11:47] <floater_afb60> it is pixelformat
[11:47] <burek> -of +if
[11:47] <floater_afb60> recently added in v42l 
[11:48] <burek> out of curiosity, try this on your camera: ffmpeg -f image2pipe -i http://... out.avi
[11:48] <floater_afb60> fmt_ff2v4l: start pix_fmt=-1 codec_id=0
[11:49] <ubitux> 10:31:24 < floater_afb60>     { PIX_FMT_NONE,    CODEC_ID_MJPEG,    V4L2_PIX_FMT_PJPG    },
[11:49] <floater_afb60> currently i am unsure if ffmpeg even extracts these informations from the cam
[11:49] <ubitux> did you test it?
[11:49] <floater_afb60> yes, recompiled with this, does something with a lot of errors, and no video being recorded currently
[11:50] <floater_afb60> fmt_ff2v4l: i=0 fmt_conversion_table[i].codec_id=14
[11:50] <floater_afb60> fmt_conversion_table[i].v4l2_fmt=842093913 
[11:51] <burek> floater_afb60, can you do: wget http://camera... output.dat
[11:51] <floater_afb60> the 480x640 should be 640x480
[11:51] <floater_afb60> ;/
[11:51] <burek> for like 10 seconds and upload it somewhere
[11:51] <floater_afb60> yea cool
[11:51] <floater_afb60> how to do that?
[11:51] <burek> use wget
[11:51] <floater_afb60> address?
[11:52] <floater_afb60> my cam is not on http, it is in /dev/video1
[11:52] <burek> cat /dev/video1 bla.dat
[11:53] <burek> > bla.dat
[11:53] <floater_afb60> o yea that's what i thought 
[11:56] <ubitux> seems PJPG stands for "progressive jpeg", it should work...
[11:56] <burek> I'm betting it's a kind of camera that sends jpgs one after another
[11:57] <ubitux> maybe you need to add the fourcc PJPG somewhere
[11:57] <burek> so you can link to it from a webpage using direct http url in <img src="..."
[11:57] <av500> well, all at the same time would not work :)
[11:57] <floater_afb60> http://jumbofiles.com/nky1qvxp6n9d
[11:57] <floater_afb60> heres the dump
[11:57] <burek> I've seen such cameras before
[11:57] <floater_afb60> a few seconds
[11:57] <floater_afb60> 100kb 
[11:57] <floater_afb60> ;/
[11:58] <floater_afb60> grabbed it using: $ dd if=/dev/video1 of=./trash
[11:58] <floater_afb60> did i miss the sector size?
[11:59] <ubitux> floater_afb60: can you paste the full output of a ffplay -f v4l2 -i /dev/video1 when you added "{ PIX_FMT_NONE,    CODEC_ID_MJPEG,    V4L2_PIX_FMT_PJPG    }," ?
[12:00] <floater_afb60> ah cat does get more data, i show that too
[12:00] <floater_afb60> http://jumbofiles.com/q75hp25t2swl
[12:01] <floater_afb60> ubitux<< 
[12:01] <floater_afb60> [video4linux2,v4l2 @ 0x7ff80c0008c0] Cannot find a proper format for codec_id 0, pix_fmt -1.
[12:01] <floater_afb60> /dev/video1: Input/output error
[12:01] <floater_afb60> heres the full output...
[12:02] <floater_afb60> haven't patched ffplay yet i think
[12:02] <floater_afb60> ah mixed up again, it went from /usr/local another version
[12:03] <floater_afb60> ok ffplay displays a 480x640 window
[12:03] <floater_afb60> and tries to render something, but stays grey
[12:04] <ubitux> can you pastebin the full output?
[12:04] <floater_afb60> no
[12:04] <ubitux> .
[12:04] <floater_afb60> when i press control C the last line gets cleared
[12:04] <ubitux> the last line isn't important :)
[12:04] <floater_afb60> last message was repeated ... times
[12:05] <ubitux> and what is the repeteated message?
[12:05] <ubitux> repeated*
[12:05] <floater_afb60> nah ok pasting
[12:05] <floater_afb60> http://pastebin.com/HXi7A4Wt
[12:06] <floater_afb60> $ /usr/bin/ffplay -f v4l2 -i /dev/video1
[12:06] <floater_afb60> and here's the raw cat /dev/video1 http://jumbofiles.com/q75hp25t2swl
[12:06] <ubitux> mmh ok
[12:07] <floater_afb60> cheese shows unchangable 640x480 resolution btw
[12:07] <floater_afb60> but unsure what may cause the width and height mix up
[12:09] <burek> http://pastebin.com/wRkzzLf0
[12:09] <burek> your width/height is mixed up
[12:10] <floater_afb60> yes i see
[12:13] <burek> floater_afb60, try this: ffmpeg -f image2pipe -vcodec mjpeg -i /dev/video1 -c copy out.avi
[12:14] <floater_afb60> burek<< it says [mjpeg @ 0x23e9b60] overread 8
[12:14] <floater_afb60> then hangs
[12:15] <floater_afb60> not sure what you want with that
[12:16] <burek> I'm assuming your "stream" is just a flow of jpeg images
[12:16] <burek> one after another
[12:16] <floater_afb60> probably
[12:17] <floater_afb60> in theory, cat can get them then...
[12:17] <floater_afb60> so what's with the PJPG format ?
[12:18] <floater_afb60> i thnk you can not do this anyway, since copy means no modification to the video stream
[12:19] <CTPAX> hi, does anyone have any experience with Google Native Client port of FFMpeg?
[12:20] Action: av500 thinks this channel is being taken over by serbs slowly
[12:21] <burek> :D
[12:21] <burek> silent revolution :)
[12:21] <av500> burek for everybody!
[12:21] <CTPAX> :D
[12:22] <floater_afb60> Stream #0:0: Video: mjpeg, yuvj422p, 480x640, -
[12:22] <floater_afb60> so stream dimension is incorrectly determined by ffmpeg
[12:23] <floater_afb60> and i dod this: /usr/bin/ffmpeg -s 640x480 -f v4l2 -r 20 -i /dev/video1 -c copy out_t.mjpeg
[12:23] <floater_afb60> -s should force the dimensions no?
[12:23] <floater_afb60> ;/
[12:27] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r60c950ac08 10ffmpeg/libavcodec/dwt.c: 
[12:27] <CIA-105> ffmpeg: dwt: use av_freep()
[12:27] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:28] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * re15f1286e6 10ffmpeg/libavcodec/ffv1.c: 
[12:28] <CIA-105> ffmpeg: ffv1: fix decoder state_transition table
[12:28] <CIA-105> ffmpeg: This fixes a regression with ffv1.1 coder=1 decoding
[12:28] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:30] <peuhkura> Hi! License question. I am compiling ffmpeg against openssl on android. There are some minor mods made to FFmpeg. Is it enough to publish the sw linking to modded ffmpeg and openssl (android's existing .so) libs with the "LGPL + using OpenSSL" license text + link to ffmpeg version, patches, and configure flags?
[13:18] <michaelni> peuhkura, iam developer not lawyer, if you send me good patches iam happy to include them :)
[13:40] <ubitux> michaelni: i think the -filter_complex path needs some updates for filters
[13:40] <ubitux> like using abuffersink_old, or aresample instead of asyncts
[13:40] <ubitux> maybe vol & stuff
[13:41] <ubitux> (ah! i forgot to push the resample patch.)
[13:53] <michaelni> ubitux, hmm, seems i missed that it was duplicated :/
[13:53] <michaelni> do you want to fix it or should i look into it ?
[13:53] <ubitux> i can do it if you want, but not now
[13:54] <ubitux> i tried some random amerge stuff with complex filters, but it crashes (even with Nicolas patches)
[13:54] <ubitux> (and even replacing bufferskink with buffersink_old)
[13:54] <ubitux> so i think there are some more things to dig into, and i don't have the time right now for it
[13:56] <ubitux> (example: ./ffmpeg -i ~/fate-samples/4xm/dracula.4xm -filter_complex amerge -f null -)
[13:59] <michaelni> ubitux, ok, ill let you work on it, but if you run into some bigger issues dont hesitate to throw your work in progress toward me
[14:00] <ubitux> ok, but as i said, i won't do it right now
[14:00] <ubitux> and i wonder if nicolas isn't working on it ATM
[14:00] <ubitux> since he's playing with -amerge ATM, he might be debugging it :p
[14:00] <ubitux> s/-amerge/amerge/
[14:01] <michaelni> you maybe should mail him ...
[14:01] <michaelni> so no duplicated work happens and or if he wants to work on it ...
[14:02] <michaelni> and isnt aware of the issue ...
[14:02] <ubitux> yep i'll do :)
[14:17] <peuhkura> michaelni: ok :) I am just on white area here 'cos the FFmpeg needs "nonfree" flag to compile itself and then it's non-redistributable, or that's what the configure says. So it's OK to proceed if the OpenSSL is just mentioned? Would like to skip meeting the lawyers ;)
[14:21] <burek> peuhkura, you just need to provide your modifications to the public, the same way FFmpeg does
[14:22] <burek> and of course, you need to ask openssl developers if it's ok to link it with ffmpeg and your project under that kind of license
[14:39] <peuhkura> burek: thanks, I'll try to get answer from openssl side.
[15:13] <CIA-105> ffmpeg: 03Nicolas George 07master * re6aee46760 10ffmpeg/libavfilter/af_amerge.c: af_amerge: set samples rate in query_formats.
[15:13] <CIA-105> ffmpeg: 03Nicolas George 07master * r0da28d689c 10ffmpeg/libavfilter/af_amerge.c: 
[15:13] <CIA-105> ffmpeg: af_amerge: compute output timestamp.
[15:13] <CIA-105> ffmpeg: Use the timestamps from the first input.
[16:13] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * r9ddc47227d 10ffmpeg/libavcodec/Makefile: Fix compilation with --disable-everything --enable-demuxer=flac (--disable-yasm).
[16:13] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * r7a0d00dc79 10ffmpeg/libavcodec/Makefile: Fix compilation with --disable-everything --enable-demuxer=ogg (--disable-yasm).
[16:32] Action: Daemon404 wonders if there is any such simple format as "y4m with timestamps"
[16:32] <Daemon404> to pipe
[16:34] <JEEBsv> not really, people used nut for that IIRC
[16:34] <Daemon404> crap.
[16:34] <Daemon404> i need to get ffms2 output piped to 
[16:34] <Daemon404> ffmpeg with timestamps*
[16:48] <ubitux> michaelni: do you plan to rework the nut specs just like you did for ffv1?
[16:48] <Daemon404> yes i just looked at th plausability of writing a simple nut muxer
[16:48] <Daemon404> i ran away screaming
[16:49] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * rfb6c6b15d7 10ffmpeg/libavcodec/avuienc.c: avuienc: Use field_order to determine if a stream is interlaced.
[16:49] <CIA-105> ffmpeg: 03Carl Eugen Hoyos 07master * r5e268633d1 10ffmpeg/libavcodec/avuienc.c: Do not leak extradata when encoding avui.
[16:54] <michaelni> Daemon404, maybe you could use libnut
[16:54] <Daemon404> isnt that terribly dead
[16:55] <michaelni> ubitux, would be a good idea, if i find the time
[16:56] <michaelni> its not actively developed currently AFAIK, but maybe you should ask ods15 
[16:56] <Daemon404> so far the sanest thing seems to be to use lavf to mux...
[16:56] <Daemon404> and thats not at all sane
[16:58] <michaelni> lavf "The insane" :)
[16:59] <Daemon404> alternatively, if ffmpeg input timecodes files like x264cli can
[16:59] <Daemon404> that'd also solve it
[17:00] <RobertNagy> michaelni: is there any possibility for atomic operations or memory barriers in ffmpeg?
[17:06] <michaelni> RobertNagy, theres ff_lockmgr_cb(), theres use of pthreads under #if theres asm() with stuff like sfence
[17:07] <RobertNagy> ok, I'll look into those
[17:07] <RobertNagy> thx
[17:08] <RobertNagy> btw, where can I find the coding conventions?
[17:08] <RobertNagy> for ffmpeg
[17:09] <michaelni> http://ffmpeg.org/developer.html#Coding-Rules-1
[17:11] <michaelni> Daemon404, if theres some input format lavf doesnt support, then adding support for it is welcome
[19:52] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r121d3875b6 10ffmpeg/libavformat/utils.c: 
[19:52] <CIA-105> ffmpeg: lavf: print the invalid tb in avpriv_set_pts_info()
[19:52] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:52] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r3cb83b225e 10ffmpeg/libavcodec/xsubdec.c: 
[19:52] <CIA-105> ffmpeg: xsubdec: print frame size when too small
[19:52] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:52] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * r212f129fb8 10ffmpeg/libavformat/avidec.c: 
[19:52] <CIA-105> ffmpeg: avidec: discard 0 packets even for sample_size==0 NI
[19:52] <CIA-105> ffmpeg: Fixes Ticket1332
[19:52] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:52] <CIA-105> ffmpeg: 03Michael Niedermayer 07master * rd0f78e77e1 10ffmpeg/libavformat/avidec.c: 
[19:52] <CIA-105> ffmpeg: avidec: simplify avi_read_idx1()
[19:52] <CIA-105> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:55] <Compn> oo
[20:55] <Compn> interesting idea
[20:56] <Compn> downloaded avi file , has small missing piece in the middle
[20:56] <Compn> lavf fails to demux it after the corruption
[20:56] <Compn> divfix++ is able to fix it 
[20:56] <Daemon404> i usually use virtualdub
[20:57] <Daemon404> as it can rebuild avi indices
[20:57] <Compn> http://svn.divfix.org/viewvc/divfixpp/trunk/
[20:57] <Compn> using mplayer -forceidx , stops at the error 
[20:57] <Compn> ahh its c++
[20:58] <Daemon404> god. sf.net's viewvc is so horrible.
[20:58] <Daemon404> makes my cry every time
[21:04] Action: Daemon404 wonders if it's possible to specify his own pts/dts (similar to loading a timecodes file)
[21:04] <Compn> isnt it the same thing ?
[21:04] <Daemon404> i see vf_fps and vf_pts, but i havent read what they actually do
[21:04] <Daemon404> Compn, yes but is it in any way possible
[21:04] <Daemon404> lol
[21:04] <Compn> oh
[21:04] <Compn> ffmpeg timecodes file
[21:05] <Daemon404> x264 has --tcfile-in
[21:06] Action: Daemon404 settles for watching MIPS hilarity instead
[21:06] <Daemon404> i do wonder who in their right mind transcodes or plays back on MIPS...
[21:07] <Compn> the future is embedded
[21:07] <Compn> laugha while you can, monkey boy!
[21:07] <Daemon404> teh future is ARM
[21:07] <Daemon404> do i have a MIPS phone? no.
[21:07] <Daemon404> mips is my router.
[21:19] <JEEB> I have something that plays H.264 and that is MIPS, but it indeed doesn't do it on the CPU
[21:19] <JEEB> (PSP)
[21:19] <Daemon404> ps vita is arm
[21:19] <Daemon404> irc
[21:19] <Daemon404> iirc*
[21:19] <JEEB> yeah
[21:19] <JEEB> PS1,2 and the PSP were MIPS
[21:19] <Daemon404> n64 too
[21:22] <Daemon404> this mips thread is sitll pretty funny
[21:22] <Daemon404> bad design all around
[21:30] <kierank> Daemon404: have you got a vfr sample?
[21:30] <kierank> i want to see what facebook does to it
[21:30] <Daemon404> kierank, no but i can make one trivially with l-smash or mkvtoolnix
[21:35] <JEEB> http://www.cccp-project.net/beta/test_files/cbed_lavc_vfr.mkv
[21:35] <JEEB> old vfr sample myrs made
[21:36] <kierank> as long as it's not anime i guess it's ok to upload to facebook
[21:37] <Daemon404> its anime.
[21:37] Action: Daemon404 bets cbed = cowboy bebop ed
[21:37] <JEEB> yes, unfortunately it is >_>
[21:38] <Daemon404> https://japland.org/foreman_vfr.mkv
[21:39] <Daemon404> here
[21:39] <Daemon404> i gotta run now
[21:39] <JEEB> oh nice
[21:39] <JEEB> vfr foreman
[21:43] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=ffplay.c;h=337a39066eeb5aacaa0e6febb18d4337d26a2599;hb=HEAD#l841
[21:43] <ubitux> is this some kind of windowing?
[21:43] <ubitux> double w = (x-nb_freq) * (1.0 / nb_freq);  this makes the value ranging linearly from -1 to 1
[21:45] <ubitux> (afaiu)
[21:45] <kierank> facebook seems happy with vfr
[21:47] <ubitux> and the 1 - w*w next line looks like it's making some kind of hyperball (or at least a triangle)
[21:47] <ubitux> so it looks like basic windowing
[21:47] <ubitux> but i can't tell what is the name
[22:01] <CIA-105> ffmpeg: 03Diego Biurrun 07master * rcc7fac9970 10ffmpeg/libavcodec/ (dwt.c dwt.h snowdec.c): dwt: return errors from ff_slice_buffer_init()
[22:01] <CIA-105> ffmpeg: 03Justin Ruggles 07master * r713548cbad 10ffmpeg/libavcodec/x86/dsputil_yasm.asm: 
[22:01] <CIA-105> ffmpeg: x86: lavc: use %if HAVE_AVX guards around AVX functions in yasm code.
[22:01] <CIA-105> ffmpeg: This is needed for older versions of yasm/nasm that do not support AVX.
[22:01] <CIA-105> ffmpeg: Signed-off-by: Diego Biurrun <diego at biurrun.de>
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * rf297dd3812 10ffmpeg/libavfilter/af_asyncts.c: 
[22:01] <CIA-105> ffmpeg: asyncts: reset delta to 0 when it's not used.
[22:01] <CIA-105> ffmpeg: Fixes an invalid write.
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * r0629f6128e 10ffmpeg/avconv.c: 
[22:01] <CIA-105> ffmpeg: avconv: rename transcode_audio/video to decode_audio/video.
[22:01] <CIA-105> ffmpeg: Those functions do only decoding now, encoding is run from
[22:01] <CIA-105> ffmpeg: poll_filters().
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * rccefa89f39 10ffmpeg/libavfilter/avfiltergraph.c: avfiltergraph: smarter sample format selection.
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * r43c7a01e98 10ffmpeg/ (10 files in 2 dirs): 
[22:01] <CIA-105> ffmpeg: lavfi: remove avfilter_default_* from public API on next bump.
[22:01] <CIA-105> ffmpeg: Those functions are only useful inside filters. It is better to not
[22:01] <CIA-105> ffmpeg: support user filters until the API is more stable.
[22:01] <CIA-105> ffmpeg: 03Mina Nagy Zaki 07master * r88c3b87bd8 10ffmpeg/libavfilter/ (Makefile avfilter.c avfilter.h defaults.c version.h): 
[22:01] <CIA-105> ffmpeg: lavfi: deprecate default config_props() callback and refactor avfilter_config_links()
[22:01] <CIA-105> ffmpeg: Link properties have to be checked after config_props() is called to
[22:01] <CIA-105> ffmpeg: make sure everything is sane, so the default config_props() for output
[22:01] <CIA-105> ffmpeg: links was redundant.
[22:01] <CIA-105> ffmpeg: Remove now empty defaults.c
[22:01] <CIA-105> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * rc04c533f62 10ffmpeg/libavfilter/ (21 files): 
[22:01] <CIA-105> ffmpeg: lavfi: remove avfilter_null_* from public API on next bump.
[22:01] <CIA-105> ffmpeg: Those functions are only useful inside filters. It is better to not
[22:01] <CIA-105> ffmpeg: support user filters until the API is more stable.
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * rce353a881f 10ffmpeg/libavfilter/avfilter.h: 
[22:01] <CIA-105> ffmpeg: lavfi: fix incorrect comment.
[22:01] <CIA-105> ffmpeg: poll_frame()/request_frame() are used for audio too.
[22:01] <CIA-105> ffmpeg: 03Anton Khirnov 07master * r0ce4a627c9 10ffmpeg/libavfilter/avfilter.h: lavfi: fix comment, audio is supported now.
[22:02] <CIA-105> ffmpeg: avio: Add a function for signalling end of reading/writing
[22:02] <CIA-105> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[22:02] <CIA-105> ffmpeg: 03Samuel Pitoiset 07master * r4a9ca93556 10ffmpeg/libavformat/ (os_support.h tcp.c): 
[22:02] <CIA-105> ffmpeg: tcp: Allow signalling end of reading/writing
[22:02] <CIA-105> ffmpeg: tcp_shutdown() isn't needed at the moment, but is added for
[22:02] <CIA-105> ffmpeg: consistency to explain how the function is supposed to be used.
[22:02] <CIA-105> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[22:02] <CIA-105> ffmpeg: 03Samuel Pitoiset 07master * re5773d8bc3 10ffmpeg/libavformat/http.c: 
[22:02] <CIA-105> ffmpeg: http: Add support for reading http POST reply headers
[22:02] <CIA-105> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[22:02] <CIA-105> (6 lines omitted)
[22:18] <ubitux> oh, hi michaelni :)
[22:18] <ubitux> michaelni: so no known name for that windowing?
[22:54] <michaelni> ubitux, you can call it 2nd order polynomial windowing but i suspect it has a more standard name :)
[22:54] <ubitux> ok :)
[22:55] <ubitux> btw, didn't you rewrite some kind of fifo stuff recently?
[22:55] <ubitux> i remember you were mentioning a more effective way of doing sth
[22:55] <ubitux> it was related to audio iirc
[22:55] <ubitux> but audio_fifo.c seems all ruggles stuff
[22:55] <ubitux> i can't find again what it was
[22:56] <michaelni> libavcodec/audio_frame_queue.c ?
[22:56] <ubitux> ah right, should be that
[22:56] <ubitux> indeed that was it.
[22:57] <ubitux> mmh ok, actually i need audio_fifo.
[23:06] <floater_afb60> Daemon404<< i did this:
[23:06] <floater_afb60> fdate=$(stat -c %y "$1" | sed -E 's;[^0-9]*([0-9]+-[0-9]+-[0-9]+)[ ]+([0-9]+:[0-9]+:[0-9]+)\.[0-9]+[ ]*(.*);\1T\2\3;')
[23:06] <floater_afb60> fbn=$(sed -E 's;(.*)\..*$;\1;' <<< "$1")
[23:07] <floater_afb60> convert "$1"  -font Arial -pointsize 10 -draw "gravity SouthEast  fill black  text 0,12 '""$fdate""'  fill white  text 1,11 '""$fdate""' "   t_"$fbn".png
[23:07] <floater_afb60> using imagick, and i have timelapse video
[23:24] <ubitux> saste: the spectrum filter is pretty fun to work on :)
[23:24] <ubitux> do you plan to commit the showwaves soon?
[23:24] <saste> yes, i need to fix the points raised by nicolas, which are not hard to fix but still may require some thinking
[23:25] <ubitux> ok :)
[23:52] <ubitux> saste: i need to raise a bufferef slightly changed from the previous one in my filter, what's the best way to do that?
[23:53] <ubitux> get a video buffer once, and raise it (with some kind of copy flag?) regularly?
[23:54] <saste> "raise"?
[23:54] <ubitux> send to the next filter
[23:54] <ubitux> sorry, weird representation in mind :)
[23:55] <saste> not sure we have an efficient way to deal with that
[23:55] <ubitux> avfilter_start_frame(outlink, avfilter_ref_buffer(mybuf, ~0)); ?
[23:56] <saste> you may try this one
[23:56] <saste> you reference a frame, and send a new reference with no write permission
[23:56] <saste> so following frames can't change it (but will need to copy a new buffer in case they need)
[23:57] <saste> in your filter you can now reuse the buffer, which was guaranteed not to change
[23:57] <saste> the problem is that you're just moving the problem, since other filters may need to do the copy anyway
[23:57] <saste> so i don't know how much efficient this will be
[23:57] <ubitux> "not my business" ;)
[23:59] <ubitux> assuming i did the get_video_buffer once in the config_props (is that even possible at that point?), i will need to unref it only once, in the uninit?
[00:00] --- Thu May 24 2012