[Ffmpeg-devel-irc] ffmpeg.log.20130710

burek burek021 at gmail.com
Thu Jul 11 02:05:01 CEST 2013


[00:06] <johnsmith_> Hi, has anyone ever had any success whatsoever installing the latest ffmpeg onto CENTOS 5/RHEL 5?
[00:09] <ulkesh> johnsmith_: sorry, I installed it recently on a centOS 6 VM, but not tried on centOS 5
[00:27] <goldfndr> Zeranoe: the current ffmpeg.exe shared (http://ffmpeg.zeranoe.com/builds/win32/shared/ffmpeg-20130709-git-00b1401-win32-shared.7z) gives an error on WinXP SP3, "ffmpeg.exe - Unable To Locate Component", "This application has failed to start because libstdc++-6.dll was not found. Re-installing the application may fix this problem."
[00:27] <goldfndr> http://ffmpeg.zeranoe.com/builds/win32/shared/ffmpeg-1.2-win32-shared.7z (from http://ffmpeg.zeranoe.com/builds/win32/shared/) doesn't have this error.
[00:30] <tm512> Does anyone know how to stream with ffmpeg, without the audio and video being completely out of sync?
[00:30] Action: goldfndr didn't see this documented/mentioned in the IRC logs at http://lists.ffmpeg.org/pipermail/ffmpeg-devel-irc/
[00:57] <curfont> When trying to compile something that uses -lswscale with gcc it works, but g++ doesn't find the reference
[00:57] <curfont> Is there some option I am missing?
[01:00] <curfont> g++ test.cpp -L/home/functor/ffmpeg_build/lib -I/home/functor/ffmpeg_build/include -lavformat -lavcodec -lswscale -lavutil -lm -D__STDC_CONSTANT_MACROS
[01:00] <curfont> all I am doing in the test.cpp is calling sws_scale with 0,0,0,0 params, but its not finding it as a ref
[01:00] <curfont> if I use gcc instead of g++ it works
[01:01] <mark4o> curfont: did you put extern "C" { } around the #include?
[01:02] <curfont> oh I see, let me try that
[01:03] <curfont> extern "C" { #include <libswscale/swscale.h> }
[01:03] <curfont> Like this?
[01:03] <curfont> yeah it didn't work :/
[01:03] <mark4o> you need 3 lines
[01:03] <curfont> complains about "#" being stray now
[01:03] <curfont> ah ok
[01:04] <curfont> yes that worked thanks
[01:20] <curfont> whats the new version of av_set_parameters?
[01:24] <durandal11707> there is no new version for that one. iirc you use av_set_opt
[01:27] <curfont> I was reading that they use avformat_write_header instead now
[01:27] <curfont> hmm
[04:30] <tm512> Does anyone here know how to stream from ffmpeg without getting a/v desyncs?
[04:45] <praveenmarkandu> hi, may i know what are the factors involved in the time it takes to transcode?
[04:45] <praveenmarkandu> is it bitrate? codec? resolution?
[09:17] <Crusher4> anyone around that can direct me to the docs on opencl for the 2.0 release? what platforms/configs it's supported in/etc?
[09:34] <Mavrik> Crusher4, opencl support was just released, there isn't even an announcement on ffmpeg page yet.
[09:34] <Mavrik> plus, what do you think OpenCL support does?
[09:35] <Crusher4> saw it on phoronix. the shortlog wasnt specific in how it was integrated, but it was even less specific on which platforms ffmpeg runs on that opencl was supported on
[09:36] <Crusher4> hence i came here, to see if anyone had more info other than "adds opencl support"
[09:36] <Mavrik> yes, the framework has preliminary openCL support
[09:36] <Mavrik> right now it's used by deshake and denoise filters
[09:36] <Mavrik> that's about it.
[09:37] <Crusher4> that's a start. on which platforms?
[09:37] <Mavrik> no idea
[09:37] <Mavrik> ones that support opencl I'd say.
[09:37] <Mavrik> check the configure script.
[09:38] <Crusher4> hm. good idea
[09:41] <Crusher4> ah. requires OpenCL 1.2, so nothing on osx for awhile then
[09:45] <sunjun> hi I have a problem about x86 compile
[09:46] <sunjun> libavcodec/x86/mpegvideo_mmx_template.c:144:9: error: can't find a register in class 'GENERAL_REGS'
[09:46] <Crusher4> Mavrik: thx for the help, laters
[09:46] <sunjun> while reloading 'asm'
[10:15] <Bor0> what is the best way to get current frame? currentpts*fps works but not too good
[10:28] <someone-noone> hello!
[10:28] <someone-noone> ./configure --enable-gpl --enable-libx264 --enable-libvpx --enable-libvorbis --enable-libopus --prefix=/usr
[10:28] <someone-noone> ERROR: opus not found
[10:28] <someone-noone> of course opus exists in /usr/include/opus * /usr/lib
[10:28] <someone-noone> ffmpeg version N-54577-g3c5071d
[10:28] <someone-noone> looks like a bug?
[10:29] <someone-noone> in config.log:
[10:29] <someone-noone> check_pkg_config opus opus_multistream.h opus_multistream_decoder_create
[10:29] <someone-noone> ERROR: opus not found
[10:30] <spaam> you forgot to install libopus-dev ?
[10:30] <spaam> ah nvm
[10:31] <someone-noone> I have /usr/include/opus/opus_multistream.h and there exists declaration of opus_multistream_decoder_create()
[10:31] <someone-noone> ls /usr/lib | grep opus
[10:31] <someone-noone> libopus.0.dylib
[10:31] <someone-noone> libopus.a
[10:31] <someone-noone> libopus.dylib
[10:31] <someone-noone> libopus.la
[10:31] <someone-noone> so, looks like opus is installed correctly
[10:32] <saste> someone-noone, did you check config.log?
[10:32] <someone-noone> yes, look above
[10:32] <saste> someone-noone, how did you install libopus?
[10:32] <saste> do you have the pkg-config files around?
[10:33] <someone-noone> cd opus-1.0.2 ; ./configure prefix=/usr; make install
[10:34] <someone-noone> I didn't do anything related to pkg-config
[10:34] <someone-noone> looks like I should do?
[10:38] <someone-noone> saste, how pkg-config files look like and where they should be located? I've never worked with pkg-config...
[10:38] <saste> someone-noone, which system?
[10:39] <saste> pkg-config libopus --version
[10:39] <someone-noone> mac os x
[10:40] <saste> usually you should find a lib/pkgconfig containing pkg-config metadata files
[10:40] <someone-noone> hm, it printed: 0.26 . Looks like libopus is installed somewhere
[10:40] <saste> which are looked after by pkg-config
[10:41] <saste> someone-noone, weird...
[10:41] <someone-noone> ?
[10:49] <someone-noone> Okay, I've reinstalled libopus
[10:50] <someone-noone> now I do: ls /usr/lib/pkgconfig/ | grep opus
[10:50] <someone-noone> It outputs: opus.pc
[10:50] <someone-noone> however:  pkg-config opus --cflags
[10:50] <someone-noone> Package opus was not found in the pkg-config search path.
[10:50] <someone-noone> Perhaps you should add the directory containing `opus.pc'
[10:50] <someone-noone> to the PKG_CONFIG_PATH environment variable
[10:50] <someone-noone> No package 'opus' found
[10:51] <saste> someone-noone, what's the value of PKG_CONFIG_PATH?
[10:52] <someone-noone> Sorry for dummy question, but how can I know? echo $PKG_CONFIG_PATH is correct?
[10:52] <saste> yes
[10:52] <someone-noone> it's empty
[10:52] <someone-noone> 0_0
[10:54] <saste> someone-noone, which should be fine, since it should look in /usr/lib/pkgconfig by default
[10:55] <someone-noone> saste, where is a problem?
[10:56] <mark4o> to show default package config path: pkg-config --variable pc_path pkg-config
[10:56] <mark4o> should be same hierarchy as pkg-config itself
[10:57] <someone-noone> mark4o, looks like you showed me the problem
[10:57] <someone-noone> macports have installed pkg-config and pathes are: /opt/local/lib/pkgconfig:/opt/local/share/pkgconfig
[10:58] <someone-noone> how should I add /usr/lib/pkgconfig path correctly ?
[10:58] <mark4o> set PKG_CONFIG_PATH?
[10:58] <mark4o> why do you want to install in /usr?
[10:59] <someone-noone> I know that it's in general not correct
[10:59] <mark4o> that is for the os, and may be overwritten on os update
[11:05] <someone-noone> mark40, I'm just studying shell and some programs don't see "/usr/local". Now I know how to work with environment variables and I just need to forget about this silly prefix...
[11:10] <someone-noone> finally it works :) thanks guys for help
[11:13] <mark4o> great, did you install it in /usr/local or /opt?
[11:14] <someone-noone> first of course, /opt is for macports
[11:15] <mark4o> ok so I guess you didn't need pkg-config, cool
[11:32] <ningujoan> hi there
[11:33] <ningujoan> does anyone know if there is any version where the ffserver does work? I keep getting error messages same as everybody searching on google but no answers :/
[11:33] <ningujoan> it might be a version issue thoug, anyone has it running?
[11:34] <Mavrik> ffserver works just fine.
[11:37] <Bor0> what is the best way to get current frame? currentpts*fps works but not too good, especially for low fps vids
[11:37] <ningujoan> it gives me a 404 when I try to connect:
[11:37] <ningujoan> [tcp @ 0x25feec0] TCP connection to localhost:8090 failed: Connection refused
[11:37] <ningujoan> [http @ 0x25f1380] HTTP error 404 Not Found
[11:37] <ningujoan> http://localhost:8090/feed1.ffm: Input/output error
[11:37] <ningujoan> and on the server: File '/feed1.ffm' not found
[11:38] <ningujoan> it's so simple, don't know what i'm doing wrong then :/
[11:38] <Mavrik> Bor0, "get" where?
[11:38] <Bor0> I want to retrieve current frame, let's say in ffplay
[11:38] <Mavrik> Bor0, command-line? source? you using libraries? is't not clear.
[11:38] <Bor0> in ffplay.c
[11:39] <Mavrik> ningujoan, no, it's giving you connection refused
[11:39] <Mavrik> ningujoan, which means it cannot connect to anything on port 8090
[11:39] <Bor0> I want to modify ffplay.c so that it'll display currentframe in console.
[11:39] <Mavrik> which means you either have a firewall or ffserver isn't running on that port
[11:39] <Bor0> currentpts * fps is not stable, especially for low-fps vids
[11:39] <Mavrik> Bor0, hmm, counting frames would be better probably
[11:39] <Bor0> when I press 'S' to frame step forward, it gives 1, 2, 3, 4, 4, 4, 8 or something
[11:39] <Bor0> but how can I count them? what if the user seeks somewhere, how will the counting be handled?
[11:40] <Mavrik> good question
[11:40] <Mavrik> Bor0, anyway, currentpts * fps is totally wrong
[11:40] <Mavrik> since PTS doesn't show time in seconds.
[11:40] <Bor0> it works for videos with good fps
[11:41] <Mavrik> current_time * fps would give you a good estimate
[11:41] <Bor0> well this is the exact formula that I use: currentframe = (double)(pts_int * is->slotinfo->fps/(is->video_st->time_base.den));
[11:41] <Mavrik> but your problem is that videos don't have to have fps
[11:41] <ningujoan> well, on the server it does connect: it gives a: File '/feed1.ffm' and then  127.0.0.1 - - [GET] "/feed1.ffm HTTP/1.1" 404 149 not found, and there are a number of questions like this one with no answer on google
[11:41] <Mavrik> and that you can't really know how many frames you had in the middle of video
[11:41] <Bor0> let's take out VBR for this case :)
[11:41] <Bor0> and focus on constant rate
[11:41] <d-fens_> hi, i need to generate a h264 from image series in a .net application, is there a recommened wrapper or way to archive this?
[11:42] <JEEB> d-fens_, either use libavcodec and friends as a native C library, or use x264 itself as a native C library?
[11:42] <Bor0> do you have any suggestions Mavrik ? or is this a tough thing to do
[11:43] <Mavrik> Bor0, it's certanly not easy
[11:43] <Mavrik> depends what your desired result is
[11:43] <d-fens_> JEEB: ok but then i have to deal with all pinkvoke and stuff my myself
[11:43] <Mavrik> with constant framerate you need to grab first PTS, calculate difference between current PTS and start PTS and then calculate number of frames with help of timebase
[11:44] <JEEB> d-fens_, let's just say that if someone actually made you a wrapper if that person wasn't really motivated it would be out-of-date by now :)
[11:44] <JEEB> there were lots of changes in APIs during 2011-2012
[11:44] <Mavrik> ningujoan, well, you DID define the feed1.ffm sink in ffserver configuration right_
[11:45] <d-fens_> ok, may be thera is a autogen-wrapper thingy hiden somewhere
[11:45] <JEEB> most probably there isn't, though :)
[11:45] <Bor0> so that's |pts_start - pts|/av_timebase against pts*currentframe/av_timebase ?
[11:46] <d-fens_> i see, it just that i have to deep dive into libavcodec internals
[11:46] <Bor0> pts*fps/av_timebase, sorry
[11:46] <JEEB> well, another alternative is to use libx264 itself without the libavcodec around it
[11:46] <d-fens_> which i would like to avoid
[11:46] <Bor0> is the former an approximation as well? can I be certain that it will give the correct current frame?
[11:46] <JEEB> and I don't think making AVFrames for an encoder nowadays is /that/ hard
[11:47] <d-fens_> jepp adding some mp3 afterwards should be easy too, i just thought about wrapping the exe as a dirty approach
[11:47] <mark4o> Bor0: don't you want to cast to double before the division?  otherwise it is integer division and you'll lose precision
[11:48] <JEEB> yes, just wrapping the exe and pushing stuff to stdout and piping to it is one hackjob way of doing it :)
[11:48] <Bor0> I'm doing all the needed casts
[11:48] <Mavrik> Bor0, don't convert to seconds until last second.
[11:48] <Mavrik> otherwise you're just throwing away precision
[11:48] <Bor0> you mean x = pts/av_timebase, then currentframe = fps * x ?
[11:48] <Mavrik> (current_pts - start_pts) / stream_time_base  * fps
[11:48] <d-fens_> JEEB: thanks for your input i'll check and read up on my options now
[11:49] <Bor0> what's an easy way to get the start_pts?
[11:49] <mark4o> Bor0: your (double) cast is after the division, should be before like (double)(current_pts - start_pts) * fps / tb
[11:50] <Bor0> what difference would start_pts make in that formula? isn't it a constant?
[11:53] <Mavrik> yes, it is. and?
[11:53] <Bor0> will not it also be an approximation?
[11:56] <Mavrik> pts values are absolute.
[11:59] <mark4o> Bor0: if you don't want to consider start_pts then this should be an improvement over what you have: currentframe = (double)pts_int * is->slotinfo->fps / is->video_st->time_base.den;
[12:01] <Bor0> let me try the one with startpts and currentpts
[13:42] <t4nk577> Hi, i'm trying to compile ffmpeg and x264 and got the following error in make
[13:42] <t4nk577> libx264.o : error LNK2001: unresolved external symbol _x264_bit_depth libavcodec/avcodec-54.dll : fatal error LNK1120: 1 unresolved externals make: *** [libavcodec/avcodec-54.dll] Error 1
[13:42] <t4nk577> can some help me with this?
[13:43] <Mattias> I have 'ffmpeg -i "filename" -y -ss "00:00:10" -s 720x460 -f image2 "thumb_filename"' but I want to have it save the aspect ratio and just go with the width
[13:43] <Mattias> How can I do that?
[13:46] <t4nk577> http://pastebin.com/RSeh09D3
[13:47] <t4nk577> this is my configure command and i have x264 builds in /home/username/x264-builds folder
[13:47] <JEEB> can you pastebin your config.log?
[13:49] <t4nk577> yes
[13:52] <t4nk577> http://pastebin.com/A3HUvW4F
[13:52] <t4nk577> this is configure log
[13:54] <JEEB> ok
[13:54] <JEEB> so configure doesn't even check for it :P
[13:54] <JEEB> oh it does
[13:56] <t4nk577> ok
[14:04] <mark4o> Mattias: ffmpeg -i infile -y -ss 10 -vf scale=720:-1 -vframes 1 thumbfile
[14:05] <Mattias> mark4o, thanks, got it to work now :)
[16:49] <mrAlmond> Hi all
[16:50] <mrAlmond> I neet to write custom metadata tags in an MP4 file muxed by my application in C code with ffmpeg
[16:50] <mrAlmond> which is the best way to do this?
[17:57] <juanmabc> 'tsup
[19:35] <klaxa> https://gist.github.com/klaxa/5968374
[19:36] <klaxa> where do i put -strict -2 so it gets recognized?
[19:36] <durandal_1707> after codec for which it is relevant
[19:36] <klaxa> as you can see i put it there
[19:38] <durandal_1707> than try it with before vp9
[19:39] <klaxa> nope
[19:39] <klaxa> argument gets parsed correctly even wtf
[19:39] <klaxa> >Reading option '-strict' ... matched as AVOption 'strict' with argument '-2'.
[19:40] <klaxa> doesn't look like it's getting applied though
[19:42] <durandal_1707> -strict:v ?
[19:43] <klaxa> doesn't work either
[19:44] <klaxa> https://gist.github.com/klaxa/5968451
[19:44] <klaxa> with -loglevel debug
[19:56] <klaxa> filed bugreport http://www.ffmpeg.org/trac/ffmpeg/ticket/2772
[20:24] <tm512> does anyone at all know how to stream with ffmpeg, without getting desyncs?
[20:33] <killown> ffmpeg now has opencl support? there is a tutorial for the opencl usage?
[21:58] <humble_user> I'm looking for a little guidance regarding ffprobe. I'm wondering if it's feasible to use ffprobe to parse a transport stream of broadcast tv in order to get channel information (i.e. the name of the channel or call sign). Has anyone here ever done anything like that?
[00:00] --- Thu Jul 11 2013


More information about the Ffmpeg-devel-irc mailing list