[Ffmpeg-devel-irc] ffmpeg.log.20151004

burek burek021 at gmail.com
Mon Oct 5 02:05:01 CEST 2015

[00:37:29 CEST] <p8r> Hi, playing a stream with ffplay I do get audio dropouts and the message "ALSA lib pcm.c:7905:(snd_pcm_recover) underrun occurred". Can I raise ffplay's audio playback buffer or is there other ways of working around it?
[00:48:58 CEST] <klaxa> alsa underruns are something i never got fixed
[00:49:12 CEST] <BtbN> foreverska, could you pastebin the output of nvidia-smi?
[00:49:42 CEST] <BtbN> nvenc is entirely X independend. It also works headless. Only the nvidia module has to be loaded.
[00:50:48 CEST] <foreverska> http://pastebin.com/kwz5PWuY
[00:52:11 CEST] <maduro> hi all, I can't seem to get ffmpeg above ~15% cpu usage (on a 40 core/160gb EC2 instance) when transcoding prores to x264 - all cores are used, but they sit at ~15% usage, with one much higher - where's the bottleneck?
[00:52:14 CEST] <BtbN> Looks fine, should just work.
[00:52:51 CEST] <BtbN> Could you also paste your ffmpeg commandline and and output(With -v verbose)?
[00:54:39 CEST] <foreverska> http://pastebin.com/pA30ABvQ
[00:56:14 CEST] <c_14> maduro: pastebin your commandline and output
[00:56:50 CEST] <dekcuf> maduro: did you compile yasm?
[01:00:41 CEST] <maduro> http://pastebin.com/LSZguVcu    -   i terminated it early since it's a ~52 min run
[01:00:50 CEST] <foreverska> Based on it's blurb about my cards I'm finding posts from last year saying GeForce isn't supported in linux,  that's not still true is it BtbN
[01:01:04 CEST] <maduro> dekcuf: i installed the binaries from the debian-multimedia repo
[01:07:32 CEST] <c_14> maduro: it's probably the scale fliter, iirc it doesn't thread well
[01:08:17 CEST] <maduro> c_14: i see, is there anything that can be done about it?
[01:10:19 CEST] <BtbN> foreverska, hm? Everything that's Kepler or newer works.
[01:10:34 CEST] <BtbN> There are a few exceptions, but Desktop GTX cards aren't.
[01:11:05 CEST] <c_14> maduro: Besides rewriting the filter api/code so it threads better? Cutting the input files into chunks, running separate ffmpeg processes and then concatting them.
[01:12:53 CEST] <maduro> c_14: would it be possible to do that one (long-ass) line, using -map, -filter_complex, and multiple inputs/outputs?
[01:13:09 CEST] <maduro> do that one/do that in one
[01:14:06 CEST] <BtbN> foreverska, [ GPU #0 - < GeForce GTX 660 > has Compute SM 3.0, NVENC Not Available ]
[01:14:14 CEST] <BtbN> NVENC is available starting from SM 3.5 Cards.
[01:14:20 CEST] <BtbN> So your card actualy does not support it.
[01:14:37 CEST] <BtbN> Missed that paste, sorry
[01:14:50 CEST] <foreverska> Oh... that's disappointing.
[01:16:18 CEST] <c_14> maduro: No, you'd need n ffmpeg processes processing n chunks and then one process at the end that combines them
[01:16:35 CEST] <maduro> c_14: ok, ty
[01:17:15 CEST] <foreverska> well what is this libmfx thing I need then?  Maybe I can get my processor's gpu to take up the good fight
[01:22:24 CEST] <BtbN> That stuff was painfull the last time i tried it. You have to patch your kernel and stuff
[02:50:22 CEST] <DHE> maksbotan: that HLS error has been fixed and committed to the git version of ffmpeg
[02:51:33 CEST] <fred1807> canI get the direct link to this audio stream? http://www.bbc.co.uk/radio/player/bbc_6music
[04:12:03 CEST] <Matviy> I'm using ffmpeg to convert x264 to x265, is there a way to force the encoding process to always convert at a specific framerate, and dynamically adjust the other paramaters to achive that framerate?
[04:13:00 CEST] <Matviy> Like, if the CPU is free, it will use a lower preset to achive a lower KBPS, at 30fps. But if CPU utilization is high, it will raise the preset, higher kbps, but still 30fps
[04:20:39 CEST] <c_14> no
[05:41:14 CEST] <pinPoint> sounds like some coding is needed for that to work
[06:47:53 CEST] <Matviy> I'm trying to encode with VP9, and it's going at ~5fps using <10% of my CPU. Is this normal? I can't seem to find a way to speed it up
[06:49:21 CEST] <Matviy> I read some forum posts that suggest splitting my video into NUMBER_OF_CORES pieces, and running parallel instances of ffmpeg, wtf...
[06:49:33 CEST] <pinPoint> heh
[06:49:49 CEST] <pinPoint> just curious, what is teh code you parse?
[06:50:25 CEST] <Matviy> Code?
[06:51:19 CEST] <pinPoint> ffmpeg command line
[06:54:51 CEST] <Matviy> ffmpeg.exe -i C:\original.mp4 -c:v libvpx-vp9 -minrate 750K -maxrate 750K -b:v 750K -c:a libvorbis -threads 36 output.webm
[06:55:15 CEST] <Matviy> (I do have 36 cores)
[06:56:00 CEST] <Matviy> h265 encodes at ~150fps ~30% cpu usage, vp9 goes at like 5fps 5% usage
[06:56:37 CEST] <pinPoint> what about using libvpx and take out -threads 36
[06:57:23 CEST] <c_14> libvpx is vp8, not vp9
[06:57:43 CEST] <pinPoint> ok take threads 36 out then
[06:57:56 CEST] <c_14> Also, I think you need to set -tile-columns to something for multithreaded encoding with libvpx-vp9
[06:59:04 CEST] <Matviy> removing the threads makes it slower by half
[06:59:06 CEST] <YaMoonSun> ffmpeg -i "Original.mp4" -c:v libvpx-vp9 -b:v 650k -c:v libvorbis -b:a 128k "Output.webm"
[07:03:42 CEST] <YaMoonSun> Pretty sure that VP8 and VP9 encoding only utilize once core when encoding, and there's a difference between threads and cores I believe.
[07:06:23 CEST] <c_14> The vp9 encoder has threading support. It's not very good, but it's there.
[07:06:59 CEST] <Matviy> Tried setting the -title-columns, doesn't make a difference. CPU usage caps at 10% no matter what
[07:07:21 CEST] <c_14> Like I said, it's not very good.
[07:07:51 CEST] <Matviy> :( Would that be an issue with all vp9 encoders everywhere? Or just ffmpeg?
[07:07:57 CEST] <Matviy> I'm guessing they all use the same SDK
[07:08:02 CEST] <c_14> There is only one vp9 encoder (afaik)
[07:08:37 CEST] <YaMoonSun> Should see how many threads each core has and then reassign the number of threads you're using.
[07:08:57 CEST] <c_14> That doesn't make sense.
[07:09:18 CEST] <c_14> Unless you're talking about Hyperthreading, in which case it doesn't matter (mostly).
[07:09:24 CEST] <YaMoonSun> Not much in life does, but we make it work anyways.
[07:09:42 CEST] <Matviy> How is google trying to push vp9 without good multi-threading support
[07:09:44 CEST] <Matviy> i dont even
[07:09:59 CEST] <pinPoint> not for long... with Open Alliance in the way
[07:10:01 CEST] <c_14> Matviy: google doesn't want you to encode vp9. They just want you able to watch vp9 on youtube.
[07:10:18 CEST] <pinPoint> I can try vp9 on my side. got a 6core cpu I can test against on the hackintosh
[07:10:29 CEST] <YaMoonSun> They're allowing vp9 on youtube now? I thought they only take x264 and libvpx?
[07:10:44 CEST] <Matviy> I want to upload VP9 because i pay .09 per GB. I can save 50% by using HEVC/VP9, but youtube doesn't allow h265 upload
[07:10:50 CEST] <pinPoint> v10 is already in the way no?
[07:11:04 CEST] <pinPoint> you give them a file and they make many versions of it including v9
[07:11:20 CEST] <YaMoonSun> Resolution?
[07:11:40 CEST] <pinPoint> I have seen 2700p
[07:12:48 CEST] <YaMoonSun> Ah, was directed @ Matviy
[07:12:51 CEST] <YaMoonSun> 2700p, lol, shit
[07:12:57 CEST] <YaMoonSun> Excuse my language
[07:13:23 CEST] <Matviy> 720p is what I'm uploading.
[07:17:39 CEST] <YaMoonSun> Ah, well damn dude, how long is the media file?
[07:17:40 CEST] <pinPoint> yeah, I'm watching a vp9 sampe go at fps 4 at best. The cpu is barely cranking... my fans are low
[07:17:47 CEST] <pinPoint> sample*
[07:20:39 CEST] <YaMoonSun> I'd upload it using vp8 @ 1600k
[07:20:49 CEST] <YaMoonSun> .09 isn't that much
[07:21:11 CEST] <pinPoint> or mp4/h264 and high bitrate, let google its nonsense in the background
[07:21:23 CEST] <pinPoint> let google do its*
[07:21:32 CEST] <YaMoonSun> It's not nonesense =/
[07:22:33 CEST] <Matviy> @pinPoint, Seems like the correct solution really is to split the file and run multipe encoding processes, lol
[07:26:12 CEST] <YaMoonSun> Doesn't -threads 0 give you the max?
[07:26:44 CEST] <pinPoint> mine does not. :/
[07:27:01 CEST] <YaMoonSun> Huh, maybe I should try things before I type them =p
[07:27:28 CEST] <pinPoint> i think libvpx/vpx-v9 is just slow for now.
[07:29:01 CEST] <YaMoonSun> Yeah, it maxed out one core for me.
[07:29:41 CEST] <YaMoonSun> I just got it to use three cores =0
[07:30:07 CEST] <YaMoonSun> Should check speccy or what have you to see how many threads your CPU has.
[07:30:35 CEST] <Matviy> On intel, Threads = (physical cores * 2)
[07:31:00 CEST] <YaMoonSun> You're using what processor?
[07:31:24 CEST] <YaMoonSun> I did -threads 4 and I got it up to 85% CPU vs 25%
[07:32:01 CEST] <Matviy> I'm using the c4 AWS instance, which has 36 vCPU's
[07:32:14 CEST] <Matviy> Intel Xeon E5-2666 v3 (Haswell)
[07:32:27 CEST] <Matviy> There isn't a spec page for it on intel because it's a custom processor made for AWS
[07:32:50 CEST] <YaMoonSun> I don't think ffmpeg was written with that processor in mind, but could be wrong.
[07:53:06 CEST] <TD-Linux> Matviy, indeed you need -tile-columns in order to get parallel encoding
[07:53:23 CEST] <TD-Linux> (or split up your video into segments. this is what youtube does internally)
[08:01:40 CEST] <Matviy> @TD-Linux, may i ask how you know about Youtube? :D
[08:02:16 CEST] <pinPoint> the boss
[08:02:18 CEST] <pinPoint> :)
[08:06:29 CEST] <pinPoint> c_14: would you say prores and prores_aw are the same?
[08:08:46 CEST] <TD-Linux> Matviy, it's obvious if you look at their dash encodes
[08:08:58 CEST] <pinPoint> youtube-dl
[08:10:46 CEST] <pinPoint> Matviy: http://pastebin.com/Z1Eery8F TD-Linux
[08:48:17 CEST] <Matviy> I'm so confused
[08:48:53 CEST] <Matviy> Using normal paramaters, i get 10% CPU usage, 850kbps, and the conversion speed is ~25fps. Great
[08:49:12 CEST] <Matviy> If i set "-quality realtime"
[08:49:23 CEST] <Matviy> I'm suddenly converting at 200FPS
[08:49:28 CEST] <Matviy> Still 10% CPU usage
[08:49:39 CEST] <Matviy> Still 850kbps (i deliberatly chose 850)
[08:49:44 CEST] <Matviy> Same file size
[08:50:20 CEST] <Matviy> That doesn't make sense. If the quality is lower, why is the rate still 850kbps. What is all that extra data?
[08:50:35 CEST] <Matviy> If the quality is the same, how am i converting at 8x the speed with the same CPU usage
[08:51:33 CEST] <TD-Linux> so encoding faster changes the quality per bit
[08:52:06 CEST] <TD-Linux> realtime changes VP9's deadline to go really fast. it doesn't control bitrate or target quality
[08:53:16 CEST] <Matviy> Would using realtime decrease the quality?
[08:54:02 CEST] <TD-Linux> yes. if you're getting the same bitrate, then the quality is worse
[08:56:06 CEST] <TD-Linux> rather than using -quality realtime, you can use -cpu-used (-speed in ffmpeg?) to set a performance level between 0 and 7
[08:56:21 CEST] <TD-Linux> there was a comparison done recently on the effect on quality: https://blogs.gnome.org/rbultje/2015/09/28/vp9-encodingdecoding-performance-vs-hevch-264/
[09:03:14 CEST] <Matviy> Thanks TD, i think my issue is that i can't get it to use more than 1 CPU
[09:03:34 CEST] <Matviy> I thought "-cpu-used" was my ticket to success, apparently not.
[09:11:30 CEST] <xerox> (also asked on #x264 because I think I'm using *that* wrong, but) what am I doing wrong? http://pastebin.com/raw.php?i=kSMQ7XM2
[09:12:04 CEST] <YaMoonSun> Can I use ffmpeg to export every frame in my video as a .jpg?
[09:12:13 CEST] <xerox> YaMoonSun: yeah I think so
[09:13:51 CEST] <xerox> I wonder if it's enough to use the -r option with the inverse of the frame rate
[09:15:27 CEST] <xerox> YaMoonSun: how about http://superuser.com/a/729351 maybe you don't even need the -r
[09:20:21 CEST] <YaMoonSun> Currently using a scene filter via vlc to do it, when I ask for 1 jpg per frame it glitches and ruins all my photos.
[10:36:56 CEST] <Skilly> is it desired that -scodec copy -acodec copy -vcodec copy doesn't keep the names of the streams it copies?
[11:07:44 CEST] <anYc> Hi, I reencode an h264 stream into a MKV and avcodec_encode_video2 sometimes returns a packet with an I frame before a B frame with a lower PTS, e.g., I: PTS=5 DTS=1, B: PTS=4 DTS=2, P: PTS=6 DTS=3. Does someone have an idea what I might doing wrong or is this valid behavior?
[11:07:49 CEST] <anYc> corresponding code and output: http://pastebin.com/rxqNWkyz
[11:11:25 CEST] <anYc> (I also removed my code that overwrites the DTS after avcodec_encode_video2 but this doesn't change the imho odd behavior)
[11:13:20 CEST] <anYc> I also checked that the the input frames are ordered by PTS before I give them to avcodec_encode_video2
[11:19:56 CEST] <ayusun> Hi, I have a doubt regarding license of ffmpeg. Can I ask it here?
[11:33:24 CEST] <ayusun> anybody here?
[12:17:40 CEST] <ayusun> hi
[12:18:57 CEST] <durandal_1707> just ask
[12:26:43 CEST] <ayusun> JavaCv uses apache license while FFMpeg uses GPL/LGPL license. So if I use JavaCv to build a program that uses FFMpeg wrapper, what sort of license will I be using(if I ever try to release that work) may be for commercial use?
[12:27:42 CEST] <ayusun> If I use it for commercial purpose, do I have to make anything open sourced?
[12:28:36 CEST] <ayusun> A page on FFMPEG says that https://www.ffmpeg.org/legal.html
[12:29:26 CEST] <ayusun> that I have to do few things that are listed there, So do I have to do those things even If use JavaCv over FFMPEG?
[12:35:55 CEST] <ayusun> The point 9 in the page of FFMpeg says that "Add "This software uses code of <a href=http://ffmpeg.org>FFmpeg</a> licensed under the <a href=http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html>LGPLv2.1</a> and its source can be downloaded <a href=link_to_your_sources>here</a>" to every page in your website where there is a download link to your application."
[12:36:35 CEST] <ayusun> what if my application is a web application and has jsp page which deals with the video
[12:37:02 CEST] <ayusun> and there is no download link of my application. Where and how will I write it?
[12:58:36 CEST] <durandal_1707> You should provide source for ffmpeg you use
[13:09:45 CEST] <sopparus> anyone with an amd card who tried encoding with opencl? does it help performance? quality is not that important to me, only need speed
[13:10:30 CEST] <durandal_1707> there is no code for encoding with opencl
[13:10:47 CEST] <durandal_1707> Only for filtering
[13:16:19 CEST] <sopparus> ah ok, thanks
[13:16:43 CEST] <sopparus> so no way I could use my amd card then?
[13:19:18 CEST] <durandal_1707> there are hw encoders in ffmpeg
[13:22:59 CEST] <sopparus> yeah, but intel only?
[13:23:03 CEST] <sopparus> from what I can fin
[13:23:03 CEST] <sopparus> d
[13:23:09 CEST] <sopparus> possibly nvidia too
[13:27:42 CEST] <Mavrik> You'll probably have to use something non-ffmpeg for AMD I guess.
[18:31:00 CEST] <maksbotan> DHE: nice, thanks!
[18:48:52 CEST] <iH2O> how can I replace the audio of a .flv file by a .mp3 file of the same duration
[18:53:50 CEST] <Mavrik> pass both files as an input (-i twice), use "-c copy" as codec and then use "-map" parameter to map video track from flv and audio track from mp3 to output.
[18:58:43 CEST] <iH2O> thank you
[19:11:00 CEST] <Renari> Does anyone know if there's a way to make ffmpeg overlay frames in a gif to reduce file size? (all the things I've tried output all of each frame rather than the difference).
[19:12:28 CEST] <YaMoonSun> Maybe add -framerate 1/3 at the beginning of the command, not absolutly possitive though
[19:15:27 CEST] <Renari> I don't want to lower the frame count rather I want the gif to use mostly transparent pixels.
[19:33:10 CEST] <Mavrik> Renari, you mean you want a proper intra-coded video format not GIF.
[19:42:26 CEST] <Renari> Mavrik, thanks but it seems imagemagick supports what I want with it's optimizetransparency flag.
[19:48:44 CEST] <Mavrik> Yeah, I'm rather sure ffmpeg doesn't.
[19:48:51 CEST] <Mavrik> Also stop using GIFs for use-cases where you need a video format.
[20:01:34 CEST] <wolfman2000> Afternoon. I am aware of the compilation guide on https://trac.ffmpeg.org/wiki/CompilationGuide/MSVC, but I feel it is outdated. Allegedly, FFmpeg should be compilable on visual studio due to snprintf working correctly. Is there an updated guide explaining how WITHOUT using ./configure && make which doesn't work by default with the normal command line?
[20:03:41 CEST] <Mavrik> Uh.
[20:03:57 CEST] <Mavrik> How would you compile it without configuring it first?
[20:04:04 CEST] <wolfman2000> Mavrik: cmake
[20:05:19 CEST] <Mavrik> Sadly that's not how it works.
[20:05:40 CEST] <ln-> eww, cmake
[20:07:14 CEST] <wolfman2000> ln-: I'm curious as to what your beef with cmake is, but we can take it off channel if necessary. For now, let's confirm something else. Mavrik: are you officially sure that no one could write their own configure script to generate an appropriate config.h file? cmake or otherwise?
[20:07:54 CEST] <Mavrik> wolfman2000, the thing is, configure script from ffmpeg isn't an autotools standard script
[20:08:09 CEST] <Mavrik> It's a rather horrible piece of bash and while, yes, someone could port it.
[20:08:23 CEST] <Mavrik> It's significantly more work than just installing msys.
[20:08:32 CEST] <Mavrik> It'll also keep breaking as git master is updated.
[20:08:47 CEST] <wolfman2000> do not tempt me. ;)
[20:09:20 CEST] <Mavrik> I guess you could generate config.h once with msys and commit it if you need it for automated compilation.
[20:09:26 CEST] <Mavrik> Also.
[20:09:29 CEST] Action: Mavrik slaps ln-.
[20:09:50 CEST] <wolfman2000> alright, let's take a different approach instead. http://ffmpeg.zeranoe.com/builds/  The page I linked to contains a number of builds for ffmpeg. However, they are all GPL. Is there a similar website out there but with LGPL builds?
[20:10:28 CEST] <JEEB> just use msys(2) or cygwin to configure and run make :P
[20:10:51 CEST] <JEEB> not exactly rocket science and msys2 and cygwin both have package managers
[20:11:03 CEST] <JEEB> also the lack of LGPL builds might be because usually only corporations want them
[20:11:24 CEST] <JEEB> or people otherwise doing closed source stuff
[20:11:24 CEST] <wolfman2000> JEEB: the (few) windows developers for a separate project like staying in visual studio. mingw support currently...doesn't work
[20:11:47 CEST] <wolfman2000> also, said project wants to stay MIT compatible. GPL would...well, force GPL if I understand the licensing terms right
[20:12:18 CEST] <JEEB> your source code would be still MIT, but the binary would become GPL. since MIT is compatible with GPL, that shouldn't be an issue
[20:13:14 CEST] <JEEB> the sources of the combined package with GPL'd FFmpeg would become GPL, but your project in general would be MIT
[20:13:27 CEST] <JEEB> the same thing happens with, say, ffms2
[20:13:52 CEST] <JEEB> where the thing itself is less limited but if built with a GPL FFmpeg the combined package should also be available as GPL
[20:14:44 CEST] <JEEB> but to be honest, if you just want to build FFmpeg with MSVC just setting up the build system for that on one machine isn't too bad IMHO :P You don't need mingw to build it, you just need a shell and make
[20:15:04 CEST] <JEEB> which will then run MSVC's compiler and linker
[20:15:37 CEST] <wolfman2000> then let's make sure I'm not unintentionally breaking policy. Previously, we enabled GPL building for general distribution. Source code stays MIT, binaries were GPL, license files...slightly disorganized mess, but working on it. For taking the engine to, say, a commercial game, just build without --enable-gpl and --enable-nonfree and we're good. Any flaws in my thinking?
[20:16:08 CEST] <JEEB> enable-nonfree in general means you can't distro it at all, so you can forget about that altogether
[20:16:21 CEST] <wolfman2000> we don't have --enable-nonfree enabled normally
[20:17:12 CEST] <JEEB> but yes, that is the general idea. the GPL-licensed package does not affect your general license. and the MIT license is GPL-compatible so it's possible to link it with GPL stuff.
[20:17:30 CEST] <JEEB> so just have your sources available as both with the binary, and FFmpeg as with GPL.
[20:17:40 CEST] <JEEB> and if someone builds it with LGPL FFmpeg then it will matter even less
[20:18:09 CEST] <JEEB> not sure if I'm explaining it well, but basically even if a binary of yours ends up being GPL, your whole project isn't
[20:18:42 CEST] <wolfman2000> source code is available on github. We state our code is MIT, all third party libraries we use are respective of their licenses. Trying to use submodules more, but that's in a test branch.
[20:18:43 CEST] <JEEB> you still have to release sources for it under GPL, but that shouldn't be an issue I think?
[20:18:47 CEST] <Mavrik> Maybe it's better to think in terms of what GPL says.
[20:18:59 CEST] <JEEB> (for that binary, basically)
[20:19:49 CEST] <wolfman2000> just explain the "so just have your sources available as both with the binary" part a little better and I should be good
[20:19:55 CEST] <JEEB> https://github.com/FFMS/ffms2/blob/master/COPYING
[20:19:59 CEST] <JEEB> something like this :P
[20:20:42 CEST] <JEEB> wolfman2000: if you have a binary release with GPL FFmpeg, then the user who receives that has the right to get the source code for that binary release as per GPL
[20:20:57 CEST] <JEEB> which is what is meant with "that release binary becomes GPL"
[20:21:11 CEST] <JEEB> that doesn't change your general license of course
[20:21:29 CEST] <JEEB> so if someone uses the library for closed source and LGPL FFmpeg is used, there is no issue with that
[20:21:48 CEST] <JEEB> (of course the LGPL license has to be followed)
[20:23:53 CEST] <wolfman2000> all users can get our source code on github. We do not make any patches to ffmpeg.
[20:23:55 CEST] <JEEB> but really, building an LGPL FFmpeg shouldn't be that hard with the MSVC compiler/linker :) but yes, you will not get rid of the reason to configure with a shell interpreter and running the build with ye olde make :)
[20:24:09 CEST] <JEEB> wolfman2000: whether or not you have patches doesn't matter
[20:24:46 CEST] <JEEB> LGPL = sources for the LGPL thing exactly, and the capability of the user to switch the library should he or she want that
[20:25:23 CEST] <JEEB> which basically means not much extra with shared linking and with static linking you'd have to share the object files
[20:26:00 CEST] <wolfman2000> so binary GPL is fine, request for code is already handled as long as github doesn't break down, and.........there's an actual distinction with static vs sharing I'm not aware of, isn't there?
[20:26:15 CEST] <JEEB> with GPL there really isn't
[20:26:27 CEST] <JEEB> with LGPL you just have to follow the general idea of "letting the user replace binary"
[20:26:35 CEST] <JEEB> which I noted there regarding static linking
[20:26:44 CEST] <JEEB> which is why almost nobody does static linking with LGPL components
[20:27:26 CEST] <wolfman2000> okay, there may be some things to improve on in the end.
[20:28:51 CEST] <JEEB> but yeah, building with GPL FFmpeg makes the _binaries_ created there unusable in closed source or other cases where LGPL is required
[20:29:03 CEST] <JEEB> otherwise it has no effect on your primary project
[20:30:01 CEST] <wolfman2000> that was mostly what I wanted to confirm. as far as cmake vs bash configure......well, consider that a challenge for the future. :)
[20:30:34 CEST] <JEEB> not sure how a cmake configuration would make anything simpler :/
[20:30:46 CEST] <JEEB> you would still have to run cmake in a command line and have the cmake tool available
[20:31:04 CEST] <wolfman2000> JEEB: we already do require cmake.
[20:31:08 CEST] <wolfman2000> even for windows
[20:31:33 CEST] <JEEB> yes, but I just don't see it any different :P it's a requirement for FFmpeg's build system (whatever OS or compiler etc. you're building it with)
[20:31:58 CEST] <JEEB> you can run configure and make from cmake if you really want to, which would then become things in your MSVS project or whatever
[20:32:07 CEST] <wolfman2000> windows users can use the GUI, which means no terminal needed. You know not all windows developers care to use terminal commands.
[20:32:37 CEST] <wolfman2000> some just love their Visual Studio too much
[20:33:03 CEST] <JEEB> well, as I said nothing stops you from calling those binaries and you can have other ways of making sure they have those
[20:33:06 CEST] <JEEB> also you need yasm
[20:33:11 CEST] <JEEB> for the hand-written assembly
[20:33:38 CEST] <wolfman2000> JEEB: --disable-asm is available. ;)
[20:33:46 CEST] <JEEB> you really don't want to use it lol
[20:33:55 CEST] <JEEB> esp. on x86(_64)
[20:34:01 CEST] <JEEB> it really kills your performance in a lot of places
[20:34:15 CEST] <wolfman2000> I never tested it. I currently require yasm or nasm to be available before building ffmpeg with my project.
[20:34:18 CEST] <JEEB> on ARM it *might* be less of an issue since there's just less asm
[20:34:34 CEST] <JEEB> but in general it's never good idea to disable it
[20:35:11 CEST] <JEEB> anyways, if you're going to do it with cmake I recommend you just wrap FFmpeg's build system with it rather than make a new one. the new one's bound to fail at some point :P
[20:35:24 CEST] <JEEB> just make it make sure that the user has specific binaries available in PATH and that should be it
[20:35:39 CEST] <wolfman2000> again, trying to not require windows users to need msys
[20:36:05 CEST] <JEEB> you literally need a couple of binaries and you need cmake and yasm already :P
[20:36:22 CEST] <JEEB> you are definitely not winning much by trying to play around FFmpeg's build system
[20:36:51 CEST] <JEEB> or well, playing around it is more or less OK, but in the way where you use it underneath
[20:37:05 CEST] <JEEB> I know you can do it with cmake because lachs0r does his cross-compilation of mpv with cmake :P
[20:37:22 CEST] <wolfman2000> Remind me to have a meeting with him then
[20:37:25 CEST] <JEEB> which basically builds everything required with their original build systems underneath
[20:38:00 CEST] <JEEB> and if you really want to have them be lazy you could even automate download of the shell interpreter, make and yasm from the cmake script :P
[20:38:28 CEST] <wolfman2000> lazy is one thing. invasive is another. :P
[20:39:21 CEST] <JEEB> basically my bottom line is that you really don't want to replace FFmpeg's build system since that's going to kill you real quickly regarding compatibility. but you *can* wrap it.
[20:39:38 CEST] <JEEB> and that should be less breaking
[20:40:28 CEST] <wolfman2000> ...alright. Guess what I have to figure out then is how to get configure scripts to work in mingw without cmake errors.
[20:44:40 CEST] <JEEB> looking at lachs0r's thing he seems to at least be utilizing a module called ExternalProject
[20:44:57 CEST] <JEEB> you can poke him more on #mpv if you want to
[20:45:02 CEST] <JEEB> https://github.com/lachs0r/mpv-build-cmake/blob/master/CMakeLists.txt
[20:45:57 CEST] <wolfman2000> JEEB: I'm familiar with externalproject_add. What was throwing me off was HOW to run it cleanly with......huh? I don't recognize the ${EXEC} command.
[20:47:07 CEST] <wolfman2000> but if it does what I think it does, that may have solved some of my issues.
[20:52:10 CEST] <wolfman2000> ...okay, maybe not. I'll figure something
[21:14:59 CEST] <foreverska> Anyone have experience with AMD's encoder?
[21:20:05 CEST] <foreverska> Also edge case solution:  Has anyone ever paired a GTX card with an A10 processor and used both NVENC and VCE at the same time?
[22:08:28 CEST] <IRCMonkey> Hello, is there a way for ffmpeg to output a warning if frames are dropping during a stream?
[22:49:10 CEST] <YaMoonSun> Is there an issue extracting jpeg images from .mkv files? I made a bat file that says ffmpeg -i "Where are you.mkv" -r 30 "Where Are You Now"-%4d.jpeg and it outputs one image.
[22:49:53 CEST] <c_14> Does it output -0000.jpeg ? -%4d.jpeg?
[22:49:56 CEST] <klaxa> hmm... does %4d work?
[22:50:08 CEST] <klaxa> i always use it as %04d
[22:50:21 CEST] <c_14> ^I think you have to
[22:50:25 CEST] <YaMoonSun> Yeah, it worked last night on some lightning storm videos.
[22:50:32 CEST] <YaMoonSun> They were in .mov though
[22:51:57 CEST] <YaMoonSun> http://pastebin.com/89JsbdhZ
[22:52:27 CEST] <c_14> There's no %
[22:52:46 CEST] <YaMoonSun> There is in the .bat
[22:52:57 CEST] <c_14> There isn't in that line
[22:53:19 CEST] <YaMoonSun> So I have to do it manually because the .bat file is being a.. lousy.
[22:53:21 CEST] <YaMoonSun> Thanks
[22:53:24 CEST] <YaMoonSun> Smh
[22:53:32 CEST] <c_14> Might need to escape the '%' in bat files
[22:53:34 CEST] <c_14> *shrug*
[22:54:23 CEST] <YaMoonSun> Worked via cmd without a .bat
[22:54:25 CEST] <YaMoonSun> So weird.
[22:57:37 CEST] <YaMoonSun> Some really awesome fan art in that video, nice to actually be able to see it (y)
[00:00:00 CEST] --- Mon Oct  5 2015

More information about the Ffmpeg-devel-irc mailing list