[Ffmpeg-devel-irc] ffmpeg.log.20191107

[burek]

[00:28:37 CET] <Atlenohen> hello, is there some kind of a list what external packages are required for ffmpeg build?
[00:29:46 CET] <DHE> relatively little is required. you just won't get the features that come with it. no openssl? http won't support ssl, etc. now there's what I'd call a recommended list...
[00:29:46 CET] <cehoyos> None should be required
[00:30:08 CET] <cehoyos> zlib and bzlib are needed for some but not many files
[00:30:37 CET] <Atlenohen> I want all the features, not that it's needed, but I don't want this silently continue. Is there a flag to break compilation if somethings missing?
[00:30:48 CET] <cehoyos> No, you have to enable them
[00:31:00 CET] <cehoyos> Note that it is not possible that all optional features work on your system
[00:31:46 CET] <Atlenohen> No I mean, the ones that I enable in configure, I if there's something missing and MPEGTS can't be used for example, ffmpeg's compilation should tell me, will it do that?
[00:33:57 CET] <nicolas17> you can't enable "all" the features, some are platform-specific
[00:35:23 CET] <AmyMalik> if someone wants to enable all the features
[00:35:30 CET] <AmyMalik> they mean they want to enable all the applicable features
[00:36:33 CET] <Atlenohen> I didn't meant that, I already have the feature set known and selected for windows, from the things I selected, will ffmpeg warn me if anything's missing. I don't like the sound of "if it's not present on system it will just not use it"
[00:37:00 CET] <Atlenohen> I consider that an error, and I hope I would see a warning or an error message. Can that be enforced?
[00:38:12 CET] <nicolas17> I just tried ./configure --enable-libaom
[00:38:33 CET] <nicolas17> and it said "ERROR: aom >= 1.0.0 not found using pkg-config" and didn't continue
[00:38:55 CET] <Atlenohen> Oh okay
[00:40:11 CET] <AmyMalik> canister canister canister canister canister canister canister canister canister canister
[01:07:08 CET] <Atlenohen> So would it build with these packages? Probably not but what do you guys see that may be missing? Ofcourse I'm going to just run the compiler and perhaps it'll just throw me a list, or it will keep breaking 1 by 1 which may take a long time.
[01:07:09 CET] <Atlenohen> https://pastebin.com/cVxKVakQ
[01:24:10 CET] <cehoyos> Except zlib and bzlib I don't think you will need any of these
[01:24:56 CET] <cehoyos> gnutls can be very helpful, needs --enable-gnutls
[03:05:43 CET] <Atlenohen> cehoyos: notice there is no make on that list ;)
[03:21:22 CET] <DHE> well if you're going to be that pedantic, there's no gcc either
[03:22:44 CET] <pink_mist> pretty sure you need a libc too
[09:21:09 CET] <wowa_putin> test
[09:21:21 CET] <wowa_putin> does anyone see my messages?
[09:22:28 CET] <another> yes
[09:22:36 CET] <wowa_putin> hello everybody. I work on app that heavily uses INTEL TBB and want ffmpeg to use also the same TBB's thread pool. The only relevant links google gave me: https://stackoverflow.com/questions/6049798/ffmpeg-mt-and-tbb and https://software.intel.com/en-us/forums/intel-threading-building-blocks/topic/283743#comment-1491456 . I followed steps described but it didn't work out. After that i digged breifly into _ff_mpeg sources code and it appeared that
[09:22:36 CET] <wowa_putin> avcodec_open2()->ff_frame_thread_encoder_init() somehow just overwrite my AVCodecContext->execute and AVCodecContext->execute2(). Maybe i do something wrong? Or just SO's answer is outdated for current ffmpeg version?
[10:00:58 CET] <AlexApps> Hello, I'm making a script to automatically edit a video and I need to perform two FFmpeg commands to finish editing the video
[10:01:20 CET] <AlexApps> Currently I encode it fully the first time and save it to disk, and then do the same the second time
[10:01:40 CET] <AlexApps> is there any way to pipe the raw video from the first command into the input of the second?
[10:01:57 CET] <AlexApps> It would save a lot of time and disk resources
[10:12:28 CET] <AlexApps> Suppose I'll ask later
[10:16:56 CET] <Lypheo> why do you need to encode it twice?
[10:20:29 CET] <wowa_putin> no any ideas on my question?
[11:19:57 CET] <Atlenohen> I have a problem with building ffmpeg on windows now
[11:20:42 CET] <Atlenohen> seems like after updating VS a couple of times lately, and updating fmpeg to 4.2.1, I doesn't look for libx264
[11:22:21 CET] <Atlenohen> I had it all figured out, even had to make fixes to the configure file, I later found the places link.exe is looking for libx264.lib, and how to adjust he ldflags and stuff
[11:25:44 CET] <Atlenohen> but now I see that link.exe isn't even looking for libx264.lib, in all the cases when it's looking inside the supplied lib path, it looks for some other files
[11:25:58 CET] <Atlenohen> such as kernel, libcmt, oldnames, etc
[11:26:34 CET] <Atlenohen> https://i.imgur.com/NQJXZJ5.png
[11:31:29 CET] <Atlenohen>  ./dependency/x264/include\x264.h(46): fatal error C1083: Cannot open include file: 'x264_config.h': No such file or directory
[11:31:45 CET] <Atlenohen> interesting, i guess it's needed for ffmpeg
[11:36:49 CET] <Atlenohen> ah now it worked, it started searching for libx264 only after I gave it all the headers, even tho there's nothing in that file that would make a difference technically
[11:37:33 CET] <Atlenohen> these tests should be improved and error reporting as well, not that one can't look at config.log but to make things easier on windows, I might offer patch myself
[11:38:54 CET] <Atlenohen> x86 assembler             nasm
[11:39:10 CET] <Atlenohen> but it says ffmpeg needs yasm? I have both installed since x264 needs nasm
[11:41:50 CET] <c_14> nasm should be fine as long as it's new enough
[11:42:07 CET] <c_14> and I believe yasm is no longer actively developed in favor of nasm, but don't quote me on that
[11:47:43 CET] <Atlenohen> https://github.com/FFmpeg/FFmpeg/commit/4f9297ac3b39098547863d28fbc8d2a906d5be49
[11:47:58 CET] <Atlenohen> I guess the docs should be updated
[11:49:05 CET] <Atlenohen> But I'd rewrire the whole windows compilation part
[11:49:17 CET] <Atlenohen> rewrite*
[11:50:23 CET] <Atlenohen> I'd doing my own notes on the process so it may come useful for the online version as well
[14:05:46 CET] <another> "but don't quote me on that" --- c_14
[15:24:51 CET] <GenTooMan> hmm I kind of like to think of it as rewire when it comes to windows compilation :D
[15:32:22 CET] <Atlenohen> libavutil -> avconfig.h and ffversion.h have LF instead of CRLF line endings when compiling on windows(msys2,mingw64)
[16:02:18 CET] <montana> why does so many programs have problem displaying proper duration of .vob files (including ffmpeg)
[16:07:17 CET] <another> does it?
[16:10:05 CET] <montana> another here:  https://x0.at/--7.png
[16:11:15 CET] <montana> another did you see the picture
[16:12:51 CET] <another> DVDs are a bit tricky
[16:12:59 CET] <montana> another did you see the picture?
[16:13:42 CET] <another> because they are split into multiple VOBs
[16:13:42 CET] <DHE> well the DVD is supposed to be the concatenation of all the VOBs for the particular title
[16:14:42 CET] <another> titles can also start in the middle of a vob, can#t they?
[16:14:46 CET] <furq> yes
[16:15:03 CET] <furq> you can't get the correct duration without reference to the ifo
[16:15:31 CET] <montana> did you people see the picture
[16:15:39 CET] <furq> no
[16:15:42 CET] <montana> https://x0.at/--7.png
[16:16:39 CET] <montana> if you see the picture, one program correctly displays the duration: how is that possible then. if
[16:16:43 CET] <montana> if you see the picture, one program correctly displays the duration: how is that possible then. if   <furq> you can't get the correct duration without reference to the ifo"
[16:17:00 CET] <DHE> if one file holds multiple titles, yeah that would easily explain it. the end of the file represents a different title and the timestamp could be earlier than the start of the file
[16:17:20 CET] <montana> then how does mediainfo able to display the correc time
[16:17:38 CET] <montana> if you see the picture
[16:18:26 CET] <another> on a dvd there are index files (.ifo) which have to be evaluated as well
[16:18:40 CET] <another> ffmpeg doesn't handle that
[16:19:02 CET] <montana> then how does mediainfo able to display the correct time without the IFO
[16:19:22 CET] <another> maybe they read the IFO?
[16:20:14 CET] <montana> another it's not
[16:20:26 CET] <montana> i just opened ONLY .vob file with mediainfo
[16:23:57 CET] <jose1711> hello, i am getting illegal instruction and coredump right after running ffmpeg on my test freebsd machine. is there a way to find out more what's causing the issue?
[16:29:06 CET] <kepstin> jose1711: well, if you're getting an illegal instruction, then the ffmpeg was probably compiled targetting a newer cpu than you have, or possibly was compiled with runtime cpu detection for the assembly instructions disabled.
[16:30:03 CET] <jose1711> hm, i recompiled with runtime cpu detection disabled to no cure
[16:30:42 CET] <DHE> you got -march set in your cflags?
[16:31:02 CET] <DHE> or other -m parametrs
[16:33:50 CET] <jose1711> trying to find out (i am new to freebsd and ports)
[16:34:15 CET] <jose1711> CFLAGS='-O2 -pipe  -DLIBICONV_PLUG -fstack-protector-strong -isystem /usr/local/include -fno-strict-aliasing '
[16:34:51 CET] <jose1711> that's from ffbuild/config.log
[16:57:58 CET] <wowa_putin> Guys, cmonn, give me plz advices how to rewrite pthread_frame.c to make ffmpeg use INTEL TBB
[20:21:32 CET] <Atlenohen> LNK2019	unresolved external symbol ff_golomb_vlc_len referenced in function get_ue_golomb_31	Dolphin	C:\GitDEV\GitRepos\Github\dolphin\Source\Core\DolphinQt\avformat.lib(mxfenc.o)
[20:22:07 CET] <Atlenohen> something about mxf container?
[20:22:47 CET] <Atlenohen> I did not get this when doing 4.1.3, but I get this now when trying to compile static ffmpeg libs from version 4.2.1
[20:23:05 CET] <Atlenohen> Not the ffmpeg it self but the project that uses it.
[21:10:46 CET] <`mist> hey guys. I'm trying to make a timelapse of a bunch of jpegs but it seems i might have gotten some of my flags wrong
[21:11:08 CET] <`mist> ffmpeg -hwaccel cuvid -c:v h264_cuvid -framerate 24 -pattern_type glob -i '/home/mist/cam_renamed/*.jpeg' -s:v 1920x1080 -vf 'format=nv12,hwupload' -c:v h264_nvenc timelapse.mkv
[21:11:25 CET] <`mist> Error while opening decoder for input stream #0:0 : Operation not permitted
[21:11:27 CET] <`mist> so i run it with sudo
[21:11:52 CET] <`mist> [image2 @ 0x55e6bb0af0c0] Could not open file : /home/mist/cam_renamed/*.jpeg[image2 @ 0x55e6bb0af0c0] Could not open file : /home/mist/cam_renamed/*.jpeg[image2 @ 0x55e6bb0af0c0] Could not find codec parameters for stream 0 (Video: h264, none): unspecified size
[21:12:52 CET] <DHE> ffmpeg works on the commandline style of:   ffmpeg {[options for file]  [input or output file]} {...repeat...}
[21:13:06 CET] <DHE> so you've told it that the input video can be decoded as h264, which is incorrect
[21:13:26 CET] <`mist> aha...
[21:14:50 CET] <DHE> honestly I think you can simplify it down to: ffmpeg  -framerate 24 -pattern_type glob -i '/home/mist/cam_renamed/*.jpeg' -s:v 1920x1080  -c:v h264_nvenc timelapse.mkv
[21:16:18 CET] <`mist> i know that hwencodign usually makes the quality go bad. Are there any presets i could change to increase the quality?
[21:17:02 CET] <`mist> (oh and the encoding is running right now, your command works thanks!) almost halfway through the video atm
[21:21:56 CET] <DHE> well you didn't specify any quality options so you get defaults. I'll bet they're not very good. I'd set a bitrate or qp value.
[21:22:21 CET] <DHE> ffmpeg -h encoder=h264_nvenc   # for options specific to the nvidia encoder. other generic options like -b:v still apply as well.
[21:23:29 CET] <`mist> any recomendations? (I'm completely new to this kind of stuff)
[21:27:14 CET] <DHE> it's hard to give a recommendation without being able to see it. but slide shows tend to not need much bandwidth. for nvenc I'd try -preset slow and for a slide show maybe try -qp 25 and see how it turns out (both image quality and size)
[21:28:39 CET] <`mist> thanks! I'm trying to make a timelapse of the building of my house
[21:33:13 CET] <`mist> quality is a bazillion times better but the file is 4gb =D  I don't mind though
[21:39:21 CET] <DHE> oh... yeah it's a time lapse not a slide show. I'd raise the qp to 32 or maybe even higher.
[21:39:47 CET] <DHE> sadly time lapses tend to be bad for motion compensation codecs since there's so much that changes at once between images
[21:39:59 CET] <DHE> (I've done something similar)
[22:51:27 CET] <Atlenohen> what if the ffmpeg's windows build process forgets to output golomb.h ?
[22:53:10 CET] <Atlenohen> it's in /ibavcodec
[22:53:38 CET] <Atlenohen> but not inside the make install output
[22:56:39 CET] <BtbN> that's not a public header
[23:20:18 CET] <Atlenohen> oh
[23:21:28 CET] <Atlenohen> BtbN,  ff_golomb_vlc_len happens to be inside there, that's the only place it is, i grepped whole ffmpeg repo
[23:33:16 CET] <kepstin> Atlenohen: my guess is that you have a link order issue, libavformat is referencing a symbol defined in libavcodec.
[23:35:28 CET] <Atlenohen> don't know much about link order, you mean in the program's includes, or the way things get added in VS like the "additional lib" I think it has always been alphabetic
[23:35:34 CET] <Atlenohen> or it has to be strictly?
[23:35:46 CET] <kepstin> i don't know how link order in windows works
[23:35:51 CET] <Atlenohen> oh
[23:36:06 CET] <kepstin> in linux, with static libs, you have to put them in a specific order based on which libraries use symbols from which other libraries
[23:37:08 CET] <Atlenohen> I pretty much just compiled stock 4.2.1 with custom options the same I had with 4.1.3
[23:42:44 CET] <DHE> rm ffmpeg_g ; make ffmpeg_g V=1   # and write down the commandline used. :)
[23:43:50 CET] <Atlenohen> As additional libraries only the whole library folder is used, for the main projects there is dependencie where they're defined one by one, and it's alphabetic, first it's avcoded then avformat
[23:44:02 CET] <Atlenohen> at least the order in the string
[23:44:19 CET] <Atlenohen> DHE: for me?
[00:00:00 CET] --- Fri Nov  8 2019