[Ffmpeg-devel-irc] ffmpeg.log.20161003

[burek]

[00:43:23 CEST] <SchrodingersScat> stah0121: http://ronja.twibright.com/pitchotron.php
[01:42:34 CEST] <ferdna> why does my image looks like this:
[01:42:35 CEST] <ferdna> http://i66.tinypic.com/afjwvl.jpg
[01:42:42 CEST] <ferdna> that is on my tv... on burnt dvd...
[01:42:46 CEST] <ferdna> on my dvd player
[02:13:59 CEST] <SchrodingersScat> ferdna: very modern art
[02:14:08 CEST] <ferdna> SchrodingersScat, lol... yes
[03:42:15 CEST] <mrelcee> having a curious x11 issue..   is there an X11 channel here or does anyone mind me asking it here?  (yes it's on freebsd :)
[03:46:22 CEST] <mrelcee> wrong channel too. soddy
[04:02:01 CEST] <waltz> hey folks! i want to loop a video for 30 seconds. how would i do that?
[05:21:01 CEST] <eduardo> hi
[05:22:33 CEST] <eduardo> oh well, i have a problem installing the ffmpeg in linux mint
[05:23:46 CEST] <eduardo> i followed his guide here: https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu
[05:25:06 CEST] <eduardo> and whe i get to the part when i should install fmpeg
[05:26:52 CEST] <ferdna> eduardo,
[05:27:08 CEST] <SchrodingersScat> now two people are doing it ;_;
[05:27:58 CEST] <SchrodingersScat> eduardo: what you absolutely must do is
[05:30:37 CEST] <eduardo> i get this error message: ERROR: x265 not found using pkg-config
[05:31:04 CEST] <eduardo> im new on linux yet so i dont know what that means
[05:32:37 CEST] <noobe1> hello, I'd like to use ffmpeg to decompress context of an mp4 video file into individual raw yuyv422 format.  I tried ffmpeg -i video.mp4 -c:v rawvideo -pix_fmt yuyv422 out%d.dat but it outputs to a single file.
[05:32:55 CEST] <eduardo> the line that returns that message is "PATH="$HOME/bin:$PATH" PKG_CONFIG_PATH="$HOME/ffmpeg_build/lib/pkgconfig" ./configure \
[05:32:55 CEST] <eduardo>   --prefix="$HOME/ffmpeg_build" \
[05:32:55 CEST] <eduardo>   --pkg-config-flags="--static" \
[05:32:55 CEST] <eduardo>   --extra-cflags="-I$HOME/ffmpeg_build/include" \
[05:32:55 CEST] <eduardo>   --extra-ldflags="-L$HOME/ffmpeg_build/lib" \
[05:32:57 CEST] <eduardo>   --bindir="$HOME/bin" \
[05:32:59 CEST] <eduardo>   --enable-gpl \
[05:33:01 CEST] <eduardo>   --enable-libass \
[05:33:03 CEST] <eduardo>   --enable-libfdk-aac \
[05:33:05 CEST] <eduardo>   --enable-libfreetype \
[05:33:07 CEST] <eduardo>   --enable-libmp3lame \
[05:33:09 CEST] <eduardo>   --enable-libopus \
[05:33:13 CEST] <eduardo>   --enable-libtheora \
[05:33:15 CEST] <eduardo>   --enable-libvorbis \
[05:33:17 CEST] <eduardo>   --enable-libvpx \
[05:33:19 CEST] <eduardo>   --enable-libx264 \
[05:33:21 CEST] <eduardo>   --enable-libx265 \
[05:33:23 CEST] <eduardo>   --enable-nonfree"
[05:37:31 CEST] <relaxed> eduardo: use http://pastebin.com/ , don't paste in the channel. Show the end of ffmpeg's config.log
[05:37:50 CEST] <eduardo> oh sorry
[05:38:04 CEST] <eduardo> how do i do that?
[05:40:22 CEST] <relaxed> go to the website and paste the info, click "create new paste", then paste the url here
[05:40:56 CEST] <eduardo> mmm sorry i meant the config.log
[05:41:10 CEST] <eduardo> nvm i got a tutorial
[05:41:22 CEST] <eduardo> when i get it ill use the website
[05:43:23 CEST] <eduardo> i thin i don't ave a confi.lo on ffmpeg because i didnt finished intalling it
[05:44:02 CEST] <eduardo> as i wrote, i got an error message when i input that lin of comand on th terminal
[05:44:57 CEST] <eduardo> this is the line of command: http://pastebin.com/uNNGzVJb
[05:46:40 CEST] <SchrodingersScat> eduardo: did you do this bit? https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#libx265
[05:46:49 CEST] <relaxed> running ffmpeg's ./configure creates config.log
[05:47:38 CEST] <SchrodingersScat> noobe1: out%02d.dat ?
[05:47:48 CEST] <relaxed> it's in the ffmpeg source dir
[05:48:54 CEST] <eduardo> yes i did.
[05:49:00 CEST] <noobe1> SchrodingersScat, I tried, it's not outputing individual files, it's outputing a single file named "out%02d.dat"
[05:49:08 CEST] <ferdna> eduardo, use pre-compiled packages...
[05:49:14 CEST] <ferdna> apt-get install ffmpeg
[05:49:20 CEST] <ferdna> apt-get install x264
[05:49:22 CEST] <ferdna> apt-get install x265
[05:50:01 CEST] <noobe1> is the raw file just yuv frames back-to-back? I mean I guess I can calculate actual nubmer of bytes per raw frame using format+resolution and just read the file continously
[05:50:18 CEST] <noobe1> but I wanted to load individual frames into photoshop
[05:51:27 CEST] <eduardo> the x264 is fine
[05:51:54 CEST] <eduardo> but ffmpeg and x265 are mising form the repositories list
[05:52:02 CEST] <eduardo> :S
[05:52:14 CEST] <relaxed> eduardo: https://www.johnvansickle.com/ffmpeg/
[05:52:57 CEST] <eduardo> oh and relaxed: i couldnt rin ./configure on ffmpeg, shoots me the error i just described
[05:53:34 CEST] <relaxed> error or not, config.log is still created
[05:54:30 CEST] <eduardo> i see, let me look for it
[05:56:14 CEST] <relaxed> noobe1: I think you want out%04d.Y
[05:58:35 CEST] <relaxed> eduardo: cd to the ffmpeg source dir and run the following:  tail -n 50 config.log | curl -F 'sprunge=<-' http://sprunge.us
[05:59:38 CEST] <eduardo> i found it already, ill pass the pased copy in a sec
[06:00:40 CEST] <eduardo> here it is
[06:00:42 CEST] <eduardo> http://pastebin.com/PPMFFgmC
[06:02:47 CEST] <eduardo> http://sprunge.us/gMTc
[06:03:33 CEST] <ferdna> eduardo, no they are not..
[06:03:36 CEST] <noobe1> I just tried it, it still outputs single file file (named out%04d.Y"
[06:03:40 CEST] <ferdna> make sure your mirrors are up to date...
[06:04:35 CEST] <noobe1> my full command line is: ffmpeg.exe -i video.mp4 -c:v rawvideo -f rawvideo -pix_fmt yuyv422 out%04d.Y
[06:05:10 CEST] <relaxed> remove -f rawvideo
[06:05:22 CEST] <noobe1> damn, that does it!! Thanks a bunch!!
[06:06:15 CEST] <eduardo> maybe  need to add some mirrors??? i just installed this distro
[06:06:32 CEST] <relaxed> eduardo: did you install libx265-dev?
[06:07:08 CEST] <eduardo> i think so, let me double check
[06:09:00 CEST] <eduardo> i installed libx265
[06:10:57 CEST] <eduardo> i used this command sequence: http://pastebin.com/3DhgNkKP
[06:23:21 CEST] <noobe1> relaxed, I ran an experiment, took a test.mp4 files (with a ball moving horizontally across screen), dumped the mp4 to video.yuv raw.  Then took the video.yuv and converted it back to test.mp4.  Now test.mp4 does not play as smooth as original file (ball stutters at times) Any ideas? Here is 2nd command I used: ffmpeg -f rawvideo -pix_fmt yuyv422 -s:v 1920x1080 -r 30 -i video.yuv -c:v libx264 output.mp4
[06:31:33 CEST] <noobe1> never mind, it was probably a computer loading from disk glitch...playing the file over and over is smooth. Thanks.
[10:09:54 CEST] <Evert___> Hello!
[10:10:49 CEST] <Evert___> I need some help. I'm building an app for a client and that app needs ffmpeg with its non-free features. So my client will need to pay ffmpeg license costs, but I can't find anywhere how much that is.
[10:11:22 CEST] <Evert___> Can anybody tell me where I can find that? My client needs to know to decide whether they want to move forward with ffmpeg or try another strategy.
[10:11:47 CEST] <furq> that's not what non-free means
[10:12:07 CEST] <furq> non-free means the binary is no longer gpl compatible
[10:12:47 CEST] <Evert___> which means what exactly? I thought that meant that the binary requires another license, a payed license.
[10:13:26 CEST] <Evert___> On this page (http://www.ffmpeg.org/legal.html) it says:
[10:13:34 CEST] <Evert___> Q: Is it perfectly alright to incorporate the whole FFmpeg core into my own commercial product?  A: You might have a problem here. There have been cases where companies have used FFmpeg in their products. These companies found out that once you start trying to make money from patented technologies, the owners of the patents will come after their licensing fees. Notably, MPEG LA is vigilant and diligent about collecting for MPEG-relate
[10:13:44 CEST] <furq> yeah that's distinct from nonfree
[10:14:00 CEST] <furq> e.g. x264 only requires --enable-gpl, but you need a license to encode h.264
[10:14:13 CEST] <furq> certainly to do that in a commercial product
[10:14:34 CEST] <furq> nonfree just means you'll be violating the gpl if you distribute binaries
[10:14:37 CEST] <Evert___> and that license comes from ffmpeg? or mpeg itself?
[10:14:48 CEST] <furq> from whomever issues the license for h264 these days
[10:14:53 CEST] <furq> i think it's mpeg-la
[10:14:58 CEST] <furq> ffmpeg doesn't issue any licenses
[10:15:24 CEST] <Evert___> hmm okay, I'm slowly starting to get it
[10:15:49 CEST] <Evert___> I'm a little but confused about one thing, what's the difference between x264 and h.264?
[10:15:56 CEST] <furq> x264 is an h264 encoder
[10:16:07 CEST] <Evert___> ah oké
[10:16:18 CEST] <Evert___> *okay (oké is my language)
[10:16:35 CEST] <Evert___> Thank you
[10:16:41 CEST] <furq> i'm not sure what you'd do if you need nonfree components
[10:18:17 CEST] <furq> if your client isn't distributing binaries then i guess it's not an issue
[10:18:44 CEST] <furq> if they are then i don't know how you'd make that gpl compliant
[10:19:23 CEST] <Evert___> To give you an idea of what we're doing: we have an app for iOS and Android and in that app users can upload a video to our server. Then other users can play that video. But we found out that Android users can
[10:20:07 CEST] <Evert___> *Android users can't play all the videos coming from iOS. So we were planning on converting all the videos on the server after uploading, to a format that both operating systems can play.
[10:20:15 CEST] <Evert___> Which would be h.264
[10:20:30 CEST] <furq> if you're just using ffmpeg on the backend then you don't need to care about gpl compliance (afaik)
[10:20:45 CEST] <furq> you'll probably need to care about licensing h264 though
[10:20:51 CEST] <Evert___> Are you sure? Because we're still using it for commercial reasons.
[10:21:05 CEST] <Evert___> Yes okay, that's what I meant
[10:21:32 CEST] <Evert___> Okay, then I'll try to find out who we need to pay for h264 and how much.
[10:22:02 CEST] <furq> nonfree is used for libraries which have a gpl-incompatible license, which means you can't distribute binaries
[10:22:46 CEST] <Evert___> Yes I think I get that now :-)
[10:22:47 CEST] <furq> if you're not doing that then it should be fine
[10:23:45 CEST] <furq> you'll probably also need a license for aac
[10:23:56 CEST] <furq> those are issued by via afaik
[10:23:57 CEST] <Evert___> we're using mp3
[10:24:00 CEST] <furq> oh
[10:24:02 CEST] <furq> never mind then
[10:24:12 CEST] <Evert___> does that have no license costs?
[10:24:18 CEST] <furq> i don't know about mp3
[10:24:21 CEST] <furq> most of the patents have expired now
[10:24:26 CEST] <Evert___> cool
[10:24:31 CEST] <furq> you'll obviously want to check though
[10:24:36 CEST] <Evert___> Too bad that the question "how much do I need to pay for h264?" isn't easy to answer apparently
[10:24:45 CEST] <Evert___> yes of course
[10:24:46 CEST] <furq> yeah codec licensing is a pain
[10:25:00 CEST] <furq> just be glad you're not using h265, there are three different licensing bodies for that
[10:25:06 CEST] <furq> and i think you need all three licenses
[11:14:47 CEST] <Spring> is there any downside to using -static-thresh to '0' for regular, moving videos?
[11:15:28 CEST] <Spring> by moving I mean videos with a standard amount of movement in the scene
[11:16:22 CEST] <Spring> using libvpx
[13:31:57 CEST] <Heniek> Hi guys. If I am using only decoding part of ffmpeg is it possible for me to get to the original data of a frame? I mean, I am receiving a h264 stream and I would like to save its contents to disk, does any pointer in AVFrame contains original data?
[13:34:27 CEST] <nonex86> AVPacket contains original data
[13:34:46 CEST] <nonex86> if you are using ffmpeg api for demuxing
[13:35:06 CEST] <nonex86> AVFrame is a result of decoding original data
[13:36:03 CEST] <nonex86> if you are using any other api to get the real data, guess you already have it :)
[13:38:14 CEST] <Heniek> thank you:)
[14:07:02 CEST] <hamidi> hi
[14:07:28 CEST] <hamidi> please give me a way to generate or download FFmpegInvoke.cs
[14:07:50 CEST] <hamidi> i'm going to change the version of ffmpeg in our project to the latest.
[14:08:01 CEST] <hamidi> i replaced ffmpeg dll's
[14:08:20 CEST] <hamidi> but it seems that it's required to replace FFmpegInvoke.cs too.
[14:08:27 CEST] <hamidi> i don't know much about it though
[14:14:14 CEST] <hamidi> no one knows?
[14:47:40 CEST] <nonex86> this is some kind of ffmpeg interop wrapper?
[14:47:47 CEST] <nonex86> why dont you just google this name?
[16:21:04 CEST] <bornpilot> Greetings I was looking the difference between the video file in AVCHD and mp4 files both were h264 compresssion using a JVC camcorder
[16:21:21 CEST] <bornpilot> I used the following command to get the stream information.
[16:21:42 CEST] <bornpilot> ffmpeg -i file.mts and ffmpeg -i file.mp4
[16:22:02 CEST] <bornpilot> here were the difference in the stream info about h264
[16:22:03 CEST] <bornpilot> h264 (high) (avc1 / 0x31637661)
[16:22:04 CEST] <bornpilot> h264 (high) (HDPR / 0x332D4341)
[16:22:20 CEST] <bornpilot> my question is what is avc1 and HDPR?
[16:23:28 CEST] <bornpilot> I searched through the ITUs document on h264 and could not find avc1 nor HDPR.
[16:23:52 CEST] <bornpilot> avc1 came from the mp4 file and HDPR came from the AVCHD file
[16:42:12 CEST] <c_14> That's the codec tag
[16:45:01 CEST] <bornpilot> c_14 so h264 is the compression and HDPR/avc1 would be the codec?
[16:45:19 CEST] <c_14> no
[16:45:43 CEST] <ozette> is there any way to make ffmpeg -ss operate faster on large files?
[16:46:29 CEST] <Spring> ozette, which order do you have the command in within the full command?
[16:46:46 CEST] <c_14> h264 is the codec (and stands for mpeg-4 part 10/mpeg-4 avc) and HDPR/avc1 is the tag, much like xvid is a tag for mpeg-4 part 2 video <- bornpilot
[16:47:23 CEST] <c_14> ozette: https://trac.ffmpeg.org/wiki/Seeking
[16:47:34 CEST] <bornpilot> thanks c_14
[16:48:18 CEST] <Spring> the wiki doesn't mention it but there's also the trim filter
[16:48:25 CEST] <ozette> Spring: i can't see the full command at this moment, i will have to log it in my application, but it starts like ffmpeg -ss 162 -i file.mp4
[16:48:38 CEST] <Spring> so before the input, right.
[16:48:43 CEST] <ozette> yes
[16:49:48 CEST] <Spring> should be the fastest method using -ss. Could try the trim filter, alternatively. Not sure speed wise how it compares.
[16:50:21 CEST] <Spring> I'd imagine since it's a filter it would be slower though
[16:50:32 CEST] <bornpilot> c_14 so would it be correct that avc1 and HDPR are two subtypes withing h264?
[16:50:37 CEST] <bornpilot> ^within
[16:50:53 CEST] <ozette> i must say that i run it on a machine with 1gb ram, but 'top' shows me the command is only taking 5% of the mem
[16:51:23 CEST] <kepstin> yeah, trim filter is slower since it decodes the whole stream, and just throws away frames that don't match the condition. But on some files where seeking doesn't work well, trim is handy to have.
[16:51:28 CEST] <ozette> but really it has taken 94 minutes already to generate 20 thumbnails for a 5Gb video file
[16:51:33 CEST] <ozette> and it's still not done
[16:52:08 CEST] <Spring> geez
[16:53:06 CEST] <ozette> yes..
[16:53:49 CEST] <kepstin> bornpilot: better to say that "avc1 and HDPR are two different ways of storing h264 video in an mp4/iso/mov type container"
[16:54:40 CEST] <kepstin> bornpilot: those tags are defined by the container specifications, not the codec specs.
[16:55:19 CEST] <c_14> ozette: what codec is the input?
[16:55:38 CEST] <bornpilot> I see, so they dont refelct on the how of the encoding proecess but the the container storage mechanisim?
[16:56:20 CEST] <ozette> mpeg4
[16:56:49 CEST] <Spring> that doesn't sound right, unless it's a /really/ slow system or the command has been interrupted (I've had this happen where pressing a key in the command prompt has caused ffmpeg to stall indefinitely)
[16:57:16 CEST] <ozette> i've had this before as well, but i'm quite sure the command will return
[16:57:48 CEST] <c_14> Something must be weird with the command it's using.
[16:57:53 CEST] <ozette> it's part of an application, which is supposed to go to production, and two video were uploaded, one of 1.5Gb and 5Gb
[16:58:06 CEST] <ozette> the 1.5Gb took about 1 hour
[16:58:44 CEST] <c_14> ozette: check top/ps to get the full commandline it's using
[16:58:46 CEST] <ozette> the command is generated by some library and from what i remember, it's pretty standard, i'd have to log the command again to see what it's doing
[16:59:16 CEST] <ozette> c_14: the command is cut off by the screen
[16:59:39 CEST] <ozette> any other way to see it?
[16:59:54 CEST] <furq> ozette: if you're generating thumbnails you probably want -skip_frame nokey -vf thumbnail
[17:00:00 CEST] <Spring> you could use the -report option to log to file, AFAIK it includes the command used
[17:00:15 CEST] <ozette> furq: interesting
[17:00:16 CEST] <furq> you'll definitely want -skip_frame nokey because it makes decoding much faster
[17:00:24 CEST] <c_14> ozette: cat /proc/$pid/cmdline
[17:00:33 CEST] <furq> -vf thumbnail will potentially use a ton of memory
[17:00:45 CEST] <ozette> what does -vf do
[17:00:50 CEST] <furq> video filter
[17:00:51 CEST] <ozette> c_14: will try
[17:00:53 CEST] <ozette> ah
[17:01:05 CEST] <furq> thumbnail is a bit awkward to use but it gives good results
[17:02:17 CEST] <ozette> nice, will share the cmdline in a second
[17:07:37 CEST] <ozette> ok this is kind of silly, but it's one very l7ong line: http://paste.debian.net/853899/
[17:09:33 CEST] <ozette> there's no whitespace between those flags, didn't know that was possible
[17:09:45 CEST] <c_14> This has got to be one of the most inefficient ways to get at thumbnails I have yet to see
[17:09:52 CEST] <c_14> ozette: /proc/cmdline is stripped of spaces
[17:09:56 CEST] <ozette> oh
[17:10:13 CEST] <ozette> i see -map
[17:10:23 CEST] <c_14> You can get at the line with spaces somehow but I don't know how right now and the ps man page is terrible
[17:10:33 CEST] <furq> this looks like it's been cut off
[17:10:34 CEST] <ozette> haha nvm :)
[17:10:38 CEST] <ozette> cut off?
[17:10:49 CEST] <ozette> it goes all the way to the right
[17:10:50 CEST] <furq> it ends in the middle of filter_complex
[17:10:58 CEST] <c_14> furq: https://paste.debian.net/plain/853899
[17:11:01 CEST] <c_14> use the raw paste
[17:11:21 CEST] <ozette> oh right thanks
[17:11:29 CEST] <furq> oh nvm i was reading it wrong
[17:11:35 CEST] <furq> that's -map
[17:11:37 CEST] <c_14> It basically splits the input into 20 copies and then seeks to arbitrary timestamps in each output and outputs one frame with -vframes 1
[17:11:38 CEST] <ozette> the plain is much better
[17:12:20 CEST] <furq> yeah i assume this is decoding the first 162 seconds 20 times
[17:12:24 CEST] <furq> and so forth
[17:12:42 CEST] <ozette> aren't the timestamps 'precise'? shouldn't that make it more fast?
[17:13:21 CEST] <c_14> any seeking that happens in the output is slow because every frame is decoded and then the timestamps are checked on each frame
[17:13:44 CEST] <furq> yeah -ss after -i is slower
[17:13:47 CEST] <ozette> so there's no way to make thumbnails in a more efficient way?
[17:13:48 CEST] <Spring> can the .MSS even extend that far?I thought it was only three digits
[17:13:53 CEST] <furq> ozette: the thumbnail filter
[17:14:08 CEST] <furq> the issue with it is that you can't tell it how many thumbnails to spit out
[17:14:21 CEST] <furq> or you can, but only by dividing the total number of frames
[17:15:33 CEST] <c_14> furq: can't you use -frames:v number_of_frames wanted
[17:15:35 CEST] <c_14> ?
[17:15:51 CEST] <ozette> in this case it says -ss first and then -i and then more -ss
[17:15:55 CEST] <furq> yeah but i assume he wants an even distribution
[17:16:14 CEST] <ozette> in this case, i use a library and i just say i want 20 thumbs
[17:16:19 CEST] <c_14> yeah, then you'll have to count input frames (or guess input frames) and average
[17:16:25 CEST] <furq> ozette: http://vpaste.net/GKtlx
[17:16:31 CEST] <furq> that's more or less what i ended up with when i was doing something similar
[17:16:37 CEST] <ozette> i can provide additional flags, but the way it works now i leave that up to the library
[17:16:45 CEST] <furq> er
[17:16:46 CEST] <ozette> oh thanks
[17:16:52 CEST] <furq> change $file to $1
[17:17:01 CEST] <furq> also that will use a lot of memory
[17:17:06 CEST] <c_14> And you might want to adjust the scale
[17:17:20 CEST] <furq> putting scale before thumbnail is slower but uses less memory
[17:17:58 CEST] <bornpilot> so is there or difference in quality between AVCHD h264 and mp4 h264?
[17:18:31 CEST] <ozette> what does skip_frame nokey do exactly?
[17:18:40 CEST] <furq> it skips non-keyframes
[17:18:52 CEST] <c_14> bornpilot: in general, no
[17:18:58 CEST] <ozette> ah
[17:19:09 CEST] <furq> it makes decoding much faster
[17:19:19 CEST] <furq> it will also use less memory with thumbnail
[17:19:30 CEST] <furq> if it still uses too much memory then use the select filter as well
[17:21:25 CEST] <bornpilot> thanks c_14
[17:22:48 CEST] <ozette> okay, thanks i will add -skip_frame nokey to the command and see if there's a significant difference
[17:23:45 CEST] <furq> if you mean the command you pasted then i would just throw that in the bin
[17:23:54 CEST] <furq> you'll have to print it out first but it'll be worth it
[17:23:55 CEST] <ozette> i can't imagine why scale before thumbnail would be slower, unless the command is interpretted from right to left
[17:24:11 CEST] <furq> scale after thumbnail only scales the output frames
[17:24:35 CEST] <furq> thumbnail=50 will keep 50 raw frames in memory so it can pick the "best" one
[17:24:52 CEST] <furq> you can imagine that eats memory pretty quickly with higher numbers
[17:33:52 CEST] <ozette> so if scale is before thumbnail, it'd scale all raw frames?
[17:34:11 CEST] <furq> yes
[17:34:18 CEST] <ozette> ah
[17:40:21 CEST] <trueroll> hi, I'm trying to build ffmpeg 3.1.3 Android with the NDK r12b Clang compiler
[17:41:02 CEST] <trueroll> I'm getting a compiler test error from configure file
[17:42:10 CEST] <trueroll> any help would be great! (btw I can build w/ gcc but Google is not including gcc in the NDK in the next version)
[17:42:45 CEST] <c_14> trueroll: upload the last 50 or so lines from your config.log to a pastebin service
[17:43:35 CEST] <trueroll> ok just a sec
[17:45:38 CEST] <trueroll> http://pastebin.com/guj2Bkqk
[17:49:15 CEST] <c_14> Can you upload your configure line to a pastebin as well?
[17:49:32 CEST] <trueroll> sure
[17:51:01 CEST] <trueroll> http://pastebin.com/QG7aihW7
[17:55:29 CEST] <c_14> I'd start by getting rid of all the extra-cflags that don't list extra include dirs
[17:56:15 CEST] <furq> i assume you want to change -mcpu=armv7-a
[17:57:23 CEST] <trueroll> @furq yeah I'll try that real quick
[17:57:37 CEST] <trueroll> @c_14 you mean all the enable/disables?
[17:57:47 CEST] <c_14> Also, can you check config.log for "defined __ARM_ARCH_"
[17:58:18 CEST] <c_14> trueroll: no, the --extra-cflags stuff, specifically the -marm, -mcpu=, -mfpu= ones
[17:58:22 CEST] <c_14> ffmpeg _should_ set those itself
[17:58:50 CEST] <trueroll> __ARM_ARCH_ appears a lot
[17:59:00 CEST] <trueroll> ah ok I'll remove those
[18:01:12 CEST] <c_14> I'm mainly interested in what probe_arm_arch() returns in the configure script, you could just add `echo "$cpu"' to line 4298
[18:02:24 CEST] <trueroll> k
[18:05:56 CEST] <flux> hmm, I wonder if ffmpeg's mov fragmentation and http post uploading should work together?
[18:07:47 CEST] <flux> oh, user error. it does! why would anyone use rtmp over this.. :)
[18:08:46 CEST] <flux> doesn't seem like it flushed the last segment, though.. could be another user error ;)
[18:21:50 CEST] <trueroll> @c_14 that never gets hit I think
[18:25:49 CEST] <TwinTailed> flux: I made the debugging steps you told me to do. avcodec_send_frame receives and successfully processes all audio frames, but avcodec_receive_packet doesn't process all frames, there is a missing one
[18:26:07 CEST] <TwinTailed> flux: and I guess it is getting missed in an AVERROR(EAGAIN) error
[18:31:31 CEST] <TwinTailed> flux: or maybe getting lost in EOF
[18:35:34 CEST] <trueroll> @c_14 I think this might be the problem http://pastebin.com/wh8QAJZU
[18:38:03 CEST] <trueroll> __ARM_ARCH_7A_ is not getting set and neither is __TARGET_ARCH_7_A
[19:01:10 CEST] <c_14> yeah, something's weird there
[19:01:18 CEST] <c_14> I don't know much about arm though
[19:10:42 CEST] <trueroll> thanks for taking a look though
[19:11:33 CEST] <c_14> have you tried asking on the ffmpeg-user mailing list?
[19:15:15 CEST] <trueroll> not yet, that's a good next step though right?
[19:15:24 CEST] <c_14> yes
[19:16:14 CEST] <trueroll> ok I'll do that soon
[21:26:33 CEST] <TwinTailed> flux: I made the debugging steps you told me to do. avcodec_send_frame receives and successfully processes all audio frames, but avcodec_receive_packet doesn't process all frames, there is a missing one
[21:26:34 CEST] <TwinTailed> flux: or maybe getting lost in EOF
[21:26:45 CEST] <TwinTailed> flux: I made the debugging steps you told me to do. avcodec_send_frame receives and successfully processes all audio frames, but avcodec_receive_packet doesn't process all frames, there is a missing one
[21:29:08 CEST] <Polochon_street> Hi! I use ffmpeg to decode songs into raw arrays, but I've encountered a case where ffprobe says the file is « pcm_s16le ([1][0][0][0] / 0x0001) » which I assume is already RAW data. So it obviously fails at the avcodec_open() step; what should I do to handle these kind of files ?
[21:31:30 CEST] <c_14> avcodec_open2 should work just fine
[21:31:36 CEST] <c_14> There is AV_CODEC_ID_PCM_S16LE
[21:57:10 CEST] <Polochon_street> c_14: okay, so I need to look for other clues then. Thanks!
[22:58:34 CEST] <Wilawar> I tried to reverse a video (including audio) with ffmpeg.exe. The command I used is  >> ffmpeg.exe -i '.\dj quicksilver ameno en reversa-jiRAePUYlvM.webm' -vf reverse -af reverse -vcodec vp8 -acodec libvorbis Ameno_Quicksilver_Rw.webm <<
[22:59:02 CEST] <c_14> you will most likely run out of memory on that
[22:59:13 CEST] <c_14> and I'm not sure it works for audio
[22:59:28 CEST] <c_14> ah, areverse insteod of reverse for audio
[22:59:48 CEST] <Wilawar> When I tried it with the build dated 2016-04-25, the video I got stuttered in VLC (I just remembered I wanted to try it in some other player; sorry, will do that now) &
[23:00:14 CEST] <Wilawar> but when I used the 2016-10-03 build from Zeranoe (Im on Vista), it outright crashed
[23:00:51 CEST] <c_14> Like I said, if your video is long enough or of high enough resolution you'll run out of memory and the kernel will kill it
[23:01:42 CEST] <Wilawar> So, Ill try it in MPC, then if it doesnt work paste the log. There was no weird stuff in it besides "[libvorbis @ 0207d700] Queue input is backward in time", which I got very often and "[webm @ 02116c60] Non-monotonous DTS in output stream 0:1; previous: 38780, current: 38763; changing to 38780. This mayresult in incorrect timestamps in the output file."
[23:01:58 CEST] <Wilawar> c_14: First, the Windows kernel wont do that, second, its short enough
[23:02:02 CEST] <Wilawar> :)
[23:02:22 CEST] Action: c_14 doesn't actually know what the windows kernel does when it runs out of memory but it's got to do something
[23:02:54 CEST] <Wilawar> It will make the machine come to a halt through paging
[23:03:09 CEST] <c_14> What's it do when your hdd runs out of space?
[23:03:14 CEST] <Wilawar> Also, the video is rather low res; as this was just for a quickly done try out of curiousity, I requested the 360p version
[23:03:51 CEST] <Wilawar> c_14: The program will get an error from the WriteFile call!?
[23:04:38 CEST] <c_14> Shouldn't it get an error from malloc?
[23:05:21 CEST] <Wilawar> What does malloc have to do with the HDD?
[23:05:35 CEST] <c_14> well, if you run out of ram the kernel starts paging
[23:05:43 CEST] <c_14> So what if it runs of out space for the pagefile
[23:06:18 CEST] <Wilawar> The program will get an error response to VirtualAlloc (or whatever it used to allocate memory)
[23:06:40 CEST] Action: c_14 is pondering if he can completely lock a windows system by requesting infinite memory and not responding to errors
[23:06:44 CEST] <Wilawar> If Windows really runs out of commit space, then yes, it might kill programs, though I havent seen this happening
[23:07:03 CEST] <c_14> Out of purely scientific curiosity, of course.
[23:08:23 CEST] <Wilawar> Getting the physical memory used percentage into the realm of over 95% has the same effect and since Im always short on RAM (only 4GiB here and not even patched my kernel for PAE), I experience this way too frequently :|
[23:08:51 CEST] <Wilawar> Btw, the video is still stuttery in MPC-HC, though it doesnt outright drop out as with VLC
[23:09:49 CEST] <Wilawar> Windows does a very shitty job on taming ill-behaving programs
[23:10:32 CEST] Action: c_14 is testing the command on one of his videos
[23:10:36 CEST] <c_14> I keep forgetting how slow vp8 is
[23:11:13 CEST] <Wilawar> Like, I dont think it even attempts at all. Setting the priority to Idle in Process Hacker or even the IO priority to Very Low (which the Task Manager, the one that comes with Windows, doesnt even offer) can do wonders
[23:13:02 CEST] <c_14> Output video works fine here
[23:13:19 CEST] <Wilawar> Ill retry with the MP4 version. If you wanted to try it on the video Ive got: I grabbed it with 'youtube-dl.exe --no-mtime https://www.youtube.com/watch?v=jiRAePUYlvM -f 43'
[23:13:30 CEST] <Wilawar> -f 18 is the MP4 version
[23:14:47 CEST] <c_14> All these fun libvorbis warnings
[23:16:26 CEST] <c_14> output works fine here
[23:16:30 CEST] <c_14> no stuttering no nothing
[23:16:33 CEST] <Wilawar> Hm :|
[23:17:02 CEST] <Wilawar> At least libx264 and aac arent giving off warnings all the time
[23:17:34 CEST] <c_14> https://dl.c-14.de/t/out.webm
[23:17:42 CEST] <c_14> try playing that in your player to see if it causes the same problems
[23:18:14 CEST] <c_14> That way we can find out if it's the player or your version[s] of ffmpeg
[23:18:19 CEST] <Wilawar> The mpeg4 one seems to be fine
[23:18:57 CEST] <Wilawar> It just occured to me, maybe my CPU isnt strong enough to do the SW decode?
[23:19:11 CEST] <Wilawar> But its very low-res &
[23:19:12 CEST] <c_14> for vp8?
[23:19:16 CEST] <c_14> don't think that should be an issue
[23:19:52 CEST] <Wilawar> I can watch h264 720p video in Firefox although it doesnt use my GPU
[23:20:05 CEST] <Wilawar> But I do have to make sure nothing else is running in the background
[23:20:18 CEST] <Wilawar> But yeah, MP4 seems to be fine
[23:20:26 CEST] <Wilawar> Ill grab your file
[23:22:14 CEST] <Wilawar> Ach, a Deutscher! ;)
[23:22:33 CEST] <Wilawar> <-- Frange
[23:24:38 CEST] <Wilawar> c_14: Yours cuts out, too
[23:24:46 CEST] <Wilawar> (in VLC)
[23:25:31 CEST] <Wilawar> Its only the audio, however, the video continues to run
[23:26:05 CEST] <Wilawar> Since the other one worked fine, maybe I should take this to #videolan?
[23:26:34 CEST] <c_14> probably
[23:26:39 CEST] <Wilawar> Lots of "core warning: playback way too early (-1624945): playing silence" in the logs
[23:27:27 CEST] <Wilawar> CPU usage was in check during playback
[23:27:48 CEST] <Wilawar> Do you mind linking to your version?
[23:28:09 CEST] <Wilawar> Or is your monthly bandwith use restricted?
[23:28:26 CEST] <c_14> I don't mind
[23:28:35 CEST] <c_14> But it gets deleted in 6 hours
[23:28:43 CEST] <c_14> If you think you need it for longer I can create a permanent link
[23:28:46 CEST] <Wilawar> Hmm, should suffice)
[23:30:51 CEST] <Wilawar> On the other hand, there are also audio cutouts in MPC &
[23:32:06 CEST] <Wilawar> So maybe it _is_ ffmpeg that fails encoding the file properly. Im undecided
[23:32:32 CEST] <Wilawar> c_14: Did you get errors about wrong timestamps, too?
[23:32:39 CEST] <c_14> yep
[23:32:43 CEST] <c_14> warnings
[23:32:48 CEST] <Wilawar> Right
[23:36:17 CEST] <Wilawar> Oh, didnt know WMP could play back webm
[23:36:35 CEST] <Wilawar> Less errors than VLC
[23:37:34 CEST] <Wilawar> Any other media player (which doesnt rely on FFmpeg) to try?
[23:39:16 CEST] <TD-Linux> a web browser
[23:40:02 CEST] <c_14> recent chrome uses libav* though, doesn't it?
[23:40:14 CEST] <c_14> so firefox or ie I guess
[23:41:04 CEST] <TD-Linux> for vp8 and vp9? chromium uses libvpx, firefox uses ffvp8/9, but sounds like you have a container issue
[23:44:01 CEST] <Wilawar> TD-Linux: Good point, no issues in Firefox 51.0a2 (Dev Edition)
[23:50:09 CEST] <Wilawar> Other than trying to destroy my hearing (seriously, why is 100% volume the default everywhere!?) and my speakers, it doesnt cut out with ffplay, though the video seems to slow down often
[23:53:21 CEST] <Wilawar> (if you wonder, I wanted to write 'its fine with ffplay when I noticed the slowdowns  'Other than [&] it doesnt cut out' doesnt make sense, of course)
[23:58:27 CEST] <Wilawar> Ok, I opened the MP4 one for comparison (that ones truly fine) and I found that whereas the video slowdowns are rather easily noticable with ffplay, there are, indeed, short audio cut-outs with it, though I didnt notice them on the first go.
[23:59:51 CEST] <Wilawar> I & think
[00:00:00 CEST] --- Tue Oct  4 2016