[Ffmpeg-devel-irc] ffmpeg.log.20180224

[burek]

[01:56:24 CET] <TexanMD> Hey all.  I'm attempting to transcode an x264 video to x265 on Windows 10 using 3.4.2, or the 20160223-b6652f5 nightly builds and shortly after the encoder starts, it crashes.  I downloaded an older build (3.3.4) and it works fine with the same command line.   I can run it under windbg but don't have symbols so the stack trace isn't very helpful.   Are symbols posted somewhere?  or will i have to recreate the problem under linux to f
[01:57:00 CET] <TexanMD> Being that it's windows I've tried both the shared and static builds as well
[01:58:14 CET] <JEEB> if the build has debug symbols included gdb will be able to load them
[01:58:56 CET] <JEEB> https://kuroko.fushizen.eu/random/gdb_builds/gdb-8.0.1.7z
[01:59:04 CET] <TexanMD> Can I use gdb to debug windows binaries?  I've actually never tried that.  Windbg can only load the export symbols which are better than nothing... but not great.
[01:59:07 CET] <JEEB> if this one after `gdb ffmpeg`
[01:59:12 CET] <JEEB> says symbols not loaded
[01:59:16 CET] <JEEB> then you don't have symbols
[01:59:59 CET] <JEEB> (that is a 64bit windows build of gdb)
[02:00:39 CET] <TexanMD> JEEB, cool give me a sec and I'll test it out
[02:01:19 CET] <JEEB> if you didn't build the FFmpeg binary yourself it's highly likely it doesn't have the debug symbols, though :P
[02:01:32 CET] <JEEB> (there are no official windows builds)
[02:03:28 CET] <TexanMD> they're not official?   I guess that makes sense.  Since they're hosted on a different domain
[02:05:17 CET] <TexanMD> yeah no symbols
[02:05:28 CET] <TexanMD> I guess i'll see if i can set up a build env
[02:06:10 CET] <JEEB> do note that windows compilation has gotten ridiculously slow on the configure step
[02:06:23 CET] <JEEB> so WSL or a linux VM are recommended if you want to get any sort of sane speeds
[02:06:37 CET] <JEEB> the build itself isn't too slow, but you'd be waiting for the configure script for like 10min
[02:07:21 CET] <TexanMD> just looking at the setup requirements its more than i'm gonna do on this box.... can you run a test in linux?  you should just need an x264 source file
[02:07:36 CET] <TexanMD> if it doesn't crash for you then i'll figure out a way to get a useful crashdump later this weekend
[02:08:12 CET] <JEEB> I think my last x265 build is from before they started doing a lot of changes over at x265 and I don't generally build it for windows
[02:08:25 CET] <JEEB> at least not with FFmpeg
[02:08:46 CET] <TexanMD> I was thinking test on linux
[02:08:57 CET] <TexanMD> if its a cross platform bug it'll be way easier to get debug info
[02:09:44 CET] <JEEB> ok, my current FFmpeg build on my linux boombox doesn't have x265 built in :P
[02:09:53 CET] <JEEB> and it's 3 am so I'm not going to start doing that right now
[02:10:00 CET] <TexanMD> lol seems fine
[02:10:07 CET] <TexanMD> I appreciate the help either way
[02:10:31 CET] <TexanMD> what I need to get done is working with an old build, but I figured if I can document a crash i'd do it
[03:12:16 CET] <luxio> how do I convert a VP9 WebM to a VP8 one?
[03:22:36 CET] <furq> -i foo.webm -c:v libvpx -c:a copy bar.webm
[03:22:40 CET] <furq> i'm not sure why you'd bother though
[05:15:58 CET] <xxzz> Hi, how can I convert .exo files to mp4?
[05:33:10 CET] <furq> isn't exo that weird encrypted youtube thing
[05:34:18 CET] <furq> if it's not a paid video and it's still up then you're probably better off just downloading it with youtube-dl
[05:45:08 CET] <xxzz> furq: yeah YouTube stuff
[05:46:42 CET] <xxzz> I use youtube-dl but some videos I have on my phone, the uploader deleted those last week and in the current region I am not able to play them on phone youtube
[05:47:32 CET] <xxzz> I used to think YouTube is global but they are not
[05:53:07 CET] <furq> i don't know of anything that breaks the encryption
[05:53:27 CET] <furq> i would assume they're just encrypted mp4 fragments
[05:56:25 CET] <xxzz> yes
[05:56:54 CET] <xxzz> I will move them to desktop and run ffprobe, see if SMTube picks it up or not
[05:59:45 CET] <xxzz> There is ExoPlayer will check that
[10:28:40 CET] <ferz> Hi, can I use ffserver to stream a mp4 (h264) file?
[10:39:05 CET] <durandal_1707> ferz: ffserver is no more
[10:55:38 CET] <ferz> durandal_1707: Thank you.   What's next?
[11:21:34 CET] <stelonix_> Can I use ffmpeg to capture a single window by its id?
[12:02:59 CET] Action: zyme wonders if his window has id's - being a windows user. Also reminds me I wonder if the Win10 linux env supports gui interaction yet, since they had a similar function to that already in the HyperV XP env that was part of windows 7...
[12:05:12 CET] <stelonix_> You window has an hWnd ;)
[12:05:25 CET] <stelonix_> well I guess the type was HWND or whatever
[12:15:32 CET] <furq> you can't do that with gdigrab
[12:15:41 CET] <furq> it has to be the window title or a desktop region
[12:16:01 CET] <furq> you're generally better off using OBS for desktop capture stuff anyway
[12:21:06 CET] <stelonix_> Not using gdi by the way
[12:21:08 CET] <stelonix_> x11
[12:21:13 CET] <stelonix_> x11grab
[12:21:15 CET] <stelonix_> But doesn't matter
[12:21:40 CET] <stelonix_> Figured a way to get window position, just need to find out if I can set start x y pos
[12:21:48 CET] <stelonix_> With x11grab
[15:19:51 CET] <klaxa> ffmpeg -f x11grab -s widthxheight -i :0+x_offset,y_offset iirc
[16:08:22 CET] <classict> Hi everybody
[16:09:10 CET] <classict> I want to know how hd speed affect ffmpeg when I use vcodec copy.
[16:09:50 CET] <classict> ffmpeg  -i rtmp -vcodec copy ...rtmp:server, I get near 20 even 30% cpu usage
[16:10:04 CET] <sfan5> unless your cpu is very slow, I/O will definitely be the bottleneck when copying
[16:20:27 CET] <classict> can I setup use tmpfs, because this doesn't write on disk
[16:20:46 CET] <classict> from rtsp ip camera, to rtmp server
[16:20:51 CET] <classict> ?
[16:26:43 CET] <durandal_170> if tmpfs is big enough
[16:35:59 CET] <shtomik> Hi guys, can you help me? I'm using transcoding.c example, and on finish a have a spoiled sound, what am I doing wrong? video with sound(https://www.youtube.com/watch?v=4oe4Ub2fBp4). Thanks!
[18:38:24 CET] <ArsenArsen> can I overlay a sound file with a blank video
[18:43:41 CET] <durandal_1707> yes, you can
[18:48:55 CET] <arooni> question:  i have a 4 .mkv files that are quite large (500mb - 4gb).  whats the *fastest lowest cpu way* to convert them to audio only in a .mp4, or .mp3 format ?
[18:50:24 CET] <furq> -i foo.mkv -c:a copy -vn bar.m4a
[18:50:29 CET] <DHE> mp3 is a specific codec in its own container, so if the source video isn't already in MP3 format a transcode is needed. even mp4 has a list of allowable codecs which might not work
[18:50:39 CET] <furq> what he said
[18:51:06 CET] <furq> if that command doesn't work then you'll need to either transcode or use mka
[20:01:58 CET] <saucecode> Wait a second! One of you explained to me that some audio formats have the moov atom at the end of the file and that's why ffmpeg wont process them until it can read the whole file.
[20:02:30 CET] <saucecode> If this is the case, how is youtube downloading and playing those files at the same time?
[20:03:15 CET] <JEEB> the stuff provided by most streaming services (be it video on demand which is not really streaming per se, or actual live streaming services) nowadays have things in segments
[20:03:23 CET] <JEEB> and each segment is independently parse'able if not decode'able
[20:03:57 CET] <JEEB> of course specifically regarding moov, if you have the mp4 file with the MOOV atom (index of the file) at the back, then if the web server lets you seek in the file then you don't have to read the whole file
[20:04:18 CET] <JEEB> also, most things that serve you mp4 files have once written it, and then re-written the moov by moving it into the beginning
[20:04:27 CET] <JEEB> because you can do that after you've learned the whole structure of your file
[20:04:37 CET] <saucecode> Hmm
[20:04:38 CET] <JEEB> in ffmpeg.c you have -movflags faststart that does that second pass
[20:04:55 CET] <JEEB> which, after finishing writing once, rewrites the whole file again
[20:04:58 CET] <JEEB> with the moov in front
[20:05:15 CET] <furq> you might be conflating it with ffmpeg not being able to stream mp4 because it can't write the moov atom until the whole file is done
[20:05:21 CET] <furq> stream it as an output, that is
[20:05:59 CET] <saucecode> I take it that that seeking-to-the-moov is non-standard, otherwise ffmpeg would also be doing it?
[20:06:06 CET] <furq> ffmpeg does do it
[20:06:17 CET] <JEEB> no, as far as I know ffmpeg does that if the HTTP server lets it do it
[20:06:22 CET] <furq> yeah it definitely does
[20:06:23 CET] <JEEB> but that's for reading
[20:06:24 CET] <saucecode> I mean doing a separate HTTP request specifically to retrieve the moov
[20:06:29 CET] <furq> that's what i mean
[20:06:34 CET] <furq> except it usually does several requests
[20:07:04 CET] <saucecode> Wat
[20:07:08 CET] Action: saucecode checks
[20:07:29 CET] <furq> that's how everything does it
[20:08:08 CET] <furq> you read until you find either the moov or the mdat header, if you find the mdat then you deduct some arbitrary amount from Content-Length, do a range request from that until the end, and then keep doing that until you find the moov atom
[20:09:21 CET] <furq> you probably need a relatively recent ffmpeg
[20:09:21 CET] <JEEB> anyways, back in the days when people were actively providing (normal, non-fragmented) mp4 files over HTTP, everyone used the "move moov to front" stuff after finishing an encode pretty much. that way was simpler for clients. but nowadays everyone's moved to fragmented/segmented stuff where each fragment or segment is independently readable (and they're smaller so no-one cares if you have to read the whole
[20:09:28 CET] <JEEB> segment/fragment) :P
[20:09:56 CET] <furq> it's still recommended to do that for e.g. uploading to youtube
[20:10:03 CET] <furq> so it can start processing the file while you're still uploading it
[00:00:00 CET] --- Sun Feb 25 2018