[Ffmpeg-devel-irc] ffmpeg-devel.log.20120213

burek burek021 at gmail.com
Tue Feb 14 02:05:03 CET 2012


[01:06] <^sandro^Zzz> hello... can anyone confirm that GPU support for FFMPEG is availab.e ?
[01:06] <^sandro^Zzz> available through vdpau and or vaapi
[01:06] <^sandro^Zzz> i have nvidia cards
[01:07] <Compn> for decoding ?
[01:07] <Compn> i dont remember if ffmpeg can use it
[01:11] <durandal_1707> --enable-vdpau           enable VDPAU code [autodetect]
[01:11] <durandal_1707> --enable-vaapi           enable VAAPI code [autodetect]
[01:14] <^sandro^Zzz> i did compile it with it.. but how would i use it in the ffmpeg -i commands
[01:18] <durandal_1707> is is only for decoding, ffmpeg -vcodec mpeg2_vdpau -i ...
[01:18] <^sandro^Zzz> :( 
[01:19] <durandal_1707> hardware encoding sucks
[01:20] <^sandro^Zzz> anyone have any experience with buffer overruns ? 
[01:20] <^sandro^Zzz> http://pastebin.com/fpu1PzD0
[01:20] <^sandro^Zzz> not sure what im doing wrong.. maybe its something obvious
[01:25] <^sandro^Zzz> anyone?
[01:25] <^sandro^Zzz> im stuck .. and can't move forward
[01:27] <durandal_1707> you can to such stuff from stream at least mpegts can not handle that
[01:28] <^sandro^Zzz> sorry.. im not sure i understood that
[01:29] <durandal_1707> and why are you usin all that funky arguments for x264, ffmpeg support parsing presets directly to x264
[01:29] <durandal_1707> you just set name of x264 preset and you are done with it
[01:30] <^sandro^Zzz> im not sure i undestand :(
[01:30] <^sandro^Zzz> what would the new line look like ?
[01:30] <durandal_1707> ^sandro^Zzz: mpegts muxer looks like does not supports streamed input
[01:31] <^sandro^Zzz> so how do i capture the raw UDP input ?
[01:31] <durandal_1707> mpegts @ 0x232c3a0] Unable to seek back to the start
[01:31] <^sandro^Zzz> its a live channel
[01:31] <durandal_1707> pick another muxer/container which is not brain dead
[01:32] <^sandro^Zzz> i feel stupid as hell righ tnow
[01:33] <^sandro^Zzz> you mean instead of mpegts
[01:33] <^sandro^Zzz> but im not even specifying mpegts
[01:33] <durandal_1707> root at bit14:/usr/src/ffmpeg# ffmpeg -y -i udp://@:5000 -f mpegts ...
[01:34] <^sandro^Zzz> ok i had remvoed it :P 
[01:34] <^sandro^Zzz> ok i was looking at a different line
[01:34] <^sandro^Zzz> ffmpeg -y -i udp://@:5000 -f mpegts -acodec libmp3lame -ar 48000 -ab 64k -s 704x480 -vcodec libx264 -b:v 2000k  -map p:"167" -flags +loop -cmp +chroma -partitions +parti4x4+partp8x8+partb8x8 -subq 5 -trellis 1 -refs 1 -coder 0 -me_range 16 -keyint_min 25 -sc_threshold 40 -i_qfactor 0.71 -bt 128k -maxrate 16000k  -qcomp 0.6 -qmin 10 -qmax 51 -qdiff 4 -level 30 /mnt/encoder/test.ts
[01:35] <^sandro^Zzz> gives me the same result
[01:35] <^sandro^Zzz> oops.. wihtout the mpegts file in the beginnig
[01:35] <^sandro^Zzz> what do you recommend i try instead of the mpegts
[01:35] <^sandro^Zzz> ?
[01:35] <^sandro^Zzz> for a UDP input
[01:35] <durandal_1707> you are still using .ts at the end
[01:37] <^sandro^Zzz> but it has to be a ts file
[01:38] <^sandro^Zzz> because i have to segment it to m3u8
[01:38] <^sandro^Zzz> im creating an HLS stream
[01:38] <^sandro^Zzz> its weird because i got it going before.. not sure what i did.. but it just stopped
[01:38] <^sandro^Zzz> but i did get it to a .ts file before
[01:39] <durandal_1707> than just use -acodec/-vcodec copy
[01:43] <^sandro^Zzz> hum.. i see your point
[01:43] <^sandro^Zzz> but can i still bring it down from its original 15Mbit/s to the 2Mbit/s for example
[01:43] <^sandro^Zzz> and into h264 
[01:43] <^sandro^Zzz> i want to have it write in H264
[01:44] <durandal_1707> try that on file
[01:45] <^sandro^Zzz> didn't work :(
[01:45] <^sandro^Zzz> doesn't do anything still same error
[01:45] <^sandro^Zzz> what should the line look like?
[01:46] <^sandro^Zzz> even as simple as ffmpeg -y -i udp://@:5000 -map p:"167" test.ts
[01:46] <^sandro^Zzz> still gives me buffer overrun
[01:46] <^sandro^Zzz> :(
[01:47] <durandal_1707> i dont see that messages on pastebin
[01:50] <^sandro^Zzz> no... im just saying.. same exact error
[01:50] <^sandro^Zzz> nothing changed 
[01:50] <^sandro^Zzz> i can paste it for you
[01:50] <^sandro^Zzz>  one sec
[01:51] <^sandro^Zzz> http://pastebin.com/aQB5S9hQ
[01:52] <^sandro^Zzz> im really stsuck on this.. :( 
[01:52] <^sandro^Zzz> im in awww
[01:52] <durandal_1707> complaining all the time will get you nowhere
[01:53] <^sandro^Zzz> i understand.. im trying hard.. been at it since morning.. reading all google stuff... can't seem to find the solution ( frustrating ) 
[01:53] <^sandro^Zzz> im sure i will happen eventually but.. :(
[01:53] <durandal_1707> try to use -threads 1
[01:54] <^sandro^Zzz> tried.. same error
[01:57] <durandal_1707> try increasing buffer -rtbufsize (guessing now)
[01:59] <durandal_1707> or edit libavformat/udp.c and increase UDP_TX_BUF_SIZE (another guessing, not my code)
[02:06] <michaelni> you can also try old releases and if one works git bisect
[02:06] <michaelni> if its a regression and we know what caused it then its likely easy to fix
[02:13] <^sandro^Zzz> well.. actually it was working before so it would appear maybe it is a problem with the new release of ffmpeg
[02:15] <^sandro^Zzz> that would make sense
[02:18] <^sandro^Zzz> i think it was working with ffmpeg-export-snapshot-2009-12-02
[02:19] <durandal_1707> that is soo old
[02:20] <^sandro^Zzz> yup.. i know i got it writing to file on another ffmpeg..
[02:20] <^sandro^Zzz> it makes sense.. it was working ok for me yesterday
[02:20] <^sandro^Zzz> today it started giving me the buffer overurns.. and i have been palying around and recompiling ffmpeg a few times
[02:20] <^sandro^Zzz> but each time i went to the repo.. latest git
[02:20] <^sandro^Zzz> maybe it was not that version.. not sure.
[02:28] <CIA-67> ffmpeg: 03Anton Khirnov 07master * r3f77c41171 10ffmpeg/libavcodec/bmpenc.c: bmpenc: switch to encode2().
[02:28] <CIA-67> ffmpeg: 03Anton Khirnov 07master * r89829242a6 10ffmpeg/libavcodec/dnxhdenc.c: dnxhdenc: switch to encode2.
[02:28] <CIA-67> ffmpeg: 03Anton Khirnov 07master * r177bb4bf50 10ffmpeg/ (3 files in 3 dirs): cljr: set the properties of the coded_frame, not input frame.
[02:28] <CIA-67> ffmpeg: 03Anton Khirnov 07master * r3dffa1b46c 10ffmpeg/libavcodec/cljr.c: cljr: implement encode2.
[02:28] <CIA-67> ffmpeg: 03Andrey Utkin 07master * rc22e2ec9d1 10ffmpeg/avconv.c: 
[02:28] <CIA-67> ffmpeg: avconv: set AVFormatContext.duration from '-t'
[02:28] <CIA-67> ffmpeg: Set output files duration to recording_time option, if given.
[02:28] <CIA-67> ffmpeg: Rationale: to save duration into metadata for file that is written to
[02:28] <CIA-67> ffmpeg: non-seekable output, for formats like FLV (with metadata at beginning).
[02:28] <CIA-67> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[02:28] <CIA-67> ffmpeg: 03Andrey Utkin 07master * re3cc617263 10ffmpeg/libavformat/utils.c: 
[02:28] <CIA-67> ffmpeg: lavf: fix aspect ratio mismatch message.
[02:28] <CIA-67> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[02:28] <CIA-67> ffmpeg: 03Ronald S. Bultje 07master * ref66a0ed2e 10ffmpeg/libswscale/x86/output.asm: 
[02:28] <CIA-67> ffmpeg: swscale: use named registers in yuv2yuv1_plane() place.
[02:29] <CIA-67> ffmpeg: Most of the function had been converted before, but I forgot this
[02:29] <CIA-67> ffmpeg: particular location.
[02:29] <CIA-67> ffmpeg: 03Ronald S. Bultje 07master * r8c433d8a03 10ffmpeg/libswscale/x86/output.asm: 
[02:29] <CIA-67> ffmpeg: swscale: rename "dstw" to "w" to prevent name collisions.
[02:29] <CIA-67> ffmpeg: "dstw" can collide with the word-version of the "dst" argument, causing
[02:29] <CIA-67> ffmpeg: all kind of weird stuff down the pipe.
[02:29] <CIA-67> ffmpeg: 03Ronald S. Bultje 07master * r3e23badd83 10ffmpeg/libswscale/x86/output.asm: swscale: convert yuv2yuvX() to using named arguments.
[02:29] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * r4e04e1b81e 10ffmpeg/: (log message trimmed)
[02:29] <CIA-67> ffmpeg: Merge remote-tracking branch 'qatar/master'
[02:29] <CIA-67> ffmpeg: * qatar/master:
[02:29] <CIA-67> ffmpeg:  swscale: convert yuv2yuvX() to using named arguments.
[02:30] <CIA-67> ffmpeg:  swscale: rename "dstw" to "w" to prevent name collisions.
[02:30] <CIA-67> ffmpeg:  swscale: use named registers in yuv2yuv1_plane() place.
[02:30] <CIA-67> ffmpeg:  lavf: fix aspect ratio mismatch message.
[08:34] <LordRPI> Silly question: in the multithreaded part of ffmpeg decoding, does each additional thread cause an additional frame to be buffered before you get access to the first frame?
[09:00] <^sandro^Zzz> hello
[09:00] <^sandro^Zzz> the buffer overrun issue doesn't exist in the 0.10 version
[09:00] <^sandro^Zzz> so its a bug in the latest version
[09:11] <^sandro^Zzz> hello
[09:13] <^sandro^Zzz> this is silly.. had it working a min ago
[09:13] <^sandro^Zzz> ran command again
[09:13] <^sandro^Zzz> [libx264 @ 0x10ae020] Input picture width (704) is greater than stride (0)
[09:13] <^sandro^Zzz> Video encoding failed
[09:13] <^sandro^Zzz> anyone have any ideas?
[10:23] <av500> LordRPI: yes
[15:20] <^sandro^Zzz> hello
[15:20] <^sandro^Zzz> [libx264 @ 0x1267640] Input picture width (704) is greater than stride (0)
[15:20] <^sandro^Zzz> anyone know what that means?
[15:20] <ohsix> stride is supposed to be >= the width
[15:21] <^sandro^Zzz> how do i set stride
[15:21] <^sandro^Zzz> i dont know what it means
[15:49] <^sandro^Zzz> anyone?
[15:50] <av500> stride is the real width of the buffer in memory
[15:50] <av500> stride >= width
[15:50] <^sandro^Zzz> but i dont know how to fix this error
[15:51] <^sandro^Zzz> im trying to trancode.. here is the url i am using
[15:51] <^sandro^Zzz> ffmpeg -y -i udp://@:5000 -f mpeg2video -acodec libmp3lame -ar 48000 -ab 64k -vcodec libx264 -map p:"167" -flags +loop -cmp +chroma -partitions +parti4x4+partp8x8+partb8x8 -subq 5 -trellis 1 -refs 1 -coder 0 -me_range 16 -keyint_min 25 -sc_threshold 40 -i_qfactor 0.71 -maxrate 3000k  -qcomp 0.6 -qmin 10 -qmax 51 -level 30 -bufsize 2M -metadata service_name="CNC WOrld English" -threads 0 
[15:51] <^sandro^Zzz> /mnt/encoder/test.ts
[15:52] <^sandro^Zzz> errors out with : [libx264 @ 0x2018760] Input picture width (704) is greater than stride (0)
[15:52] <^sandro^Zzz> Video encoding failed
[15:53] <av500> thats not an url
[15:53] <Tjoppen> try putting the sample on disk and try transcoding that to eliminate any udp issues
[15:54] <^sandro^Zzz> what do you mean ? instead of udp:// ? 
[15:55] <^sandro^Zzz> but its a live capture 
[15:55] <Tjoppen> so grab the stream using the appropriate tool
[15:55] <^sandro^Zzz> what kind of tool ? 
[15:56] <Tjoppen> nc seems able to do it
[15:56] <^sandro^Zzz> nc ?
[15:56] <Tjoppen> man nc
[15:57] <^sandro^Zzz> so.. we nc 
[15:57] <^sandro^Zzz> so we nc | ffmpeg -i -
[15:57] <^sandro^Zzz> something like that ?
[15:58] <av500> no
[15:58] <av500> man nc
[15:58] <Tjoppen> no, something like nc -u ip port > file, then run ffmpeg on the file so it doesn't depend on the health of the network, and so you can  upload the sample and possibly file a bug report
[15:58] <av500> then you read it 
[15:58] <Tjoppen> like grab a couple of megs off the stream
[15:59] <^sandro^Zzz> but that file will grow huge
[15:59] <Tjoppen> if the problem persists with the file on disk then you might want to file a ticket at https://ffmpeg.org/trac/ffmpeg/ and attach the file
[15:59] <Tjoppen> just ctrl+c until you have enough
[15:59] <Tjoppen> *when
[16:00] <^sandro^Zzz> ho ic
[16:00] <^sandro^Zzz> one sec.. let me try
[16:00] <^sandro^Zzz> prblem is i need to capture live to transcode into segmenter
[16:01] <Tjoppen> one problem at a time
[16:10] <^sandro^Zzz> funny thing is that it sometimes works :( 
[16:10] <^sandro^Zzz> i had it working yesterday 
[16:10] <^sandro^Zzz> today nada
[16:10] <^sandro^Zzz> not even sure what i changed
[16:47] <^sandro^Zzz> can anyone help with this ? 
[16:47] <^sandro^Zzz> im stuck and no place to go
[16:47] <^sandro^Zzz> im all googled out
[16:48] <^sandro^Zzz> i think its a bug
[16:52] <kierank> who was the guy interested in dvb subtitles
[17:40] <t4k1s> sandro: what is sending your UDP packets?
[17:43] <^sandro^Zzz> a capture device
[17:43] <^sandro^Zzz> sending the mp2 udp 
[20:05] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * r47e015e6f1 10ffmpeg/libavformat/matroskadec.c: (log message trimmed)
[20:05] <CIA-48> ffmpeg: matroskadec: properly fall back to generic seek.
[20:05] <CIA-48> ffmpeg: In particular, detect when the index is obviously broken.
[20:05] <CIA-48> ffmpeg: This fixes the worst symptoms of trac issue #958 and makes
[20:05] <CIA-48> ffmpeg: sense to allow seeking in files without index.
[20:05] <CIA-48> ffmpeg: However it is possible that there still is an index parsing bug
[20:05] <CIA-48> ffmpeg: with that file.
[20:05] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * rc18899d432 10ffmpeg/libavformat/matroskadec.c: 
[20:05] <CIA-48> ffmpeg: matroskadec: reset num_levels on seek.
[20:05] <CIA-48> ffmpeg: Otherwise when we run into levels beyond the max. allowed
[20:05] <CIA-48> ffmpeg: playback will be permanently broken.
[20:05] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[21:49] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * r048cc80292 10ffmpeg/tests/ref/seek/lavf_mkv: 
[21:49] <CIA-48> ffmpeg: Update mkv seek tests.
[21:49] <CIA-48> ffmpeg: Seek beyond the end will now directly return an error instead
[21:49] <CIA-48> ffmpeg: of claiming to succeed and then return EOF immediately on next read.
[21:49] <CIA-48> ffmpeg: This change is because before 47e015e6f1913f7da943898eb7716a954f947ff7
[21:49] <CIA-48> ffmpeg: mkv seek incorrectly never failed.
[21:49] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[22:20] <ubitux> ok, got a new box
[22:21] <ubitux> i will be able to run a valgrind with and without undef option
[22:25] <michaelni> ubitux: \o/
[22:49] <ubitux> can we raise the number of jobs?
[22:49] <ubitux> (with fate)
[22:50] <ubitux> mmh makeopts
[23:53] <ubitux> ok it's way more faster with the new box
[23:53] <ubitux> 1 hour instead of 7
[00:00] --- Tue Feb 14 2012


More information about the Ffmpeg-devel-irc mailing list