[Ffmpeg-devel-irc] ffmpeg-devel.log.20150331

[burek]

[00:45:58 CEST] <prelude2004c> hello, Can anyone help ? http://pastebin.com/U1hFQc7i .... works well with 2 outputs but i can't seem to get it with just one
[00:46:00 CEST] <prelude2004c> sorry
[00:46:01 CEST] <prelude2004c> with 3
[00:46:11 CEST] <prelude2004c> it works well with 1 and 2 but with 3 it fails
[00:46:22 CEST] <prelude2004c> anything over 2 outputs the ffmpeg doesnt like it.. any ideas?
[01:00:58 CEST] <cone-225> ffmpeg 03Diego Biurrun 07master:87f87ccf3205: mjpeg: Use a more precise type for put_marker()
[01:00:59 CEST] <cone-225> ffmpeg 03Michael Niedermayer 07master:48183f144d1d: Merge commit '87f87ccf32055aca25fa08e29e22cca1ba9976d5'
[01:36:18 CEST] <bluck> anyone with knowledge of the filtering code around?
[01:36:33 CEST] <bluck> I'm looking for a way to frame thread a single filter chain
[01:37:30 CEST] <cone-225> ffmpeg 03Diego Biurrun 07master:4978850ca2cb: build: Split JPEG-related tables off into a separate component
[01:37:31 CEST] <cone-225> ffmpeg 03Michael Niedermayer 07master:a105931d3edf: Merge commit '4978850ca2cb1ec6908f5bc79cc592ca454d11e8'
[01:43:54 CEST] <prelude2004c> hey .. can anyone help.. with ffmpeg. and libnvenc .. how do i set it all to just 1 GPU
[02:14:41 CEST] <cone-225> ffmpeg 03Michael Niedermayer 07master:a38758a97efe: avcodec/h264_refs: fix null pointer dereference with -DTRACE
[04:34:21 CEST] <cone-225> ffmpeg 03Michael Niedermayer 07master:385fe984fc6c: avcodec/snowenc: Check av_frame_ref() return value
[04:46:06 CEST] <cone-225> ffmpeg 03Rodger Combs 07master:5cba5291a767: lavf/segment: style nits
[05:53:10 CEST] <jamrial> rcombs: telling someone to build ffmpeg on a Pentium MMX is not the best recommendation :p
[05:54:28 CEST] <rcombs> jamrial: I'm hoping he realizes that sounds hard and goes away
[05:54:36 CEST] <jamrial> lol
[05:54:46 CEST] <rcombs> building ffmpeg on Win32 is annoying enough without having to do it in the stone age
[10:02:01 CEST] <nevcairiel> its probably not even sse or anything like that, just i686 instructions =p
[10:02:33 CEST] <nevcairiel> but seriously, people still have working Pentium MMX boxes around?
[10:02:37 CEST] <nevcairiel> what a waste of space
[10:07:43 CEST] <JEEBsv> I still have a working pentium !mmx box around
[10:07:48 CEST] <nevcairiel> what a waste of space
[10:07:49 CEST] <nevcairiel> :D
[10:08:01 CEST] <JEEBsv> it's pretty with a carpet on it
[10:08:13 CEST] <JEEBsv> and in russia there's plenty of space for it
[10:08:25 CEST] <JEEBsv> I think it still has a 4GB HDD and windows 2000 installed on it
[12:11:57 CEST] <cone-818> ffmpeg 03Himangi Saraogi 07master:0c9abf211931: avfilter/vf_detelecine: Fix uninitialized array index error
[12:35:06 CEST] <cone-818> ffmpeg 03Rainer Hochecker 07master:8b152c355f6c: h264: avoid unnecessary calls to get_format
[12:45:45 CEST] <cone-818> ffmpeg 03Himangi Saraogi 07master:919d13d2383b: avformat/http: Fix null check on allocated value
[13:36:45 CEST] <haasn> http://ffmpeg.org/doxygen/trunk/structAVMotionVector.html This documentation doesn't explain the interpretation of width/height as they related to src_x/src_y; does src_x/y point to the center of the block, or to the top left (or something entirely else)?
[13:39:06 CEST] <kierank> it's from block to block
[13:39:11 CEST] <kierank> so internal position shouldn't matter
[13:44:14 CEST] <haasn> Then I don't understand the meaning of this structure at all
[13:49:05 CEST] <haasn> I thought it gave me the position of one of the motion vectors as relative to the frame
[13:53:17 CEST] <kierank> motion vectors are relative to the current block
[13:55:34 CEST] <haasn> current block?
[13:57:10 CEST] <haasn> Maybe I don't understand how to use access/use this struct, either
[13:57:29 CEST] <haasn> I query it using av_frame_get_side_data(frame, AV_FRAME_DATA_MOTION_VECTORS)
[13:59:05 CEST] <haasn> The documentation seems to suggest that the resulting AVFrameSideData->data will be a single AVMotionVector, but the example file in https://ffmpeg.org/pipermail/ffmpeg-devel/2014-August/160938.html seems to suggest it's actually an array of count sd->size / sizeof(*sd->data)
[13:59:18 CEST] <haasn> that example is also outdated, though; hmm
[14:01:00 CEST] <haasn> yeah, the up to date version also treats it like an array
[14:01:04 CEST] <haasn> (misleading documentation?)
[14:08:07 CEST] <haasn> the actual numbers seem to suggest that src_x/y point to the center of the block, as I anticipated
[14:08:08 CEST] <haasn> framenum,source,blockw,blockh,srcx,srcy,dstx,dsty,flags
[14:08:11 CEST] <haasn> 2,-1,16,16,   8,   8,   8,   8,0x0
[14:08:17 CEST] <haasn> (next step is at 24/8)
[14:10:19 CEST] <kierank> it's an array of blocks with associated motion vectors
[14:12:34 CEST] <haasn> Can I rely on this array always covering the entire frame, incidentally?
[14:19:21 CEST] <cone-818> ffmpeg 03Marton Balint 07master:c84b75c89c01: vf_drawtext: add support for setting box border width
[16:34:35 CEST] <cone-818> ffmpeg 03Michael Niedermayer 07master:65a646eb0b9a: avcodec/internal: Fix doxy for ff_alloc_packet2() return value
[16:34:36 CEST] <cone-818> ffmpeg 03Donny Yang 07master:749fbfd08175: png: Clear up the calculation of max packet size
[16:57:45 CEST] <michaelni> wm4, ive posted a patch which changes the default allocation limit to SIZE_MAX, iam not aware of any usecase this fixes or improves but i thought might want to comment as you seemed to dislike this limit
[16:58:20 CEST] <michaelni> thought, YOU might
[18:17:53 CEST] <Daemon404> anyone know what unite the pcr retransmit time is
[18:17:55 CEST] <Daemon404> for mpegtsenc
[18:17:57 CEST] <Daemon404> unit*
[18:18:23 CEST] <Daemon404> it doesnt seem to be documented anywhere
[18:30:18 CEST] <kierank> milliseconds
[18:30:21 CEST] <kierank> very likley
[18:30:41 CEST] <Daemon404> really?
[18:30:43 CEST] <Daemon404> default is 20...
[18:30:47 CEST] <Daemon404> that would be insane
[18:30:50 CEST] <kierank> yes that's normal
[18:31:02 CEST] <Daemon404> ... pcr packet every 20ms?
[18:31:03 CEST] <kierank> enables fast receiver channel cahnge
[18:31:19 CEST] <Daemon404> wouldnt that be an insane amount of overhead
[18:31:39 CEST] <Daemon404> (tryign to reduce it for hls segments, which only need it once for many segs)
[18:31:40 CEST] <kierank> not if you piggyback on the video PID
[18:32:24 CEST] <Daemon404> i somehow doubt the muxer does this
[18:32:44 CEST] <Daemon404> ill poke some segments with dvbsnoop
[18:32:46 CEST] <Daemon404> see whats going on
[18:39:18 CEST] <iive> 90kHz or 27MHz
[18:46:26 CEST] <iive> iso13818-1
[18:56:37 CEST] <selsta> anyone interested in helping me implement sample-aes hls decryption? im kinda stuck.. i did what the specs say and the result looks almost right.
[18:57:01 CEST] <selsta> (original: http://a.pomf.se/uaunsz.h264) (my decryption: http://a.pomf.se/tgszqh.h264)
[19:00:50 CEST] <selsta> but somewhere in the middle of the file, it starts to look like this http://a.pomf.se/aqevnp.png
[19:03:17 CEST] <selsta> oh ive just found that there are some extra bytes and thats why everything seems different.. that explains it
[19:22:12 CEST] <selsta> https://codesequoia.wordpress.com/2009/10/18/h-264-stream-structure/ solves my problem :)
[20:09:02 CEST] <prelude2004c> hello everone
[20:11:23 CEST] <prelude2004c> can anyone help me with something.. i am trying to encode 3 output bit rates.. " http://pastebin.com/CjVEEKQd " see code
[20:11:35 CEST] <prelude2004c> for whatever reason, i can't seem to be able to launch more than 2 streams
[20:11:39 CEST] <prelude2004c> the 3rd one always fails
[20:11:41 CEST] <prelude2004c> any idea why? 
[20:12:28 CEST] <prelude2004c> also, anyone know how i can set it to use the libnvenc but to specify which GPU to use ? eg.. the GPU is only 20% used let's say .. for one channel.. i want to put a few channels on that GPU to make it use up the rest of the power of the card. Can anyone assist with the direction i need to go ?
[20:18:04 CEST] <Daemon404> michaelni, you need to add TYPE_IDR too for TYPE_I in libx265.c
[20:18:08 CEST] <Daemon404> just found out.
[20:19:51 CEST] <michaelni> Daemon404, X265_TYPE_IDR: was in my patch
[20:20:08 CEST] <Daemon404> ah
[20:20:08 CEST] <Daemon404> yes
[20:20:11 CEST] <Daemon404> i missed it
[20:20:15 CEST] <Daemon404> same idea though.
[20:20:29 CEST] <Daemon404> LGTM when formatted then.
[20:38:16 CEST] <prelude2004c> can anyone assist?
[20:39:57 CEST] <prelude2004c> i am willing to pay for someone with experience in GPU encoding and H265 ... please PM me.. What i want done is.. Enable MultiGPU encoding for channels.. ( eg. 5 channels on 2 GPU's ) .. and I am looking to also use h265 GPU encoding if possible intead of h264. anyone interested?
[20:42:40 CEST] <kierank> prelude2004c:  good luck with
[20:46:33 CEST] <cone-818> ffmpeg 03Michael Niedermayer 07master:2ddedfd39c0d: avcodec/libx265: export choosen picture types
[20:49:19 CEST] <jamrial> prelude2004c: you could try emailing the people that added support for nvenc_h265 to libavcodec
[20:50:00 CEST] <BBB> so helpful kierank & :)
[20:50:24 CEST] <kierank> I thought he wanted to write a GPU encoder
[20:50:31 CEST] <kierank> Like the usual questions
[20:51:37 CEST] <BBB> I think hes looking to use something pre-packaged
[20:53:13 CEST] <selsta> out of 230k bytes, 16 bytes are randomly decrypted wrong. makes no sense ):
[20:54:26 CEST] <Daemon404> makes perfect sense
[20:54:30 CEST] <Daemon404> aes uses 128 bit blocks
[20:55:04 CEST] <selsta> i know that
[20:55:34 CEST] <selsta> but i decrypt 16 byte samples from the video stream.. and all work fine expect 1 in the middle of the video
[21:37:50 CEST] <Daemon404> https://mailman.videolan.org/pipermail/x265-devel/2015-March/007395.html
[21:37:55 CEST] <Daemon404> comments welcome as it would affect ffmpeg
[21:43:08 CEST] <jamrial> the x264 guys could get involved as well. both projects would benefit from a change like this
[21:46:21 CEST] <Daemon404> perhaps
[21:46:27 CEST] <Daemon404> i dont know if x264 would go do it though
[21:46:41 CEST] <Daemon404> no big changes seem to happen noewadays
[21:46:43 CEST] <Daemon404> nowadays*
[23:33:05 CEST] <electronrotoscop> is it possible to build ffmpeg with bmx as an external library?
[23:50:13 CEST] <cone-818> ffmpeg 03wm4 07master:8b76c0eb561b: avformat/mp3dec: offset seek index to end of id3v2 tag
[23:57:57 CEST] <kierank> electronrotoscop: no
[00:00:00 CEST] --- Wed Apr  1 2015