[Ffmpeg-devel-irc] ffmpeg-devel.log.20150418

[burek]

[00:14:51 CEST] <cone-263> ffmpeg 03wm4 07master:e2691b809271: hls: skip to next segment if the current is unavailable
[01:15:22 CEST] <rcombs> it seems I've just implemented BFSTM support in lavf
[01:16:28 CEST] <rcombs> though it breaks on the last block
[01:18:04 CEST] <rcombs> this is some excellent Mario Kart 8 music
[01:20:22 CEST] <jamrial> haha
[01:20:44 CEST] <jamrial> i implemented a couple game audio formats some time ago, like RSD
[01:21:06 CEST] <jamrial> and so did paul
[01:21:32 CEST] <rcombs> yeah, BFSTM is basically a new header format for BRSTM
[01:21:43 CEST] <rcombs> so I just tweaked brstm.c, which is Paul's
[01:21:58 CEST] <jamrial> cool
[01:22:04 CEST] <rcombs> still have some hacks to remove
[01:22:13 CEST] <jamrial> does it use the same adpcm codecs?
[01:22:15 CEST] <rcombs> and I need to work out why decoding the last block is failing
[01:22:17 CEST] <rcombs> yup
[01:22:50 CEST] <rcombs> referenced vgstream (winamp plugin with support for the two) to work out the format differences
[01:23:49 CEST] <rcombs> the only samples I have are from Mario Kart 8, which only uses AV_CODEC_ID_ADPCM_THP
[01:24:08 CEST] <rcombs> which naturally is the codec that involves the most work
[01:25:11 CEST] <rcombs> http://puu.sh/hhCs4/c839eedb1c.png whoooo uninitialized variables
[02:56:08 CEST] <fluffrabbit|away> BBB: vpx requires MSYS. I'm not a developer anymore. Cut me some slack. I think it's probably safe to assume that the steady degradation of framerate is NOT the fault of libvpx.
[02:56:55 CEST] <jamrial> wm4: foobar2000 for example uses libavcodec's mp3 float decoder
[02:57:25 CEST] <jamrial> but they use their own demuxer for the container and not libavformat
[02:58:29 CEST] <jamrial> libavcodec for mp3, aac and vorbis. other libraries for anything else
[03:26:45 CEST] <radio_guy> Hi - anyone here who can help me with a copyright issue. May be I discovered a ffmpeg related copyright infringement.
[03:30:15 CEST] <cone-530> ffmpeg 03Michael Niedermayer 07master:80fa48a413a0: avformat/utils: Use codec_info_duration_fields in fps calculation end decission when no timestamps and thus no duration_count is available
[03:30:15 CEST] <cone-530> ffmpeg 03Michael Niedermayer 07master:edbb9b5e8fe5: avformat/utils: Add HEVC to unreliable_tb()
[04:05:08 CEST] <Prelude2004c> hey quick question guys.. i have a source multicast ( UDP ) ... i am trying to traverse from one network to another .. its a direct fiber 10 gig link.. but i am showing artifacts on the conversion to h264... so basically i ffmpeg -i < input > vcopy and acopy and -f mpegts udp://destination:xxxx
[04:05:36 CEST] <Prelude2004c> then on destination:xxx i do ffmpeg -i localhost:xxxx transcode -f HLS
[04:05:48 CEST] <Prelude2004c> how come i see artifacts every once in a while.. likely due to the UDP right?
[04:06:04 CEST] <Prelude2004c> i don't have a lot of power on the source so how do i set it to TCP so all packets always arrive
[04:20:37 CEST] <kierank> you can't multicast tcp
[04:20:45 CEST] <kierank> you need to fix your network such that it doesn't drop packets
[04:53:02 CEST] <Compn> Prelude2004c : what kind of artifacts ?
[04:53:26 CEST] <Compn> increase keyframes
[04:53:34 CEST] <Compn> then you wont need to worry about it so much :P
[04:57:52 CEST] <lordkrondor> use traceroute & mtr to figure it out heh
[06:35:50 CEST] <Timothy_Gu> How hard would it be to implement an SSIM filter like the existing PSNR one?
[16:33:09 CEST] <rkromer> Is anyone out there working on implementing the CineForm codec?
[16:36:58 CEST] <kierank> rkromer: afaik no
[16:37:08 CEST] <kierank> if a spec appeared on the mailing list I think it would help
[16:37:28 CEST] <kierank> and some test vectors
[16:39:17 CEST] <rkromer> The available specs are sadly not complete.
[16:40:20 CEST] <kierank> rkromer: really?
[16:40:24 CEST] <kierank> smpte playing games again?
[16:40:30 CEST] <rkromer> Especially not regarding the lossless flavor.
[16:41:29 CEST] <rkromer> I guess I must do some reverse engineering, as I would use it.
[16:41:49 CEST] <kierank> rkromer: do you have the specs?
[16:43:55 CEST] <rkromer> Yep, both the horribly expensive documents by SMPTE.
[16:44:47 CEST] <kierank> I am rather fed up of expensive SMPTE documents being produced that are incomplete
[16:44:53 CEST] <kierank> I need to make an angry blog post about it
[16:46:30 CEST] <rkromer> The lossless codec version interests me for archival purposes. (I am not a production or post-production person.)
[16:46:39 CEST] <kierank> I see
[16:53:29 CEST] <kierank> rkromer: can you make a list somewhere of all the deficiencies?
[16:57:56 CEST] <rkromer> I am working on it. The "standard" use is quite well documented, but the most interesting lossless flavour only... in a side note. I am caricaturing just a little.
[17:02:22 CEST] <kierank> I think vc-5 doesn't cover lossless explicitly
[17:02:58 CEST] <rkromer> In fact, I noticed its existence by doing a mistake.
[17:08:31 CEST] <rkromer> At GoPro I could not gather useful information so far. They answer to download the GoPro Studio...
[17:09:17 CEST] <kierank> yeah they won't help you
[17:12:30 CEST] <rkromer> In fact we have a free codec without a free documentation, i.e. a proprietary codec.
[17:14:10 CEST] <kierank> well it's patented too which the american archiving people don't like
[17:18:55 CEST] <rkromer> I am Swiss but I don't like neither this pseudo-free concept.
[17:19:56 CEST] <kierank> neither do I
[17:20:26 CEST] <kierank> What annoys me more is when they say the codec is standardised but the standards are not complete
[17:20:38 CEST] <kierank> and as an independent implementor you have no way of reporting bugs
[17:20:53 CEST] <kierank> At least in mpeg if you report a bug on the mailing list someone will listen
[17:20:58 CEST] <kierank> but smpte doesn't have any way of doing that
[17:22:52 CEST] <rkromer> Hmm... Should I include this as an exercise for my students? They are often much better skilled as I am!
[17:24:38 CEST] <kierank> Could write a paper about "Just because something is standardised, doesn't mean it's independently implementable"
[17:25:26 CEST] <rkromer> :-)
[17:25:58 CEST] <rkromer> I'll check how difficult it would be to implement this codec into libavcodec next semester. Thanks a lot for the chat!
[17:53:39 CEST] <cone-200> ffmpeg 03Michael Niedermayer 07master:e739cbb2bb91: avcodec/cabac: remove START/STOP_TIMER debug/benchmark code
[18:35:43 CEST] <cone-200> ffmpeg 03Mariusz SzczepaDczyk 07master:e623e8cbb033: lavf: add documentation on directory listing API
[18:41:59 CEST] <Compn> rkromer : kostya is working on cineform
[18:42:09 CEST] <Compn> theres a patch posted to libav-devel
[18:42:15 CEST] <Compn> for cfhq i think
[18:43:15 CEST] <Compn> hmm where did it go
[18:43:38 CEST] <Compn> oh no, i'm thinking of canopus hqa
[18:43:39 CEST] <Compn> my bad
[18:43:41 CEST] <Compn> :((((((((((
[18:45:42 CEST] <Compn> rkromer : but you should contact kostya , he does some RE'ing on some codecs. http://codecs.multimedia.cx/?p=419
[18:47:07 CEST] <Compn> if cineform uses wavlets, thats going to be confusing and difficult as hell to implement.
[18:47:18 CEST] <Compn> wavelet codecs are more difficult...
[18:48:45 CEST] <jamrial> i thought kostya quit developing for multimedia projects
[18:50:32 CEST] <Compn> once a reverse engineer, always a reverse engineer.
[18:50:55 CEST] <Compn> then theres maxim person who also does some codec engineering
[18:53:18 CEST] <Compn> and mike melanson does (or did) some work on writing video specs down
[18:58:30 CEST] <rkromer> Compn: Thanks a lot!
[19:01:35 CEST] <Compn> jamrial : also probably the bink devel trolling kostya helped him into doing some more work on bink :D
[19:01:49 CEST] <Compn> development by trolling
[19:20:12 CEST] <cone-200> ffmpeg 03James Almer 07master:6863249b347c: aaccoder: use put_sbits()
[19:20:13 CEST] <cone-200> ffmpeg 03James Almer 07master:b5564e041bdc: dcaenc: use put_sbits()
[19:20:14 CEST] <cone-200> ffmpeg 03James Almer 07master:979b77eec5db: dca_xll: use av_clip_intp2()
[19:20:15 CEST] <cone-200> ffmpeg 03James Almer 07master:778bac278869: adpcm: use av_clip_intp2()
[19:20:16 CEST] <cone-200> ffmpeg 03James Almer 07master:d8c52f5d04fc: proresenc_anatoliy: use put_sbits()
[19:37:00 CEST] <BtbN> Which RGB pixel format is the most common?
[19:38:42 CEST] <nevcairiel> bgr0/bgra probably, since that maps to windows rgb32 format =p
[19:39:22 CEST] <cone-200> ffmpeg 03Andreas Cadhalpun 07master:ef16501aebed: alsdec: ensure channel reordering is reversible
[19:56:11 CEST] <BtbN> What's the difference between get_video_buffer and filter_frame in an AVFilterPad? Some filters seem to implement both and the the same work in each, others only implement one of them.
[20:29:36 CEST] <wm4> just pick what seems simpler
[20:30:01 CEST] <wm4> I think there was a lot of microoptimization in lavfi that made everything more complex, though most is gone now
[20:30:08 CEST] <wm4> (like the awful buffer permissions stuff)
[20:32:46 CEST] <thardin> what I know of lavfi is that it's hard to use for simple stuff like just filtering a buffer
[20:32:50 CEST] <thardin> or has that improved?
[20:33:37 CEST] <Daemon404> no.
[20:39:23 CEST] <wm4> thardin: well, it's like a FIFO, so yes you don't just call a function which filters an AVFrame and returns it immediately
[20:39:41 CEST] <wm4> the most annoying thing about lavfi is all the boiler plate needed to set it up, though
[20:45:57 CEST] <rcombs> s/lavfi/all ffmpeg libs/
[20:53:28 CEST] <Daemon404> ^
[20:54:33 CEST] <BtbN> Hm, does YUV even carry any transparency information?
[20:54:46 CEST] <nevcairiel> there is YUVA formats
[20:55:05 CEST] <BtbN> But they don't seem to be present in ffmpeg, specialy not for the common YUV420 formats
[20:55:36 CEST] <nevcairiel> YUVA420 should exist, afaik
[20:56:14 CEST] <BtbN> AV_PIX_FMT_YUVA420P indeed
[20:56:38 CEST] <BtbN> I wonder if the transparency information in that is propperly parsed, by for example the overlay filter.
[20:57:22 CEST] <BtbN> looks like it does
[20:57:43 CEST] <cone-200> ffmpeg 03Michael Niedermayer 07master:7e104647a355: avcodec/alsdec: Use av_mallocz_array() for chan_data to ensure the arrays never contain random data
[21:11:05 CEST] <thardin> wm4: yeah
[21:11:42 CEST] <thardin> because what I've found is I just want a function like filter(inbuffer, outbuffer, width, height, pix_fmt, "filtername")
[21:12:14 CEST] <cone-200> ffmpeg 03Andreas Cadhalpun 07master:4c2b88678b43: alsdec: limit avctx->bits_per_raw_sample to 32
[21:16:22 CEST] <nevcairiel> unfortunately such simple filter functions dont work for quite a bunch of filters, so any generic filtering framework will need a M:N in/out FIFO-like setup
[21:18:20 CEST] <wm4> but it'd be possible to write such a function
[21:19:04 CEST] <wm4> why can't lavfi just take all the parameters from an AVFrame...
[21:21:08 CEST] <cone-200> ffmpeg 03Thomas Guillem 07master:b8d7f3186e86: matroskadec: fix crash when parsing invalid mkv
[21:21:09 CEST] <cone-200> ffmpeg 03Michael Niedermayer 07master:f98cad87590e: Merge commit 'b8d7f3186e86234f6255f5e8ee9e98573b4d9a6e'
[21:27:06 CEST] <nevcairiel> doesnt it mostly do that these days
[21:31:17 CEST] <nevcairiel> although for building the graph you need to know some information sometimes
[21:54:50 CEST] <cone-200> ffmpeg 03Paul B Mahol 07master:daf73d5d1701: avcodec/g729dec: support decoding packets with multiple frames
[21:54:51 CEST] <cone-200> ffmpeg 03Paul B Mahol 07master:df7c2b2f1b75: avformat/riff: support G729A
[22:09:31 CEST] <BtbN> Any comments on this? https://github.com/BtbN/FFmpeg/commit/8b5d20f5cdfc71749ac21c3e5a52a30e0aa92f2e
[22:09:42 CEST] <lglinskih> kierank, wm4: do I really need to use framecrc, or running my program and using its return code as a result of a fate-test is okay?
[22:09:43 CEST] <BtbN> vf_colorkey filter
[22:10:03 CEST] <BtbN> Propably far from optimal, but it works and should get things started
[22:10:13 CEST] <llogan> users ask for that one on occassion
[22:10:41 CEST] <BtbN> yep
[22:10:56 CEST] <BtbN> Well, they more specificly ask for chroma key, which is what lets you do greenscreens.
[22:11:03 CEST] <BtbN> This is just simple color key in RGB.
[22:11:48 CEST] <BtbN> But i think that's not too complicated to do, will look into it tomorrow.
[22:12:00 CEST] <BtbN> There are quite a few diffrent algorithms on the net.
[22:17:36 CEST] <wm4> lglinskih: fate tests can be all different
[22:17:44 CEST] <wm4> framecrc is just a helper used by some tests
[22:17:58 CEST] <wm4> what you want should work
[22:18:11 CEST] <wm4> fate just needs to be extended
[22:19:02 CEST] <wm4> see fate-run.sh
[22:19:10 CEST] <wm4> or was it fate.sh
[22:19:17 CEST] <Timothy_Gu> fate-run.sh
[22:19:34 CEST] <Timothy_Gu> fate.sh is for submitting results to fate.ffmeg.org etc.
[22:19:50 CEST] <Timothy_Gu> and yes documentation on this area is sorely missing
[22:21:16 CEST] <llogan> Timothy_Gu: perhaps you can persuade a teacher that it would be an acceptable assignment for "technical writing".
[22:21:35 CEST] <llogan> grades4code
[22:24:51 CEST] <jamrial> lots of fate tests just run a program and its return value defines if the tests passes or not
[22:25:04 CEST] <jamrial> you don't need to expand fate-run.sh for that
[22:25:29 CEST] <jamrial> see for example libavcodec.mak
[22:25:54 CEST] <jamrial> cabac-test, golomb-test, dct-test
[22:26:07 CEST] <lglinskih> wm4, Timothy_Gu: I've already read it, and want to know is it a good way, before I implement it =)
[22:28:39 CEST] <wm4> lglinskih: I think possibly the tests should be a bit more flexible (like taking input arguments and printing results, and this can be used in fate itself to test multiple interesting cases), but you probably don't have to worry about such ideas yet
[22:33:48 CEST] <llogan> damn it. forgot version.
[22:37:44 CEST] <Timothy_Gu> llogan: lol, or maybe I  can just write it myself.
[22:38:41 CEST] <llogan> you have to make the system work for you
[22:43:01 CEST] <Timothy_Gu> llogan: thing is the student would probably have to take more time trying to understand the FATE system first than writing about it
[22:45:35 CEST] <llogan> Timothy_Gu: i meant that you would write it, and it can be considered homework or whatever
[22:54:45 CEST] <Timothy_Gu> llogan: oh haha i don't think my english classes have such assignments, at least not right now
[23:09:04 CEST] <cone-200> ffmpeg 03James Almer 07master:5cd216934406: wavpackenc: use put_sbits()
[23:42:52 CEST] <Timothy_Gu> llogan: 1-hour attempt: http://imgh.us/FATE.svg
[23:56:45 CEST] <llogan> Timothy_Gu: it looks nice, but it confuses me, and some of the boxes look too small for my ugly fonts: http://avmule.com/junk/fate.png
[00:00:00 CEST] --- Sun Apr 19 2015