[Ffmpeg-devel-irc] ffmpeg-devel.log.20150106

[burek]

[00:35] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:42411a85b78c: avfilter/vf_yadif: detect telecine content
[01:26] <nevcairiel> videolan git down again? at least the website seems unreachable
[01:33] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:a79ac73b631a: avfilter/vf_idet: Add analyze_interlaced_flag mode
[02:17] <jamrial> the repository seems to work. it's just the website
[05:08] <compn> wow
[05:08] <compn> amd getting in on that opencl action
[05:09] Action: compn remembers everyone hating on opencl support
[05:12] <jamrial> well, it is kind of broken, in a way
[05:13] <jamrial> works wonders for lavfi filters, but the installed header is broken for API users
[05:16] <jamrial> since it tries to #include config.h :p
[05:18] <compn> :p
[05:18] <compn> there is only ffmpeg, there is no api users! :P
[05:18] <compn> is/are
[05:44] <cone-342> ffmpeg.git 03Dale Curtis 07master:db42d93a61be: mov: Fix overflow and error handling in read_tfra().
[05:44] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:3859868c7531: avformat/mov: fix integer overflow in mov_read_udta_string()
[05:44] <cone-342> ffmpeg.git 03Dale Curtis 07master:3ebd76a9c575: mov: Fix negative size calculation in mov_read_default().
[05:56] <jamrial> there, patch sent
[08:13] <wm4> kierank: in addition to mplayer, there's also szap-s2 (some dvb tuning utility), which will result in ts streams without PMT
[10:21] <ffbuilder> Hi
[10:27] <durandal_1707> ubitux: see f929ab0569ff31ed5a59b0b0adb7ce09df3fca39 for dvdsubdec and first hunk
[10:28] <ubitux> ah, dvdsub
[10:28] <ubitux> yes it looks wrong
[10:29] <ubitux> but it's changed
[10:29] <ubitux> i mean, here at least
[10:30] <nevcairiel> it was fixed in the merge apparently
[10:40] <ffbuilder> I have question about libfdk-aac
[10:42] <ffbuilder> If libfdk-aac build with --enabled-shared, no license problem on ffmpeg?
[11:11] <rcombs> you'll get nonredistributable binaries
[11:12] <cone-150> ffmpeg.git 03Paul B Mahol 07master:145a84717b62: lavfi: check av_strdup() return value
[11:15] <ffbuilder> hm.. thanks i will find non redistributable 
[11:16] <ubitux> > The 1.02 scale factor is black magic, but produces correct results. I don't know what nvenc is doing
[11:16] <ubitux> +    nvenc_ctx->nvenc_cfg.darHeight    = 1024 * avctx->height * 1.02;
[11:16] <ubitux> haha wat :D
[11:17] <wm4> lol
[11:18] <wm4> darHeight is float?
[11:23] <ubitux> wm4: looks like it's u32
[11:25] <wm4> weird
[11:26] <ubitux> maybe it should be 1.024
[11:39] <nevcairiel> i imagine the 1024 is there to make  it kind-of fixedpoint
[11:40] <nevcairiel> both factors in the AR contain 1024, so it just scales the value up, doesnt change it
[11:51] <ubitux> Past duration 1.000000 too large
[11:51] <ubitux>     Last message repeated 4273 times
[11:51] <ubitux> mmh.
[11:58] <saste> ubitux, do you know how to add duration information in AVFrame?
[11:58] <saste> right now we have pts, pkt_pts, and pkt_duration
[11:59] <ubitux> no idea
[11:59] <saste> in lavfi we only use the pts information, and we ignore pkt_pts
[11:59] <ubitux> cast it in AVSubtitle
[11:59] <ubitux> @_@
[11:59] <ubitux> but yeah no, i have no idea
[11:59] <ubitux> you usually just diff with the next frame
[12:00] <saste> ubitux, yes, but sometimes it's usefult to have the duration information
[12:00] <saste> it is a main limitation at the moment
[12:00] <wm4> if you add duration, be sure not to make it int
[12:01] <wm4> and to make a difference between 0 duration and no duration
[12:01] <wm4> 2 past API mistakes
[12:01] <saste> especially considering that filters such as setpts doesn't know the next PTS and can't cache
[12:01] <wm4> though in general I'd consider the duration info lavf outputs useless
[12:01] <wm4> better don't rely on it?
[12:01] <wm4> the buffersink filter could emulate it
[12:02] <nevcairiel> yeah duration info for video is 90% useless
[12:02] <nevcairiel> and for audio its already defined
[12:02] <nevcairiel> (nb_samples and samplerate, etc)
[12:02] <saste> nevcairiel, why, because it is unreliable?
[12:02] <wm4> I wonder when will they invent variable samplerate audio
[12:02] <nevcairiel> wm4: you mean aac ssr?
[12:03] <wm4> really? stupid reality
[12:03] <nevcairiel> saste: unreliable, inexistant, hence not useful
[12:03] <nevcairiel> the only w ay to get reliable duration info is by diffing the frames
[12:03] <nevcairiel> frames timestamps*
[12:05] <nevcairiel> wm4: well technically it uses different sample rates for different frequency bands, somewhat similar to how opus splits music and speech coding .. but yeah.
[12:05] <nevcairiel> a decoder would probably just resample that to the highest possible rate
[12:06] <nevcairiel> luckily aac ssr is barely used anywhere
[12:06] <wm4> using different samplerates for different channels would also be a great way to break absolutely everything
[12:07] <ubitux> do we support odd sample rates?
[12:09] <mattlea> Hi all, I'm trying to create a filter that does a FFT followed by an IFFT, but I get horrible artifacts.  Here's the code: http://pastebin.com/7M5ma2hf
[12:13] <mattlea> If anyone has any hints that'd be great :)
[13:38] <ubitux> mattlea: your query_format is correct?
[13:40] <ubitux> mattlea: also use the public api (av_fft_init()) and check return value
[13:41] <ubitux> same for other functions, just use the public api
[13:41] <mattlea> ubitux: ah, yep, I'm using floats, I need to cast extended_data
[13:41] <mattlea> ubitux: will do, thanks!
[13:41] <ubitux> what did you request in query_format?
[13:43] <mattlea> AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE
[13:46] <ubitux> ok
[15:28] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:62eafb6a0dd6: avformat/mov: Clear array to prevent potential out of array read from av_dlog()
[15:28] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:383c6a3a07b7: avformat/mov: use av_freep() to avoid leaving stale pointers in memory
[15:28] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:6e70e4aca506: avformat/matroskadec: Use av_freep() to avoid leaving stale pointers in memory
[15:28] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:59ca6e258694: avformat/hlsproto: Use av_freep, to avoid leaving stale pointers in memory
[15:29] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:da0dadbee4e9: avfilter/buffer: use av_freep() to avoid leaving stale pointers in memory
[15:29] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:bbfca8e84b0e: avformat/utils: Clear pointer in ff_alloc_extradata() to avoid leaving a stale pointer in memory
[15:29] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:369b4cd4120b: cmdutils: Use 64bit for file size/offset related variable in cmdutils_read_file()
[17:41] <user123> Hello
[17:41] <user123> I have some question
[17:42] <compn> user123 : then ask them.
[17:43] <compn> this is the developer channel, if you have user questions ask in #ffmpeg :)
[17:43] <user123> I want build ffmpeg with libfdk_aac
[17:43] <compn> ok
[17:43] <user123> I heard libfdk_aac have license problem
[17:43] <kierank> I don't understand how the yadif telecine stuff works
[17:44] <kierank> it just detects soft pulldown?
[17:44] <user123> Somepeople said if distribute with enable-shared, then no license issue
[17:44] <kierank> user123: wrong
[17:44] <user123> oh
[17:44] <user123> wrong?
[17:44] <compn> there is a license conflict
[17:44] <compn> i dont remember the details however.
[17:44] <compn> kierank : you dont understand if it works, or how it works ?
[17:44] <kierank> compn: both
[17:45] <iive> yadif have inv-telecine support?
[17:45] <user123> Also this is license conflict right ? http://oss.netfarm.it/mplayer-win32.php
[17:45] <kierank> yes
[17:46] <iive> doesn't look like it.
[17:46] <compn> kierank : you are referring to avfilter/vf_yadif: detect telecine content ?
[17:46] <user123> If purely for free application, also license conflict?
[17:46] <kierank> compn: yes
[17:47] <kierank> user123: yes licence conflict
[17:47] <user123> ugh.. ok thank you answer... 
[17:47] <compn> kierank : user123 is wondering if theres a way to build against libfdk but not distribute the dll and let the user download the .dll himself ?
[17:47] <cone-150> ffmpeg.git 03wm4 07master:ea7af58fc6b3: avformat/id3v2: support USLT tags
[17:47] <kierank> compn: dunno
[17:48] <compn> user123 : you may want to ask the other builders of ffmpeg (like zeranoe or sherpya ) about libfdk-aac
[17:48] <compn> and how they handle it
[17:48] <compn> or ask on mailinglist
[17:48] <compn> or stick around here
[17:48] <iive> https://trac.ffmpeg.org/wiki/Encode/AAC#fdk_aac
[17:49] <iive> I would speculate that it conflicts with gpl, but it "might" be ok with LGPL
[17:49] <user123> for example... I just distribution ffmpeg (enabled-libfdk aac and shared) exclude fdk aac dll file
[17:50] <user123> then no problem?
[17:50] <Daemon404> iive, it is not
[17:50] <Daemon404> i have confirmed with SFLC in the past
[17:50] <Daemon404> it conflicts with all LGPL and GPL
[17:50] <compn> kierank : i think the telecine commit only 'fixes' the sample in that it detects telecine, and then disables yadif , http://www.mail-archive.com/ffmpeg-devel@ffmpeg.org/msg07592.html
[17:50] <iive> x264 is gpl, so you might not have both at the same time.
[17:50] <compn> but i could be misreading the commit and thread there.
[17:50] <iive> also, what Daemon404 said.
[17:51] <iive> user123: this only matters if you want to distribute the build. you can use it yourself.
[17:51] <compn> Daemon404 : bonus points if someone puts the sflc opinion in the wikipage for fdk ? :P
[17:51] <user123> hm..
[17:51] <wm4> what resolution do DVB subs have? I have a sample here that seems to use 720x576 or so, but the DVBSUB_DISPLAYDEFINITION_SEGMENT indicates 1920x1080
[17:53] <user123> some people enable libfdk-aac with shared, and build with disable-nonfree(maybe configure modified?) and don't distribution libfdk .dll file.
[17:53] <user123> it also conflicts license right?
[17:53] <compn> user123 : i think what we are saying is that you are in violation of fdk's license if you distribute fdk with gpl software.
[17:54] <compn> or gpl license, i am not sure which one conflicts...
[17:54] <wm4> ah no I was wrong
[17:54] <compn> although i dont think ffmpeg goes after such cases :P
[17:54] <user123> libfdk aac with gpl
[17:54] <wm4> sometimes DVBSUB_DISPLAYDEFINITION_SEGMENT is just missing
[17:55] <user123> ok... last question...
[17:56] <user123> I build ffmpeg with enable libfdk aac and libfdk aac with enable-shared... when I build done, and execute ffmpeg
[17:56] <user123> then get error "can not find libfdk-aac-1.dll"
[17:57] <JEEB> not in library load path then :P
[17:57] <user123> I want remove dependency
[17:57] <iive> hum... is the problem with patents?
[17:58] <user123> http://oss.netfarm.it/mplayer-win32.php - this mplayer build ffmpeg with libfdk aac, enable shared
[17:58] <Daemon404> the problem is the license has restrictions based on patents which are incompatible with (L)GPL
[17:58] <user123> but no problem when do not have libfdk-aac-1.dll
[17:58] <Daemon404> any build linking to fdk is illegal to distribute
[17:58] <Daemon404> full stop.
[17:58] <JEEB> user123, then that person has patched in a loader :P
[17:58] <user123> whats that?
[17:58] <JEEB> dynamic loading of a library
[17:58] <iive> in theory, gpl also forbids distribution of code covered by patents...
[17:59] <user123> any reference data?
[17:59] <JEEB> user123, but even that is not correct license-wise :P
[17:59] <user123> oh..
[18:00] <iive> but at least it requires that limitation to be imposed by court.
[18:00] <nevcairiel> well its rather opaque if a simple loader is actually still wrong as long as you dont distribute the libfdk binary with it .. you get into areas that noone can really verify
[18:01] <user123> aw.. ok.
[18:01] <user123> recently how about experimental aac codec? improved quality or same?
[18:01] <JEEB> I'm not sure what would be the case if you made your software use a completely LGPL-licensed lavf/lavc, and had a separate module in your application that uses fdk-aac, and you have fully aligned yourself with the fdk-aac license
[18:02] <user123> ok...hm...
[18:02] <user123> too low quality another aac codecs :(
[18:03] <JEEB> because if you can use LGPL lavf/lavc in a proprietary application I guess as long as you are adhering to the fdk-aac license you should be able to actually use it? But I don't know these things and at this point you are already getting into lulzy territory
[18:03] <JEEB> of course if you want to use LGPL lavf/lavc and corporate licensed libx264 I think you have to do something similar?
[18:03] <JEEB> because lavc libx264 wrapper makes you enable gpl
[18:04] <user123> I'm so sad hm,,,
[18:05] <wm4> blame Fraunhofer
[18:06] <nevcairiel> switch to opus
[18:06] <nevcairiel> :D
[18:06] <user123> yea..
[18:06] <user123> opus?
[18:06] <user123> Opus have aac?
[18:06] <JEEB> no
[18:06] <JEEB> it was in relation to the fact that Opus was standardized for broadcast
[18:07] <JEEB> and there'll be a mp4 spec for it soon as well
[18:07] <user123> ok..  I need good quality aac codec...
[18:07] <nevcairiel> it was in relation that opus is decent quality and free to use
[18:07] <JEEB> yes, that too
[18:07] <JEEB> but if you need AAC you are pretty fucked :P
[18:07] <JEEB> I mean, the lavc encoder isn't /that/ bad any more
[18:07] <JEEB> but it's not exactly spectacular
[18:08] <JEEB> for lower rates the vo-aacenc apachev2 one might be "alright"
[18:08] <JEEB> but in general I've seen the lavc encoder beat vo-aacenc
[18:08] <nevcairiel> vo-aacenc is terrible
[18:08] <nevcairiel> as are all the other options in lavc, the native one is already the best short of fdk of course
[18:09] <nevcairiel> should poke the mega ticket again
[18:09] <JEEB> yeah
[18:09] <user123> sorry I'm beginner...
[18:09] <user123> lavc encoder is acodec=aac ?
[18:11] <JEEB> yup
[18:12] <user123> but maybe much worse than libfdk aac ? 
[18:13] <JEEB> it is worse then fdk-aac, but there is no simple way for you to actually use it :P
[18:14] <user123> yea..
[18:14] <JEEB> you can forget about it pretty much
[18:14] <nevcairiel> you can use it yourself just fine, you just cannot distribute it
[18:14] <nevcairiel> i think its weird that debian of all places actually has it packaged
[18:15] <nevcairiel> strong lobbying going on i guess?
[18:15] <user123> my object is distribute
[18:15] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:d2838f09d6dc: cmdutils.c: Use av_realloc_array()
[18:15] <user123> make good ffmpeg based encoder like megui
[18:16] <user123> I need think about use qaac with pipe
[18:16] <user123> clearly qaac better quality than libfdk-aac.. pretty many difference..
[18:17] <Daemon404> it is not better actually
[18:17] <user123> or maybe just my feel :)
[18:17] <Daemon404> some ABX tests were done and it more or less matches it
[18:17] <Daemon404> it juts sounds slightly different
[18:17] <Daemon404> but not in a better or worse way
[18:17] <Daemon404> or so im told.
[18:17] <Daemon404> i dont have the golden ears.
[18:17] <nevcairiel> qaac is at least similar,  both have pretty good behaviour, but both are practically unusable for open-source
[18:17] <user123> yea...
[18:18] <nevcairiel> psycho-acoustics are hard
[18:19] <user123> If use pipe, can't utilize with ffmpeg's trim feature (-ss , -to) . sad about that
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:05bc6f8ba653: avcodec/h264_slice: Clear table pointers to avoid stale pointers
[19:35] <cone-150> ffmpeg.git 03Carl Eugen Hoyos 07release/2.3:d4c70c8b502a: lavc/utils: Make pix_fmt desc pointer const.
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:6ed35a66745e: avcodec/options_table fix min of audio channels and sample rate
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:b89b136c001b: avcodec/utvideodec: fix assumtation that slice_height >= 1
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:66261cfa7789: avcodec/wmaprodec: Fix integer overflow in sfb_offsets initialization
[19:35] <cone-150> ffmpeg.git 03Lukasz Marek 07release/2.3:bb2c09310cbd: lavu/opt: fix av_opt_get function
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:693d0d3ac520: avformat/hlsenc: Free context after hls_append_segment
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:17ff5d3f8826: doc/APIchanges: Fix some wrong versions
[19:35] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:dca70c593178: swscale/x86/rgb2rgb_template: handle the first 2 lines with C in rgb24toyv12_*()
[19:58] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:8327bef1c920: Update for 2.3.6
[19:58] <cone-150> ffmpeg.git 03Michael Niedermayer 07release/2.3:db27f50e0658: avformat/mov: Fix mixed declaration and statement warning
[20:10] <cone-150> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n2.3.6': unknown revision or path not in the working tree.
[20:10] <cone-150> Use '--' to separate paths from revisions
[20:10] <cone-150> refs/tags/n2.3.6:HEAD: cmdutils.c: Use av_realloc_array()
[21:36] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:a0a7b154bb49: tools/ismindex: Keep count fields consistent with arrays in case of errors
[21:36] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:dede2f7fc444: tools/ismindex.c: Use av_realloc_array()
[21:44] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:07a0c0f00050: opt: check memory allocation
[21:44] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:b20e79dbe998: Merge commit '07a0c0f0005072d115ace61e60f46be68582cc3a'
[21:51] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:1dd797e3c9f1: swscale: check memory allocations
[21:51] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:90de28befd88: Merge commit '1dd797e3c9f179f957316a0becbec048b42df8aa'
[22:01] <cone-150> ffmpeg.git 03Diego Biurrun 07master:ca09effb01e1: ffv1: Drop unnecessary casts and const qualifiers to match function signatures
[22:01] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:f4cc19cdd082: Merge commit 'ca09effb01e126b0ac74ff3de70a475423ddee82'
[22:12] <cone-150> ffmpeg.git 03Diego Biurrun 07master:0352ff102d62: ffv1: const correctness for encode_rgb_frame()
[22:12] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:3b3782d74e5a: Merge commit '0352ff102d62ee94e79e0baaf64d5ad4e66f907b'
[22:12] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:9198397115bf: avcodec/ffv1enc: Fix incompatible pointer type warning
[22:31] <cone-150> ffmpeg.git 03Stefano Pigozzi 07master:2cef68da69a1: vda: error out if decoded CVPixelBuffer is empty
[22:31] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:2ab6deb36f19: Merge commit '2cef68da69a17ed09c313ba3c3850ec1cc0a80e0'
[23:40] <cone-150> ffmpeg.git 03Agostino Sarubbo 07master:fe0829986833: build: Use -Werror=format-security
[23:40] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:696141e89819: vf_interlace: use image width rather than linesize
[23:40] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:30cf755548a9: Merge commit 'fe082998683334e7f573cfe3876810fb80c5249c'
[23:40] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:f931b1ef450e: Merge commit '696141e898193311c994b399a8dc60713709092f'
[23:40] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:ac3f6429ba5b: vfilter/vf_tinterlace: Fix issues with linesize and cols
[23:51] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:046f75a97070: vf_interlace: also assert for height
[23:51] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:6279cb3597cc: Merge commit '046f75a970701b4c947d38bfd2186dcc5f2ddae2'
[00:00] --- Wed Jan  7 2015