[Ffmpeg-devel-irc] ffmpeg-devel.log.20141013

burek burek021 at gmail.com
Tue Oct 14 02:05:02 CEST 2014


[00:05] <BBB> jamrial: thanks! patch is fine I suppose
[00:45] <cone-922> ffmpeg.git 03James Almer 07master:ccd04f43201d: fate: add test for vp90-2-trac3849.webm
[00:51] <jamrial> ^ i agree. look at these results i got with crypto-bench the other day: http://pastebin.com/0WU9utym
[00:51] <nevcairiel> gcrypt probably uses it, huh
[00:51] <jamrial> yeah
[00:52] <jamrial> hence the massive difference compared with lavu and openssl
[00:52] <jamrial> also, nice to know i wrote the fastest ripemd160 implementation :P
[00:56] <kierank> in b4 Compn says aesni is nsa controlled
[01:03] Action: michaelni kicks his ISP
[01:08] <cone-922> ffmpeg.git 03Martin Storsjö 07master:dbb472cb2f2f: movenc: Write edit lists for fragmented files as well, if necessary
[01:09] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:ca5a77191b79: Merge commit 'dbb472cb2f2f799295a12b4922a6a8be2cccfdee'
[01:16] <cone-922> ffmpeg.git 03Martin Storsjö 07master:dad12ce452a9: movenc: Print a warning for an unhandled case of nonzero start dts with empty_moov
[01:16] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:20a66868a25b: Merge commit 'dad12ce452a9d69c0d9d53c375003947d5f1b02e'
[01:21] <cone-922> ffmpeg.git 03Martin Storsjö 07master:e44ee1eb8db7: movenc: Simplify code by using an existing local pointer
[01:22] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:094b3ce8356a: Merge commit 'e44ee1eb8db7393e9d43207c2e1812720e292e6d'
[01:33] <cone-922> ffmpeg.git 03Mark McGough 07master:76c70e33d224: icecast: Do not use chunked post
[01:34] <Compn> kierank : i dont know what is nsa controlled. 
[01:34] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:25a418082597: Merge commit '76c70e33d2244a688832f03b53862eb5d9ad3b01'
[01:34] <Compn> i havent followed crypto backdoors
[01:58] <J_Darnley> While kierank makes a good point that aes-ni is probably useful, how many people have them?  How many opw applicants?
[01:58] <kierank> aes-ni is quite old
[01:58] <kierank> 4 years old now
[01:59] <J_Darnley> Is it that long?
[01:59] <J_Darnley> (damn)
[02:02] <J_Darnley> If it does get added as a task, it should have requirements that they have a suitable cpu...
[02:02] <J_Darnley> or if someone volunteers ssh access that they are familiar with using it already.
[02:19] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:7a11333387b2: libpostproc: support grayscale
[02:19] <cone-922> ffmpeg.git 03Michael Niedermayer 07master:b8c50becc8cf: avfilter/vf_pp: support AV_PIX_FMT_GRAY8
[02:22] <smarter> not sure what could be backdoored, aes is deterministic
[02:26] <J_Darnley> like this: cp private_key /nsa/storage
[02:37] <rcombs> kierank: I have an AES-NI-capable CPU and am bored at the moment; got a sample command/file that could use it?
[02:49] <pross> ls: cannot access /nsa/storage: No such file or directory
[02:54] <rcombs> pross: that's what they *want* you to think! :P
[10:25] <benoit-> hi
[10:29] <thardin> gah, outgoing mail address is different from ingonig
[10:31] <thardin> michaelni: you can drop the tomhar@ mail
[10:33] <thardin> ah, too late. oh well
[12:31] <chris1988> hello to everyone, i have one question, i am trying to load the remove logo filter but i get anytime the same error... i tried many differect ways but nothing, the same thing here is how i tried to load it: -vf "remove-logo" or -vf "remove_logo" or -vf "removelogo" but the output is  No such filter
[12:37] <cone-436> ffmpeg.git 03Mika Raento 07master:a59808dd57f8: mov.c: allow reading fragment start dts/pts from fragmented mp4
[12:58] <J_Darnley> chris1988: the obvious next step if to look at your version, run ffmpeg -filters, and head on over to #ffmpeg
[14:18] <amalia> Hello
[14:19] <amalia> I would like to participate in the OPW program with ffmpeg this year.I am interested in working on the mpeg-4 audio lossless coding encoder project.
[14:20] <Compn> great amalia :)
[14:20] <Compn> register on the wiki and put your name in as being interested to work on that project
[14:20] <amalia> I am currently looking at the qualification tasks
[14:21] <amalia> Thanks very much Compn How do I register on the Wiki ?
[14:21] <Compn> amalia : https://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12
[14:21] <Compn> amalia : here to register https://trac.ffmpeg.org/register
[14:21] <amalia> Thank you Compn
[14:22] <amalia> You are very kind
[14:22] <Compn> i think theres only about 8 days left to get a qualification task and send in an application ? 
[14:23] <Compn> we've gotten a lot of applicants so far, just letting you know :)
[14:23] Action: Compn afk
[14:23] <Compn> of course, more interested people, the better :)
[14:23] <amalia> Must I be doing Computer Science studies to participate in the OPW  ? I read Management studies but I code in C/C++ sas my Dad was a Computer Science lecturer 
[14:24] <J_Darnley> I doubt it.  You just need some programming experience.
[14:25] <amalia> he really encouraged all of us (siblings) to learn how to program computers. 
[14:25] <amalia> Compn  Must I complete the qualifications tasks to get selected ?
[14:26] <cone-436> ffmpeg.git 03Paul B Mahol 07master:a586b3d9b1df: libavcodec/libtwolame: fix null pointer dereference
[14:26] <J_Darnley> I didn't even think you had to be a university student for this.
[14:26] <Compn> amalia : no, but people who have completed tasks might be favored over people who have incomplete tasks...
[14:27] <Compn> amalia : you dont need CS learning to apply, but writing an audio encoder is fairly complicated and deals with a lot of algorithms and math. if you are good with those things, thats a bonus
[14:28] <amalia> Hmm. I am good at Coding and Algorithms too
[14:28] <Compn> oh great then :)
[14:28] <amalia> I pprogram for fun but study Management at the University
[14:28] <Compn> ffmpeg is written in C (no c++) and a lot of asm
[14:29] <Compn> and of course some simd (cpu instructions)
[14:29] <Daemon404> simd is asm.
[14:29] <Compn> ah
[14:29] <Compn> shows how much i know
[14:33] <amalia> You really know a lot - Cause I don't know what asm is. Could you help mentor me on this project Compn ?
[14:33] <Compn> no, sadly i dont know how to program :)
[14:33] <Compn> nor am i good with maths or algorithms.
[14:33] <amalia> Really ? Compn
[14:33] <J_Darnley> 1+1=3?
[14:34] <amalia> Oh... And I was beginning to.......
[14:34] <Compn> amalia : but there are a lot of smart people in here ready to mentor people
[14:34] <Compn> i'm just here on irc to let people know , that the smart people are probably asleep right now and you should stick around to get more answers later :)
[14:35] <amalia> Okay. But are they rude ? Cause' I heard programmers are very rude people
[14:35] <ubitux> there are some very small c++ layers in ffmpeg, asm is assembly (cpu instructions), and the asm is using SIMD instructions which are instructions for "parallelism"
[14:35] <Compn> haha , no. i dont think theres any rude people here, luckily
[14:35] <wm4> Compn: I'm rude
[14:35] <Compn> yes mr wm4 is rude, but i dont think hes a mentor
[14:36] Action: wm4 kicks Compn 
[14:36] <Compn> amalia : the two mentors on the mpeg4 als encoder are very nice, thilo and stefano
[14:36] <amalia> I'm beginning to get scared of wm4 and other rude developers
[14:37] <ubitux> i heard wm4 also eats children sometimes
[14:37] <J_Darnley> I suggest you email them directly for help on your qualification task
[14:37] <Compn> yes please email them directly, they are not on irc as much
[14:37] <J_Darnley> or even just to ask for one.
[14:38] <J_Darnley> We here can help you with generic ffmpeg-related problems
[14:38] <Compn> i think there is also an OPW ffmpeg mailing list to ask questions
[14:38] <J_Darnley> getting the source, setting up the environment, building...
[14:39] <amalia> Thanks J_Darnley That's sweeter than eating up children like wm4 and ubitux
[14:39] <ubitux> :D
[14:40] <J_Darnley> Oh no.  I doo that too.
[14:40] <Daemon404> certain types of children are delicious. i love veal.
[14:40] <Compn> programmers like to make jokes, thats for sure
[14:41] <benoit-> Compn: that's not that funny when it's in their code, though :)
[14:41] <superware> I have an input AVStream (h.264 over UDP), and I want to copy it to a new output context (mp4), how should I setup the output stream time_base so the recording to a file will be timed according to the input?
[14:41] <Compn> isnt there a copytb or copydts option somewhere ?
[14:42] <Compn> also that looks like an #ffmpeg question probably
[14:42] Action: Compn goes afk
[14:42] <superware> already asked there, no reply
[14:42] <superware> copytb?
[14:43] <Compn> superware : its -copyts , read the manual https://www.ffmpeg.org/ffmpeg.html
[14:43] <Compn> and copytb is there too
[14:44] <superware> I'm using libav*
[14:44] <Compn> well you can look at ffmpeg.c source and find out which av_opt that copytb and copyts use, then copy that for your code ?
[14:45] <Compn> maybe
[14:46] <Compn> maybe we should make a version of the manual which combines command line options and av_options so that libav* devels can easily find features... ?
[14:47] <amalia> J_Darnley What about black kids ? Are they delicious too ?
[15:47] <ubitux> michaelni: fate is not happy with the recent pp changes
[15:47] <ubitux> http://fate.ffmpeg.org/report.cgi?time=20141013054115&slot=x86_64-archlinux-gcc-valgrindundef
[16:18] <cone-436> ffmpeg.git 03Michael Niedermayer 07master:57884172e0c7: avformat/mov: fix mix of declaration and statement
[16:18] <cone-436> ffmpeg.git 03Michael Niedermayer 07master:0b7e5d0d75e7: postproc: fix qp count
[16:18] <michaelni> ubitux, should be fixed
[16:20] <ubitux> cool, thx
[16:42] <ubitux> http://jasmcole.com/2014/10/12/rolling-shutters/
[16:46] <wm4> fascinating
[16:53] <Compn> whuh thats strange :)
[16:53] <Compn> in the future there will be two lenses taking images top first and bottom first then combining :P
[16:53] <Compn> i like them trying to piece it back together
[17:01] <iive> Compn: it's not the lenses that are the problem.
[17:02] <iive> one way to fix this is by having a shutter in front of the camera. so you open it, let the light be detected and the close it. then the sensor could be safely scanned.
[17:02] <iive> the shutter could be done by liquid crystals, to avoid mechanical parts. 
[17:03] <iive> (like LCD)
[17:03] <iive> of course this won't help much when capturing video, or maybe it could?
[17:08] <wm4> isn't LCD generally too slow to switch instantly between full and no opacity at such a fast rate
[17:09] <nevcairiel> yeah a black/white LCD switch is still quite slow
[17:10] <nevcairiel> although i have no idea which kind of timing would be required for this use-case
[17:34] <J_Darnley> iive: did you just reinvent the SLR?
[17:34] <iive> I guess
[17:36] <iive> hum.. no, single len reflex is not related.
[17:37] <iive> lcd monitors manage 60Hz, so having 25-30fps should be feasible. 
[17:38] <nevcairiel> there is a bit of motion blur on lcd monitors tho, which comes from slow crystals
[17:40] <iive> yes, also, you need first to open the shatter , let some light expose on the sensor, then close the shatter, then start scanning the charged sensors
[17:40] <iive> less light -> more (electric) noise
[18:05] <hima> ubitux : hi! my kate probe and get_pts fucntions are ready! Could you please explain the read header function in details 
[18:05] <hima> so that i can write it a bit efficiently 
[18:09] <hima> here is my kate_probe() and get_pts() please have a look http://fpaste.org/141475/13216531/
[18:13] <hima> if anyone could guide me with the read_header() function in the demuxer. As in what is the main aim and what are we supposed to do. i just need a basic understanding.
[18:16] <cone-436> ffmpeg.git 03Benoit Fouet 07master:4da7111eb838: avcodec/aacdec: map LFE[0] to SCE[1] for 4.0 audio.
[19:56] <akira4> ubitux, I was testing the stl demuxer with this file http://nothingcantouchme.com/downloads/Subtitles/UK_FINAL_240511.stl 
[19:56] <akira4> I get a low score for the probe function again
[19:56] <akira4> here is the diff
[19:57] <akira4> http://pastebin.com/0F2xpqTQ
[19:59] <ubitux> akira4: can you print the probe buffer?
[20:02] <akira4> how do I print the buffer?
[20:12] <kierank> ffmpeg has an stl demux?
[20:12] <nevcairiel> kierank: akira4 is working on it
[20:12] <kierank> ah
[20:15] <ubitux> akira4: av_log(0,0,"buf=[%s]\n", ptr)
[20:24] <akira4> ubitux, how is the probe function called?
[20:25] <kierank> ffmpeg calls it somewhere
[20:33] <kierank> ah you're doing spruce stl
[20:33] <kierank> not ebu stl
[20:34] <akira4> yup
[20:37] <Daemon404> that format looks gross...
[20:40] <akira4> I tried using av_log() and redirected it to a text file using ./ffmpeg -i UK_FINAL_240511.stl  Uk1.srt >logfile.txt 
[20:40] <akira4> is this the correct way?
[20:41] <J_Darnley> No because ffmpeg prints to stderr
[20:41] <J_Darnley> That should be 2>
[20:41] <akira4> oh. Thanks
[20:50] <akira4> ubitux, I used av_log() but nothing seems to get printed. Not sure if  implemented it correctly...link to the probefunction http://pastebin.com/QrpCMvpA
[21:14] <sectroyer> Hello. I don't know if you are aware but Samsung has violated your LGPL license in their Smart TVs and basically stole your code :)
[21:15] <llogan> thardin: do you want me to approve or discard your message in -devel?
[21:15] <llogan> oh, it's a dupe. i'll discard.
[21:17] <sectroyer> I have send an email to them informing about the issue and after weeks of exchanging mails they did admit doing that but it looks they are not willing to release the source code.
[21:17] <BtbN> samsung even has http://opensource.samsung.com
[21:17] <sectroyer> BtbN: Yeah and search there for SDAL
[21:18] <sectroyer> they released only two files containg your code 
[21:18] <sectroyer> now they are moving it to seperate lib: libtsmutex.so but for years they have distributed as a part of much bigger lib
[21:19] <sectroyer> and according to LGPL license they have to release the source code for the parts that they did distribute and not for the parts they want to do it :)
[21:19] <BtbN> what?
[21:19] <sectroyer> It looks they have been doing it for 2-3 years...
[21:20] <BtbN> libSDAL doesn't look like ffmpeg and also not like it's using it.
[21:21] <sectroyer> BtbN: look at tsmutext source cod efiles
[21:21] <sectroyer> using your code or not is not an issue
[21:21] <sectroyer> they admited doing that :)
[21:21] <sectroyer> the issue is this:
[21:21] <sectroyer> "But when you distribute the same sections as part of a whole which is a work based on the Library, the distribution of the whole must be on the terms of this License, "
[21:21] <sectroyer> from lgpl :) They ave to release whole libSDAL at terms of LGPL :)
[21:22] <sectroyer> They know that just try to avoid it at any cost...
[21:22] <sectroyer> But that's illegal
[21:23] <BtbN> No they don't. They have to give you a way to replace the (L)GPL licensed parts. Which they do by giving out the object files of their stuff, so you can re-link it.
[21:24] <sectroyer> from lgpl: " Pay close attention to the difference between a "work based on the library" and a "work that uses the library". The former contains code derived from the library, whereas the latter must be combined with the library in order to run."
[21:24] <sectroyer> :)
[21:24] <sectroyer> it's "work based on the library"
[21:24] <BtbN> The whole point of LGPL is that you don't have to do exactly that.
[21:24] <sectroyer> for every work "based on the library: " You must cause the whole of the work to be licensed at no charge to all third parties under the terms of this License."
[21:25] <sectroyer> they broke your guidlines, didn't link with libffempeg, etc.
[21:25] <sectroyer> since they didn't link with your code it becomes "work based on the library"
[21:25] <sectroyer> copying and pasting lgpl code isn't "allowed" by lgpl license :)
[21:26] <BtbN> I don't see any copy&pasted code in there
[21:26] <BtbN> i hardly see code at all, just object files for re-linking
[21:26] <sectroyer> see two source code files :)
[21:26] <sectroyer> that's Your code :)
[21:27] <sectroyer> /libSDAL_F/Coral/Common/Spi/Swtsm/tsmuxenc.*
[21:27] <sectroyer> :)
[21:28] <sectroyer> in this "archive" they pretend your code is used in seperate library. Which containc ONLY your code
[21:28] <sectroyer> which is allowed by lgpl but they distributed it with libSDAL as a whole
[21:28] <sectroyer> Only now they try to seperate lgpl code because it was spotted that they violated lgpl :)
[21:29] <sectroyer> When I asked by email they also pretendted that your code is not part of libSDAL as is only linked with lib (libtsmuxenc.so) which contains your code
[21:30] <sectroyer> when I pointed to them that for years they have been distributing it differently (whole libSDAL with your code compiled in) I got excuses and finally silence...
[21:30] <sectroyer> They know what they did and they know it's illegal...
[21:31] <Daemon404> standard bigcorp
[21:31] <sectroyer> They often do stuff like that, same was with ExFAT...
[21:32] <thardin> llogan: yeah, I had the wrong From: configured
[21:32] <sectroyer> That's why I wanted inform you. You are the "rights holders". AFAIK you can ask FSF to intervene
[21:33] <Daemon404> they cant do anything
[21:33] <Daemon404> they have no copyright.
[21:34] <sectroyer> well somebody can :)
[21:35] <sectroyer> I think in USA there are lawyers who sue companies for this kind of stuff:)
[21:35] <sectroyer> I think that's how ExFAT source code was released...
[21:38] <llogan> sectroyer: maybe reynaldo knows what's going on
[21:40] <sectroyer> llogan: Well I just wanted to contact you to shine some light on the matter :)
[21:41] <sectroyer> Let me know if you want more information.
[21:42] <sectroyer> Besides if they did "everything right" they wouldn't lie on every single step that they DON'T link directly with your code ;)
[21:42] <sectroyer> also in "object files" that they released that they started major code rewrite... Steal code than panic ;)
[21:44] <sectroyer> Okay I have to go, you want some more info? 
[22:12] <cone-448> ffmpeg.git 03Benoit Fouet 07master:f87134c7a13f: avformat/movenc: add support for syncframes concatenation for E-AC-3.
[22:49] <akira4> ubitux, in the av_log() function the first parameter is a void pointer. Shouldn't we pass NULL(or the context) instead of 0? 
[22:49] <ubitux> NULL and 0 are the same
[22:50] <ubitux> it's just faster to type av_log(0,0,
[22:50] <ubitux> you won't keep it in the final code anyway
[22:50] <ubitux> what do you mean nothing seems printed?
[22:50] <ubitux> the buffer is empty?
[22:50] <ubitux> the function is not called?
[22:51] <akira4> I dont get any output in the redirected file except the usual stuff that is printed 
[22:51] <akira4> do I need to loop the av_log or just using it once is enough?
[22:52] <ubitux> just once
[22:52] <ubitux> it should be called automatically
[22:52] <ubitux> if nothing is printed, you're probably doing it wrong
[22:52] <ubitux> it should be printed in red
[22:53] <akira4> http://pastebin.com/QrpCMvpA is how I  used it
[22:53] <ubitux> yes
[22:53] <ubitux> it's correct afaict
[22:54] <akira4> and I used this command ./ffmpeg -i UK_FINAL_240511.stl  Uk1.srt 2>~logfile.txt
[22:54] <akira4> ah sorry its ~/logfile.txt
[22:55] <ubitux> well
[22:55] <ubitux> does it work without the redirection first?
[22:56] <akira4> it didnt. so I tried with the redirection
[22:56] <ubitux> is the probe function referenced in your file?
[22:57] <ubitux> didn't you actually recompile ffmpeg?
[22:57] <ubitux> i need to sleep, you're on your own for a few hours
[22:57] <ubitux> see you later, gl
[22:57] <akira4> cool. Thanks anyways.
[00:00] --- Tue Oct 14 2014


More information about the Ffmpeg-devel-irc mailing list