[Ffmpeg-devel-irc] ffmpeg-devel.log.20121229

[burek]

[00:11] <ubitux> here we go.
[00:23] <ubitux> mpl2, aqt, mpsub, pjs, vplayer, subviewer1& check.
[00:26] <wm4> lol mpsub
[00:28] <ubitux> wm4 :D
[00:29] <ubitux> wm4: timed *and* frame support of course
[00:29] <wm4> how does frame timing work again with these demuxers?
[00:29] <ubitux> microdvd and mpsub frame rate is configurable within the file
[00:29] <ubitux> for aqt there is a demuxer option to set the frame rate
[00:30] <ubitux> (which default to 25 iirc)
[00:41] Action: ubitux hopes no one will bump lavc & lavf minor soon
[00:44] <ubitux> hey saste :)
[00:44] <saste> ubitux, hi
[00:45] <ubitux> saste: what did spi say about sharing donators name publicly?
[00:45] <ubitux> (for those who want)
[00:45] <saste> uhm spi...
[00:45] <saste> it says nothing, but right now the donation system is pretty inflexible
[00:46] <saste> so there is no way for the donor to specify that his donations should be advertised
[00:46] <ubitux> that's too bad; didn't you ask them for improvement on that matter?
[00:46] <saste> i suppose she can contact us directly
[00:46] <saste> but right now on the donations front we're lacking in many ways
[00:47] <saste> things to do: improve SPI blurb, fix the ffis/SPI things (which is current work in progress)
[00:47] <saste> add support to paypal, crowd-funding, etc.
[00:47] <saste> t-shirts/mugs
[00:48] <ubitux> is spi new?
[00:48] <llogan> i talked to some shirt makers
[00:48] <saste> btw i should send a report for the first semester donations in the first days of january 
[00:48] <saste> right now we have ~600$
[00:49] <michaelni> \o/ i thought it was just 500
[00:49] <llogan> that's more than i expected too
[00:49] <saste> llogan, yes considering that we're not doing anything for promoting donations that's not bad
[00:50] <llogan> VLC's conspicous yet somewhat unobtrusive donation box is nice
[00:51] <saste> llogan, what about a crowd-funding task for funding bug fixing?
[00:51] <saste> yes this is not the first time i mention it
[00:51] <saste> but could be a good experiment
[00:52] <llogan> how would it work? with what croud-funding provider?
[00:52] <ubitux> saste: do you have access to the public donators?
[00:52] <ubitux> -public
[00:52] <saste> llogan, kickstarter, indiego, one or the other
[00:53] <llogan> do you have a bug(s) in mind?
[00:53] <saste> they provide similar services and fees
[00:53] <saste> ubitux, what do you mean by "access"?
[00:53] <llogan> i know some people that used kickstarter. i can ask them about how they felt. one made 65k and one made 6 or 9k, IIRC.
[00:53] <saste> i receive a mail with the name and address of the donor
[00:53] <ubitux> saste: get the name :p
[00:53] <wm4> what will ffmpeg use this money for?
[00:54] <ubitux> oh ok
[00:54] <ubitux> wm4: pay assassins
[00:54] <ubitux> for who-you-know
[00:54] <wm4> fate instances to drive global warming?
[00:54] <llogan> wm4: i was going to embezzle it into shirts
[00:55] <saste> llogan, the idea is that we collect fund, and we allocate for example 100$ for each bug fixed
[00:55] <saste> the money goes to the fixer, even if that's going to be complicated
[00:55] <michaelni> saste, this sounds like a great idea
[00:55] <wm4> 100$ for each ffmpeg bugs? you could waste millions...
[00:56] <michaelni> we have 192 open or reopened defects on trac
[00:56] <michaelni> i think 20k is pretty cheap to fix all open bugs in ffmpeg
[00:57] <michaelni> (note  this doesnt count new bugs that arent open )
[00:57] <durandal_1707> 100$ is not for new features
[00:57] <saste> 500 if you count features
[00:57] <michaelni> there are 156 new (not open) bugs
[00:58] <durandal_1707> vlc gives 7k for new features
[00:58] <llogan> wow
[00:59] <saste> durandal_1707, which kind of new features?
[00:59] Action: llogan makes some skins
[00:59] <llogan> 3. profit
[00:59] <durandal_1707> saste: not trivial ones
[00:59] <saste> many feature request are trivial, e.g. can be implemented with a component option or such
[01:00] <durandal_1707> i started to work on dts-hd-ma but lost motivation, started work on evrc and did lot of stuff but again lost motivation
[01:01] <saste> durandal_1707, yes a bit of monetary incentive would help to finish stuff
[01:01] <wm4> evrc?
[01:01] <durandal_1707> tried to work on als, did not even got motivation to start it
[01:01] <durandal_1707> wm4: cdma codec
[01:02] <llogan> crowd-funding seems to be "event" driven, IMO. i can see working for a big, singular feature, but i don't know how to implement it best for an open-ended thing
[01:02] <llogan> example: http://www.kickstarter.com/projects/1061646928/vlc-for-the-new-windows-8-user-experience-metro
[01:03] <saste> llogan, what about "fund fixing for open tickets from 1 to 2000"?
[01:03] <saste> would that count as a valid project?
[01:03] <michaelni> i think that might be a bit hard and unpredictable
[01:04] <michaelni> too much to even read to know how much work it is
[01:05] <ubitux> we need a bounty on the trac :)
[01:05] <ubitux> bounty plugin*
[01:07] <llogan> saste: i think the crowd-funding is a good idea, and worth a try...
[01:10] <llogan> j-b: do you know who made the VLC shirts n' stuff for the kickstarter campaign?
[01:11] <michaelni> ubitux, what pluging, do you have alink ?
[01:11] <ubitux> i don't know if it exists :p
[01:12] <ubitux> i just thought it could be nice
[01:13] <j-b> llogan: noone yet
[01:14] <llogan> ubitux: use this http://en.wikipedia.org/wiki/Dog_The_Bounty_Hunter
[01:14] <ubitux> wouf~
[02:40] <ubitux> http://pastie.org/5591394 dvds are funny :)
[03:13] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:76e65a1b731e: avs: silence "may be used uninitialized" warnings
[03:13] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:30872fa09be9: ivi_common: Fix use of uninitialized warnings
[03:13] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:2fbc8ea08ed4: hpel_motion_search: move code used for asserts under correct #if
[03:13] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:e0a99d541111: msmpeg4enc: fix possible use of uninitialized warning
[03:18] <michaelni> ubitux, there are strict aliassing violations in ffprobe: http://fate.ffmpeg.org/log.cgi?time=20121227120626&log=compile&slot=armv7l-panda-gcc4.4-armv4t
[03:19] <michaelni> hmm saste not here :(
[03:21] <wm4> michaelni: av_freep is the biggest violation of all
[03:26] <ubitux> is this due to a recent change?
[03:28] <michaelni> ubitux, probably not
[03:29] <michaelni> maybe aliasing between int vs enum but i didnt look closely
[03:29] <michaelni> its not happening on x86, just on arm
[03:31] Action: ubitux wonders where these children_ids are init&
[03:44] <michaelni> /tmp/ccMlWhhK.s:1622: Warning: Use of 'mov' may violate WAW dependency 'GR%, % in 1 - 127' (impliedf), specific resource number is 15
[03:45] <michaelni> assembler warning fron gcc output of plain C code if iam not mistaken :)
[04:34] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:c5bea1b02b16: xtea: fix unused variable warning for config_small
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:97660b8b137f: ffplay: move frame step pause from the video thread to video_refresh
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:4e33d8ebb39f: ffplay: only drop frames if not in frame step mode
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:0ca4e9f01483: ffplay: reset external clock to unknown on start
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:ef7f3b08703f: ffplay: hide cursor on no mouse movement
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:ec89ea302257: ffplay: always display audio visualization if mode is set
[04:40] <cone-801> ffmpeg.git 03Marton Balint 07master:92b50b71a1e4: ffplay: fix type of time_diff in waveform display
[04:40] <cone-801> ffmpeg.git 03Michael Niedermayer 07master:ba34e56c0d39: Merge remote-tracking branch 'cus/stable'
[11:05] <durandal_1707> nice, it tooke more than year for l to reinvent truehd channel layouts
[11:08] <durandal_1707> ubitux: you gonna push your subtitles before i push w64 muxer?
[11:11] <durandal_1707> also w64 selecting wav demuxer is brain dead
[11:34] <saste> ubitux, what's your first subtitle codec in queue? anyone I could review?
[11:37] <durandal_1707> the first one in changelog
[11:38] <durandal_1707> all others are added after that one
[11:50] <saste> ok, so it is MPL2, VPlayer, MPlayer
[12:14] <durandal_1707> saste: you still against my swapuv patch?
[12:15] <saste> durandal_1707, i don't understand your objections to my proposed changes
[12:16] <durandal_1707> your proposed changes are not good, my checks are more future proof
[12:17] <saste> durandal_1707, "your proposed changes are not good", elaborate
[12:18] <durandal_1707> you ask to remove many checks
[12:23] <saste> durandal_1707, for example why do you use the PLANAR flag? That should be avoided, as the flag itself is completely foobar
[12:29] <durandal_1707> filter works for planar formats only
[12:34] <saste> durandal_1707, yes but what about NV12? it is planar according to the braindead definition of planar of the flag
[12:36] <durandal_1707> so what?, ther are other checks
[12:39] <saste> durandal_1707, yes there are too many checks ;-)
[12:40] <durandal_1707> too many, come on
[12:40] <saste> what i don't understand is this thing about the thing the my change was not valid, can you provide an example where my imo simpler approach won't work?
[12:41] <saste> i'm relying on some assumptions on the format, but the assumptions are sane
[12:41] <durandal_1707> it would not work if someone adds new pixel format? are then going to be there and update your precious function to not use that format?
[12:45] <wm4> planar should mean one component per plane
[12:47] <saste> durandal_1707, planar/packed can be derived with much greater detail by inspeciting the pixdesc struct
[12:48] <wm4> saste: but who wants to do that every time?
[12:48] <saste> you don't need to set a flag for that, indeed i don't know why it was added in the first place, i suppose to support some weird hack in swscale
[12:48] <wm4> saste: pixdesc is too "compressed" to be actually usable directly
[12:48] <saste> wm4, write an helper for that
[12:48] <saste> otherwise you end with inconsistent code all the time
[12:49] <saste> we already did that, check the history of imgconvert.c in libavcodec
[12:49] <durandal_1707> like code that got in before or now is even better
[12:49] <saste> people adding new formats and forgetting to update the various hacks in utilities
[12:49] <saste> always ending in broken code
[12:50] <saste> also PLANAR, again NV12 is not planar and should not be regarded as such
[12:51] <wm4> it's both packed and planar
[12:52] <durandal_1707> now we will troll about some other code: is it half empty or half full...
[12:59] <iive> saste: nv12, as far as I remember have 2 planes. one Y luminance and one with packed UV samples.
[12:59] <wm4> I wonder why ffmpeg has to support these more weird formats directly as pixel formats
[12:59] <wm4> (I guess NV12 is still needed, though)
[13:00] <wm4> but consider endian swapped formats and the like
[13:03] <saste> how difficult would be to parse all*.c files and create component tables, instead of maintaining them in general.texi?
[13:04] <saste> we already rely on Perl for building the docs, so Perl usage is allowed
[13:07] <durandal_1707> wm4: so how would you handle endian swapped formats?
[13:08] <wm4> durandal_1707: like audio formats
[13:08] <durandal_1707> you want instead to create bunch of codecs? try to ask that on l channel
[13:09] <wm4> uh no
[13:09] <wm4> rawvideo could utilize an extra table for that
[13:11] <wm4> and then do byte swapping when an endian flag is set
[13:14] <durandal_1707> wm4: that is same, using lavc rawvideo, and forcing every decoder/encoder to do swapping
[13:14] <wm4> so there are codecs that actually encode from / decode to swapped endian?
[13:16] <durandal_1707> it is done so one just do memcpy instead of swapping when reading
[13:17] <wm4> I was assuming most decoders do non-trivial operations on the pixel data, which would require endian swapping anyway, regardless of whatever is stored in files
[13:18] <durandal_1707> it is mostly for image formats
[13:21] <wm4> also, most time you can do something useful with it only after endian swapping
[13:24] <cone-785> ffmpeg.git 03Diego Biurrun 07master:844c68520eec: configure: Add --disable-all command line option
[13:24] <cone-785> ffmpeg.git 03Diego Biurrun 07master:5ad2f0bfb23c: build: Add rtpenc_chain extra config option
[13:24] <cone-785> ffmpeg.git 03Diego Biurrun 07master:69583bd3b1eb: avfilter: Refactor unconditional filter registration
[13:24] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:2a9443a110e1: Merge commit '69583bd3b1eba471366141c945030c163e073e02'
[13:46] <durandal_1707> doing byteswapping in lavc/lavf should be avoided
[13:50] <durandal_1707> audio code may not been benchmarked enough to find out that single byteswapping and memcpy is faster than what is currently in pcm*.c
[14:01] <burek> is there an option in ffmpeg which does the same thing that qt-faststart does (is it -flags +global_header) ?
[14:07] <durandal_1707> burek: look at mov muxer options
[14:07] <durandal_1707> that feature is supported long ago
[14:07] <burek> i thought qt-faststart is meant for mp4 formats?
[14:08] <durandal_1707> mov=mp4
[14:08] <burek> oh, ok :) thanks :)
[14:08] <durandal_1707> -movflags faststart
[14:08] <burek> thanks again :)
[14:08] <durandal_1707>    faststart               E..... Run a second pass to put the moov at the beginning of the file
[14:40] <cone-785> ffmpeg.git 03Diego Biurrun 07master:c73c87b41298: cosmetics: Prettyprint codec/format/filter registration files
[14:40] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:83d4a8442ca4: Merge remote-tracking branch 'qatar/master'
[14:40] <burek> any ffmpeg license expert to suggest the answer on this: http://ffmpeg.gusari.org/viewtopic.php?f=16&t=775 ?
[14:42] <durandal_1707> answer is yes
[14:42] <wm4> is ffmpeg under AGPL now?
[14:43] <durandal_1707> if they create mp3 but do not pay for license they can be sued by license owner
[14:43] <durandal_1707> wm4: what is agpl?
[14:43] <wm4> http://en.wikipedia.org/wiki/Affero_General_Public_License
[14:44] <durandal_1707> wm4: no, why you think it is?
[14:44] <burek> i wonder how rich would the inventor of fat32 or extfs be, if he asked for a fee on each file created in such a fashion :)
[14:45] <durandal_1707> your are mixing apples and oranges
[14:46] <durandal_1707> burek: ask them to ask mp3 license owner can they do that
[14:46] <durandal_1707> dont use mp3 and you are fine
[14:47] <burek> i believe it falls into the same category
[14:47] <burek> sadly
[14:48] <burek> but ok, i was ironic obviously :)
[14:54] <michaelni> burek, i think the guy should ask a lawyer
[14:54] <michaelni> not mp3 "creators"
[14:55] <michaelni> they will always say "pay us for a license"
[14:56] <burek> i agree
[15:30] <saste> http://paste.org/59423
[15:30] <saste> something like this
[15:30] <saste> the problem is that we need to move some information to the all*.c files
[15:31] <saste> for example type of codec (A/V/S) and description
[16:14] <Compn> burek : depends if he is in a country where they honor software patents
[16:14] <Compn> no patents = no license :P
[17:04] <durandal_1707> does anything use tree instead dictionary?
[17:07] <saste> durandal_1707, drawtext
[17:08] <durandal_1707> hmm, why?
[17:10] <saste> durandal_1707, because "dictionary"  is not efficient
[17:10] <saste> it has O(N) against O(log_2(N))
[17:10] <saste> i mean the lookup operation
[17:58] <durandal_1707> i gonna push w64 muxer, i cant wait ubitux
[18:02] <wm4> why does ffmpeg not make use of BRANCHES
[18:02] <wm4> and I don't mean for relases
[18:02] <wm4> but for features
[18:02] <wm4> and if they're stable and well tested, merge them into master
[18:05] <durandal_1707> how would one found that branch is stable and well tested?
[18:05] <wm4> first off, you wouldn't feel the need to push experimental code into master
[18:06] <durandal_1707> experimental?
[18:06] <wm4> and then people who want to use this feature would test it
[18:06] <durandal_1707> are you kidding, bunch of already present code is experimental
[18:07] <durandal_1707> "and then people who want to use" that problem here is to find such people
[18:07] <durandal_1707> when there is just no enough people
[18:09] <durandal_1707> also code if far from experimental or hard to understand
[18:29] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:981baf742ce6: lavfi: replace REGISTER_BUILTIN_FILTER() by REGISTER_FILTER_UNCONDITIONAL()
[18:31] <michaelni> wm4, we use branches, see the repos of developers, for example ubitux, he has branches for the stuff he works on
[18:31] <michaelni> marton for ffplay too
[18:31] <michaelni> and so do others ...
[18:31] <durandal_1707> i have branches too, want to work on something, wm4?
[18:49] <cone-785> ffmpeg.git 03James Almer 07master:d59d39e16425: wavdec: fix duration calculation for files with invalid data size
[18:51] <ubitux> durandal_1707: ok :(
[18:54] <durandal_1707> ubitux: i dont get it, why are you sad?
[18:54] <ubitux> the w64 muxer
[18:55] <ubitux> it's my fault anyway :)
[18:55] <durandal_1707> no i will wait your subtitles, but i think l will beat you
[18:56] <durandal_1707> you need to fix allformats after diegonism
[18:59] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:53228f47d689: avf_concat: init pointers to NULL for saftey and to avoid warning of uninitialized use.
[18:59] <cone-785> ffmpeg.git 03Dave Yeo 07master:925c9f444fc9: Fix OS/2 threading
[19:00] <j-b> lol at last commit
[19:00] <durandal_1707> who to blame?
[19:03] <durandal_1707> hmm, that line is pretty old
[19:05] <durandal_1707> os/2 have threads?
[19:23] Action: durandal_1707 sigh
[19:52] <durandal_1707> ^ this is nothing, deadbeef dev stopped supporting FFmpeg for anything higher that 0,11
[19:53] <durandal_1707> this is ridiculous, that is what you get when you have no API/ABI at all
[19:55] <durandal_1707> btw, fyi, FFmpeg have community chan on google plus
[19:55] <durandal_1707> join so we can have more members than qatar
[20:04] <michaelni> ubitux, libavformat/assdec.c:135: warning: dereferencing type-punned pointer will break strict-aliasing rules
[20:06] <wm4> why is it called "qatar"
[20:08] <michaelni> btw, if someone has a list of API/ABI removials that affect deadbeef, iam interrested
[20:08] <durandal_1707> wm4: http://en.wikipedia.org/wiki/qatar
[20:11] <durandal_1707> michaelni: https://plus.google.com/s/ffmpeg%20deadbeef
[20:12] <ubitux> wm4: they made a qatar related joke once
[20:12] <ubitux> michaelni: erm& :/
[20:12] <ubitux> will try to fix it
[20:16] <michaelni> ubitux, libavformat/jacosubdec.c:231: warning: dereferencing type-punned pointer will break strict-aliasing rules
[20:17] <ubitux> :(
[20:36] <kierank> 18:55:20 <durandal_1707> join so we can have more members than qatar --> lol
[20:37] <durandal_1707> kierank: that is typo, should be _Q_atar
[20:55] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:4f1d7cc0ed66: bktr: fix warning: missing braces around initializer
[20:55] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:05d81a57f56a: motion_est_template: fix warning: "ASSERT_LEVEL" is not defined Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:55] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:aa9507cc3d56: swscale: fix warning: assignment from incompatible pointer type
[20:55] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:3cd137bfeac8: vf_deshake: fix strict aliassing errors
[20:55] <cone-785> ffmpeg.git 03Michael Niedermayer 07master:c77eb4ee9025: sws: fix warning: cast from pointer to integer of different size
[21:30] <ubitux> michaelni: i don't know how i am supposed to fix that
[21:31] <ubitux> pointer is uint8_t*, and parameter is char*, i don't see the problem with the cast
[21:32] <michaelni> uint8 and char* are compatible but uint8** and char** are apparently not
[21:33] <michaelni> or at least the compiler (path64 IIRC) thinks so
[21:36] <ubitux> well i guess i'll make an intermediate pointer
[21:38] <wm4> michaelni: that is right
[21:38] <wm4> void** and uint8** is not compatible either
[21:38] <wm4> that's why av_freep is undefined behavior galore
[22:13] <ubitux> what do you propose?
[22:14] <ubitux> wm4: av_freep is doing a void* to void**
[22:14] <wm4> ubitux: av_freep can be a macro at best
[22:16] <ubitux> i may be wrong, but i think another solution will start warning all over the codebase
[22:16] <ubitux> also, do you have a compiler which complain on this case?
[22:17] <ubitux> or anything which suggest an actual problem?
[22:18] <wm4> it's an aliasing violation and thus undefined behavior
[22:18] <wm4> I can't imagine an actual problem right now (though who knows what LTO does...)
[22:22] <ubitux> rhaa fuck that shitty all{formats,codecs} diego'ing
[22:46] <Compn> why is there a vp9 ?
[22:47] <wm4> because why not
[22:49] <Compn> lol
[22:49] <Compn> + * Copyright (c) 2012 Google, Inc.
[22:49] <Compn> + * Filter that shortens or lengths and audio stream
[22:49] <Compn> ...
[23:53] Action: ubitux is creating a nice conflicts mess with its own subtitles branches
[23:53] <ubitux> need to push some stuff&
[23:54] <ubitux> michaelni: any comment on http://ffmpeg.org/pipermail/ffmpeg-devel/2012-December/136508.html ?
[00:00] --- Sun Dec 30 2012