[Ffmpeg-devel-irc] ffmpeg-devel.log.20111126

[burek]

[00:07] <michaelni> ubitux, if it allows code reuse and simplifies things its surely possible
[00:07] <ubitux> i'm going to try
[00:07] <ubitux> but it might take some while, i'm not familiar with that stuff... yet.
[00:08] <michaelni> where to put it i dont know, where it fits in naturally best i guess
[00:08] <ubitux> :)
[00:09] <ubitux> it seems the drawutils are a good place
[00:58] <CIA-36> ffmpeg: 03Stefano Sabatini 07master * ra11eeb9215 10ffmpeg/configure: configure: sort entries
[01:17] <CIA-36> ffmpeg: 03Mans Rullgard 07master * r6b34fbba9b 10ffmpeg/libavutil/common.h: 
[01:17] <CIA-36> ffmpeg: MK(BE)TAG: avoid undefined shifts
[01:17] <CIA-36> ffmpeg: Casting the left-most byte to unsigned avoids an undefined
[01:17] <CIA-36> ffmpeg: result of the shift by 24 if bit 7 is set. This affects
[01:17] <CIA-36> ffmpeg: the rm demuxer.
[01:17] <CIA-36> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[01:18] <CIA-36> ffmpeg: 03Janne Grunau 07master * rd14d4d982c 10ffmpeg/tests/fate/aac.mak: 
[01:18] <CIA-36> ffmpeg: aacdec: add more fate tests covering SBR and PS
[01:18] <CIA-36> ffmpeg: Add all seven test bitstreams of Coding Technologies "aacPlus Decoder
[01:18] <CIA-36> ffmpeg: Check Package". The streams cover different ways to signal SBR and PS
[01:18] <CIA-36> ffmpeg: in different formats.
[01:18] <CIA-36> ffmpeg: 03Janne Grunau 07master * r117e2a30f2 10ffmpeg/libavcodec/ (pthread.c thread.h): 
[01:18] <CIA-36> ffmpeg: frame-mt: return consumed packet size in ff_thread_decode_frame
[01:18] <CIA-36> ffmpeg: This is required to fulfill avcodec_decode_video2() promise to return
[01:18] <CIA-36> ffmpeg: the number of consumed bytes on success.
[01:18] <CIA-36> ffmpeg: 03Rafaël Carré 07master * rbe1e872582 10ffmpeg/configure: 
[01:18] <CIA-36> ffmpeg: configure: Store vda lib flags in extralibs instead of ldflags
[01:18] <CIA-36> ffmpeg: This way the needed linking flags end up in libavcodec.pc.
[01:18] <CIA-36> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[01:18] <CIA-36> ffmpeg: 03Mans Rullgard 07master * r00a856e3f9 10ffmpeg/libavcodec/ (arm/dca.h dca.c): 
[01:18] <CIA-36> ffmpeg: dca: ARMv6 optimised decode_blockcode()
[01:18] <CIA-36> ffmpeg: This is a hand-tuned version of the code with impossible parts of
[01:18] <CIA-36> ffmpeg: the FASTDIV function ommitted.
[01:18] <CIA-36> ffmpeg: 2-5% faster overall on Cortex-A8.
[01:18] <CIA-36> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[01:18] <CIA-36> ffmpeg: 03Mans Rullgard 07master * rcc276c85d1 10ffmpeg/ (21 files in 4 dirs): 
[01:18] <CIA-36> ffmpeg: Make channel layout masks unsigned
[02:18] <ubitux> mmh i realize we don't have any audio filter to mix multiple streams?
[02:18] <ubitux> for instance to add a background music to a video
[02:18] <ubitux> (which already has some sound)
[02:33] <michaelni> ubitux, shouldnt be hard as long as you dont try to implement AA sync
[02:34] <ubitux> AA sync?
[02:35] <michaelni> audio audio sync ;) 
[02:35] <michaelni> like audio video
[02:35] <ubitux> ah, ok :)
[02:36] <ubitux> i was still on the overlay factoring, seems like i'm tired :p
[02:51] <ubitux> michaelni: it might need some audio standardization (number of channels for instance)
[02:51] <ubitux> so certainly some usage of swr
[02:53] <ubitux> michaelni: btw, about the call for maintainers, maybe you should just add a news to the website
[02:53] <ubitux> that might be relayed by some large news website etc
[02:54] <ubitux> we lack manpower, so that would hardly make things worse
[02:58] <michaelni> ubitux, _very_ good idea
[03:36] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r31a1342e7a 10ffmpeg/libswscale/swscale.c: 
[03:36] <CIA-36> ffmpeg: swscale: remove duplicate code from yesterdays merge.
[03:36] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[10:46] <CIA-36> ffmpeg: 03Stefano Sabatini 07master * r28338bc2a3 10ffmpeg/ (7 files in 3 dirs): lavfi: add libass based subtitles renderer
[11:00] <ubitux> \o/
[11:00] <ubitux> thx saste :)
[11:01] <ubitux> lifesrc now ;)
[13:26] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r8f37c8f0f8 10ffmpeg/cmdutils.c: 
[13:26] <CIA-36> ffmpeg: opt_pix_fmts: try to fix segfault on open solaris
[13:26] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:52] <CIA-36> ffmpeg: 03kaptnole 07master * rcb8db6423a 10ffmpeg/libavcodec/aacdec.c: 
[14:52] <CIA-36> ffmpeg: aacdec: Fix Sound fragments after seeking
[14:52] <CIA-36> ffmpeg: Fixes Ticket420
[14:52] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:50] <michaelni> ubitux, Compn bcoudurier cbsrobot do you have ideas for gci tasks ?
[15:51] <michaelni> that question is of course not limited to the people above :)
[15:53] <michaelni> Hi Anssi_ do you have some idea for a google code in task ?
[15:53] <burek> I would suggest setting up wiki and forming a better documentation
[15:53] <burek> with examples and stuff
[15:54] <burek> students can help with testing mostly, since they are weak as coders (mostly)
[15:54] <michaelni> wiki sounds good
[15:57] <michaelni> our trac bugtracker comes with a buildin wiki that can easily be used
[16:03] <burek> good :) we can also setup some smaller tasks, to get them involved, like to refactor some code or something
[16:04] <burek> i mean we can't actually expect them to fix a lot of real bugs, because i fear they have got not enough knowledge for that, i dont know
[16:05] <michaelni> yes, bugfixing is unrealistic
[16:05] <michaelni> we have fixed all easy bugs already :)
[16:06] <burek> I would go for some generic tasks (plus documenting the work while they do it), like: setting up live stream from webcamera to the outter world
[16:06] <michaelni> theres the option of fixing warnings but its possibly not worth it, i can probably do it quicker without a student than with one
[16:06] <burek> and some similar interesting tasks like that
[16:06] <michaelni> this sounds quite interresting
[16:07] <burek> yes, I believe the help with coding would more take away developers time, than it would help
[16:08] <burek> also, try to see where do you loose a lot of time, while you don't write the code, and see if that everyday stuff can be transfered to someone as a task
[16:09] <burek> does ffmpeg have a facebook page? :)
[16:09] <michaelni> not that i know
[16:10] <burek> well kids are really into that stuff and they could help with it more than we know :)
[16:10] <burek> it might promote ffmpeg more, so there would be more coders joining the project?
[16:12] <ubitux> michaelni: i don't have easy task in mind, but a lot of possible audio filters :p
[16:13] <michaelni> they could go over the various forums like doom9 and see if there are bugs mentiond that are reproduceable with ffmpeg and then put them up on trac
[16:13] <ubitux> michaelni: maybe the matrix configuration in libswr?
[16:13] <ubitux> the update of the hall of shame?
[16:14] <michaelni> you mean finetuning the matrix so itz sounds best ?
[16:14] <ubitux> no, allowing the api to configure the matrix
[16:14] <michaelni> the hall of shame is dead ...
[16:14] <ubitux> hall of shame is dead because no one wants to maitain it, no? :p
[16:14] <michaelni> ubitux, these are 13-17 year olds
[16:14] <ubitux> yeah it's not easy :(
[16:14] <burek> they will be motivated with money prizes and will be monitoring several projects to find the easiest tasks, so we might adapt tasks for massiveness, not for brainiacs? :)
[16:15] <ubitux> michaelni: what about adding fate tests? :p
[16:15] <ubitux> "find some untested codec/format, and add reg tests"
[16:16] <ubitux> might be hard too...
[16:16] <michaelni> possible though this will need help from a mentor as the fate system is not trivial to work with
[16:16] <burek> they could also create some logos, icons, brandings, etc.
[16:17] <ubitux> add a RSS stream generation from the news on the website?
[16:17] <michaelni> burek, right i forgot that, we could use modified logos for each day of the year like google is using
[16:17] <ubitux> that might require some thinking to change the way news are stored
[16:18] <ubitux> looking for typo all over the code
[16:18] <ubitux> and doc, and website
[16:18] <burek> was the ffmpeg's webpage built using cms or in notepad? :)
[16:18] <michaelni> run a spellchecker over the docs (and report the segfaults ;))
[16:19] <ubitux> :D
[16:19] <burek> :)
[16:19] <ubitux> "fix some warnings:"
[16:20] <michaelni> burek, website uses a simple Makefile + very simple pieces of html
[16:20] <burek> well, adding rss support can be added as a task then, because it would be trivial thing :)
[16:20] <ubitux> no
[16:21] <ubitux> you need to change the way news are stored
[16:21] <burek> does it use db and why no? :)
[16:21] <michaelni> no db
[16:21] <burek> i see, file based only
[16:21] <michaelni> just files + header + footer IIRC
[16:22] <burek> even no php? :)
[16:22] <ubitux> news = [{'date': ..., 'content': '<p> ... </p>'}, {...}, ...]
[16:22] <ubitux> then generate index + rss based on this is a possibility
[16:22] <burek> i see
[16:22] <ubitux> it will require python support (for instance, could be perl or sth) on server side
[16:23] <ubitux> but no cgi, just for generating static content
[16:23] <ubitux> not sure how realistic it is
[16:24] <burek> well it's better to remain that way (static) because a lot of people visit it
[16:24] <burek> so the server is not that much loaded
[16:24] <burek> overloaded*
[16:24] <ubitux> mmmh maybe update the old FFmpeg tutorial?
[16:25] <burek> i would go with docs and easy stuff, to get a lot of them involved, waiting for easy points and a little bit later adding more serious tasks
[16:25] <ubitux> http://www.inb.uni-luebeck.de/~boehme/using_libavcodec.html
[16:25] <burek> do tasks have difficulty (num of points)?
[16:25] <ubitux> this could be updated to the current API
[16:25] <michaelni> theres a difficulty but i dont know if it affects points
[16:26] <ubitux> and merged into the main documentation at some point?
[16:26] <burek> ubitux, this would really be a valuable part of ffmpeg's webiste
[16:26] <ubitux> mmh, i was thinking of this one actually: http://dranger.com/ffmpeg/
[16:26] <burek> a lot of people are coming to the #ffmpeg asking for such support
[16:26] <ubitux> yes
[16:27] <ubitux> and those tutorials are old, and not maintained
[16:27] <ubitux> syncing them with upstream and adding them to the doc could be nice
[16:27] <michaelni> i fear a ffmpeg developer would be more qualified to update these tutorials than a high scool student
[16:28] <burek> it would be perfect if you would use an auto generator for docs, just by using comments in the source code, like java is practicing, but i guess that's too big of a change
[16:28] <ubitux> michaelni: building the examples and fixing the deprecated warning could a good start, and that looks accessible to me
[16:28] <michaelni> burek, you mean ? http://ffmpeg.org/doxygen/trunk/index.html
[16:28] <burek> this all could go into wiki and is really needed
[16:29] <ubitux> mmh about this: maybe integrate the doxy to the website better?
[16:29] <ubitux> (merge css, ...)
[16:29] <michaelni> fixing some deprecated warnings is easy others is not
[16:29] <ubitux> (add main website header to it, etc)
[16:29] <michaelni> i mean wrapers that have to contain the deprecated stuff ;)
[16:29] <burek> michaelni yes :) that's what I'm talking about, just each function would have comment above itself, to explain what it does and why/how
[16:30] <burek> and when you change it, you change the comments accordingly
[16:30] <burek> and you have your docs updated
[16:30] <michaelni> the doxy linked above should change when the codes comments changes
[16:30] <burek> yes
[16:30] <burek> that's the idea :)
[16:31] <burek> also merging of such things can easily be done just using dns subdomains for ffmpeg.org
[16:31] <burek> to avoid constantly updating main website
[16:31] <burek> they could help with website a lot
[16:32] <ubitux> btw, what's the state of the forum? :p
[16:32] <ubitux> no work available on it? :)
[16:32] <burek> well, I don't receive any more emails about it, so I don't know really
[16:34] <burek> I've sent the backup of the forum at my server to Kyle, and haven't heard from him since then
[16:34] <michaelni> last mail i got was 16 Nov 2011
[16:34] <burek> me too
[16:37] <burek> michaelni, this is a good example of doxygen: http://xerces.apache.org/xerces-c/apiDocs-3/classDefaultHandler.html
[16:37] <burek> for each function there is a brief description of usage and parameters
[16:42] <michaelni> http://ffmpeg.org/doxygen/trunk/group__lavu__tree.html
[16:42] <michaelni> i think the configuration may need work (its in main git)
[16:43] <burek> this is good also
[16:43] <burek> is it accessible directly from the main webpage?
[16:43] <michaelni> but it should be shown here too:
[16:43] <michaelni> http://ffmpeg.org/doxygen/trunk/tree_8c.html
[16:43] <michaelni> theres a link on http://ffmpeg.org/documentation.html
[16:43] <burek> if you are already using it, its great then :) a lot less work then :)
[16:44] <burek> excellent :)
[16:48] <burek> one more thing, is there a tutorial or something on how to setup the evnironment for a new developer, that wants to contribute to the project by writing patches and stuff
[16:49] <michaelni> not that i know also its very different between platforms
[16:50] <michaelni> it would be a good idea for a task
[16:50] <burek> do you guys use eclipse or something else?
[16:53] <kierank> no
[17:01] <burek> i think the major problem why people don't join too often as developers is that they find it very difficult to just setup the evnironment to make even a slightest code change
[17:01] <burek> so, tutorial on that would be highly appreciated
[17:02] <michaelni> yes i agree
[17:03] <burek> also one of the tasks could be that users show what all can they do using ffmpeg, and the most exotic examples could be awarded as a task done
[17:04] <burek> just to later showcase the ffmpeg's features and usability
[17:04] <michaelni> the idea is good but i dont know if it can be fit into the gci framework
[17:05] <michaelni> i think (but am not sure) that each task is first claimed by a student and then work submited and the mentor must judge if it passes needs amendment or fails
[17:05] <michaelni> so multiple and take best wont work ...
[17:06] <burek> i see
[17:06] <michaelni> i also think (but am not sure) that each task can only be finished once
[17:07] <nevcairiel> you can have the same task more then once, though
[17:07] <michaelni> but you have to create the taskl multiple times or am i misisng something ?
[17:07] <nevcairiel> yes, you have to create it multiple times, but i think there was some bulk creation thing
[17:08] <michaelni> yes, j-b posted a link but it didnt work for me
[17:08] <michaelni> i also googled and found the same link but that still didnt work :(
[17:16] <michaelni> btw, nevcairiel, do you want to help mentoring gci ?
[18:33] <ubitux> about adding "smovie" in the src movie filter, it should be possible add a "AVSubtitle" source, but i think it would make sense to actually raise frames which could be blended already
[18:34] <ubitux> the issue is, smovie must be aware of the final WxH
[18:34] <Compn> michaelni : some documentation on creating 3d (side by side) videos using ffmpeg would be nice. someone gave me a nice command line but i havent had time to test it and create docs
[18:35] <ubitux> would it make sense to raise AVSubtitle, and make filters like overlay the ability to render and blend it?
[18:35] <Compn> let me dig up the mail
[18:35] <ubitux> i'm not sure where to put the rendering
[18:35] <Compn> michaelni : http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2011-November/116890.html
[18:36] <Compn> america is tricking lots of people into buying 3d televisions, now they have to figure out which formats those tvs support, and how to make those formats :D
[18:36] <ubitux> smovie filter raises AVSubtitle, then you would have a "burn" subtitle filter, a "text" subtitle filter (print on stdout? freetype rendering?), etc.?
[18:37] <ubitux> mmh maybe i'll do that.
[18:37] <ubitux> </monologue>
[18:43] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r61c2cec957 10ffmpeg/libavcodec/pthread.c: 
[18:43] <CIA-36> ffmpeg: pthreads: fix segfault due to the thread beimg killed before it has been allocated
[18:43] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:43] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r1b84e06244 10ffmpeg/tools/lavfi-showfiltfmts.c: 
[18:43] <CIA-36> ffmpeg: lavfi-showfiltfmts.c: fix handling of null names
[18:43] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:57] <Compn> http://www.guardian.co.uk/commentisfree/cifamerica/2011/nov/11/us-justice-department-legally-hacked-twitter
[19:30] <CIA-36> ffmpeg: 03Clément BSsch 07master * rc1ae524e2b 10ffmpeg/libavfilter/avfilter.h: avfilter: fix AVFilterPad video only comment.
[19:32] <kierank> Compn: we discussed the 3d thing at vdd. there are a lot of issues to deal with
[19:34] <Compn> vdd ? whuh?
[19:34] <Compn> oh videolan days ?
[19:35] <Compn> kierank ?
[19:36] <ubitux> i think so
[19:37] <kierank> yes
[19:39] <Compn> what 3d things did you discuss and is vdd videolan dev days ?
[19:40] <Compn> if you feel like sharing
[19:40] <kierank> yes vdd = videolan dev days
[19:40] <kierank> there are a bajillion different ways of implementing 3d
[19:44] <Compn> ah
[19:44] <Compn> yes
[19:45] <Compn> luckily mplayer has stereo3d filter :P
[19:50] <michaelni> and ffmpeg has libmpcodecs wraper that supports stereo3d :)
[19:57] <kierank> well demuxing 3d blu-rays is not trivial
[19:57] <kierank> since you have two streams which need to be fed to the same decoder
[19:58] <kierank> it is not at all clear how you do that
[19:58] <kierank> how one could do that i mean
[19:59] <Compn> oh yeah, interlaced bd 3d support
[20:01] <kierank> mvc support isn't that hard
[20:01] <kierank> the demuxing is hard
[20:01] <Compn> i wonder if its more or less harder for sending to a hardware decoder
[20:02] <michaelni> the additional data could be feeded in by using side data of AVPackets maybe
[20:20] <iive> kierank: so the 3d is done by separate streams?
[20:26] <Compn> iirc, yes
[20:26] <Compn> within the same file
[20:27] <kierank> yes within the mpegts
[20:27] <nevcairiel> mvc is two or more streams, one "base view" which is just default AVC, and one or more MVC streams for the other views (for 3D, just one)
[20:28] <nevcairiel> for decoding, the decoder needs to get both streams
[20:41] <michaelni> are there some interleaving requirements so that one can expect the ts demuxer to spit out packets from related streams in sequence ?
[20:55] <iive> so you need special decoder too? it doesn't work with 2 separate instances of standard decoder?
[20:59] <nevcairiel> thats right
[21:00] <kierank> there are dts requirements that should be compliant
[21:00] <kierank> but lavf's demuxer doesn't follow dts
[21:01] <Compn> do we have some 3d bd samples for testing? :)
[21:12] <ubitux> michaelni: i think the current about page is likely to belong to the documentation one
[21:12] <ubitux> or maybe home page presenting the project, but that index is already quite filled up
[21:13] <michaelni> feel free to move it if you find a fitting spot, we also could keep it below the new one
[21:14] <ubitux> yes maybe below
[21:15] <ubitux> i'd split the list btw
[21:15] <Compn> michaelni : did you make a new task for google code in about 3d documentation? :)
[21:15] <ubitux> "FFmpeg provides various tools: [tools list] and developers libraries: [lib list]
[21:15] <ubitux> "
[21:16] <michaelni> Compn, join as mentor and add a few tasks :)
[21:16] <ubitux> it should fit well in the about page with your introduction
[21:17] <michaelni> yes, ill see what i can come up with and post that then
[21:17] Action: Compn busy planning for his trip to florida
[21:26] <ubitux> michaelni: "Whereever" while you are at it :p
[21:27] <ubitux> oh and "dependancies" ’ "dependencies"
[21:28] <Compn> forget that, make a post-commit hook to run spell check on it :P
[21:28] Action: Compn runs
[21:29] <michaelni> Compn, we dont have enough diskspace for the output
[21:44] <ubitux> hey btw, what does "FFmpeg" means? Fast Forward MPEG? wouldn't that belong to the About page too?
[21:47] <michaelni> sounds like that should be in a seperate patch
[21:49] <Compn> hehe
[21:50] <Compn> are we sure we can have 'mpeg' in the name ?
[21:50] <Compn> motion pictures experts group might not like it :P
[21:51] <Compn> what i mean is 
[21:51] <Compn> fast forward MPEG is still an acronym
[21:51] <Compn> "fast forward motion picture experts group" ??
[21:52] <ubitux> any video? any media*
[21:52] <Compn> not that i want to bikeshed this
[21:53] <Compn> sure media works too :P
[21:54] <ubitux> it's more correct :p
[21:54] Action: Compn still waits for vivo support
[21:55] Action: Compn will have to learn to reverse engineer or something
[22:01] <Daemon404> i thought vivo was done
[22:01] <Daemon404> or close to it
[22:01] <Daemon404> according to wiki
[22:06] <Compn> i think drv had some patches, forgot the current status
[22:07] <Compn> every few months i bug everyone about it
[22:08] <Compn> i am the most unhelpful person here :)
[22:23] <drv> i have a patch that's almost ready, but it still needs some cleanup
[22:23] <drv> i will get to it one of these decades
[22:58] Action: ubitux wonders if we will have -af some day :(
[23:00] Action: michaelni wants filters that can have mixed media in and out for audio vissualization and all that
[23:02] <ubitux> i'm working on it right now :)
[23:02] <ubitux> but it will be a hack at first :(
[23:15] <michaelni> the only things that are perfect at first are the ones that are never finished :)
[23:20] <michaelni> and perfect implies finished thus to fullfill the statement there can be nothing perfect.
[23:24] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r4fd5e7639b 10ffmpeg/libavcodec/vc1.c: (log message trimmed)
[23:24] <CIA-36> ffmpeg: vc1: Fix pic_header_flag=0 (SA10143.vc1)
[23:24] <CIA-36> ffmpeg: Bug introduced in:
[23:24] <CIA-36> ffmpeg: commit 4509be3d2f46a52ada8e2ecb476faed93e19abf3
[23:24] <CIA-36> ffmpeg: Author: Michael Niedermayer <michaelni at gmx.at>
[23:24] <CIA-36> ffmpeg: Date: Tue Oct 11 11:56:42 2011 +0200
[23:24] <CIA-36> ffmpeg:  vc1: reset interlaced variables, prevent another bunch of crashes.
[23:24] <CIA-36> ffmpeg: 03Michael Niedermayer 07master * r22cb8e7b34 10ffmpeg/tests/fate/aac.mak: 
[23:24] <CIA-36> ffmpeg: fate: enable new sbr tests after our rsync server has them now.
[23:24] <CIA-36> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:50] <ubitux> michaelni: :)
[00:00] --- Sun Nov 27 2011