[Ffmpeg-devel-irc] ffmpeg-devel.log.20120209

[burek]

[01:08] <burek> how can i roll back my local git repos to a certain revision and recompile ffmpeg
[01:16] <Compn> burek : git --reset something 
[01:17] Action: Compn forgets syntax
[01:17] <durandal_1707> git checkout revision-number
[01:40] <burek> thanks
[01:40] <burek> after i do git checkout rev
[01:40] <burek> do i need to do anything else
[01:40] <burek> or all the files are already fixed
[01:40] <burek> rolled back
[01:41] <funman> it's ok but i advise a bit of rtfm even if git can be confusing
[01:46] <CIA-40> ffmpeg: 03Ronald S. Bultje 07master * r4ff6dea390 10ffmpeg/libavcodec/x86/pngdsp.asm: pngdsp: swap argument inversion.
[01:46] <CIA-40> ffmpeg: 03Ronald S. Bultje 07master * r9b027c0d07 10ffmpeg/libavcodec/ (dsputil.h utils.c): 
[01:46] <CIA-40> ffmpeg: dsputil: set STRIDE_ALIGN to 16 for x86 also.
[01:46] <CIA-40> ffmpeg: This fixes crashes in e.g. PNG decoding with SSE2 enabled. In fact, many
[01:46] <CIA-40> ffmpeg: x86 optimizations for codecs assume that our buffer strides are 16-byte
[01:46] <CIA-40> ffmpeg: aligned.
[01:46] <CIA-40> ffmpeg: 03Diego Biurrun 07master * r0144fe6995 10ffmpeg/ (13 files in 5 dirs): 
[01:46] <CIA-40> ffmpeg: Remove Sun medialib glue code.
[01:46] <CIA-40> ffmpeg: It is obscure, most likely unused and not bit-exact compared to
[01:46] <CIA-40> ffmpeg: libavcodec due to a different IDCT transform algorithm.
[01:46] <CIA-40> ffmpeg: 03Alex Converse 07master * r314af0ff94 10ffmpeg/ (libavcodec/aacdectab.h libavutil/audioconvert.h): (log message trimmed)
[01:46] <CIA-40> ffmpeg: aacdec: Use correct speaker order for 7.1.
[01:46] <CIA-40> ffmpeg: The spec says the following speaker mapping is default:
[01:46] <CIA-40> ffmpeg: center front speaker
[01:46] <CIA-40> ffmpeg: left, right center front speakers,
[01:46] <CIA-40> ffmpeg: left, right outside front speakers,
[01:46] <CIA-40> ffmpeg: left surround, right surround rear speakers,
[01:46] <CIA-40> ffmpeg: 03Ronald S. Bultje 07master * rce1e250ee9 10ffmpeg/libavcodec/x86/ (h264_idct.asm h264_intrapred.asm): 
[01:46] <CIA-40> ffmpeg: h264: manually save/restore XMM registers for functions using INIT_MMX.
[01:46] <CIA-40> ffmpeg: On Win64, these registers are callee-save, so not saving/restoring them
[01:47] <CIA-40> ffmpeg: Deprecate avcodec_encode_video().
[01:47] <CIA-40> (26 lines omitted)
[04:19] <Mista_D> what's a better gcc version for ffmpeg on Centos? requiring least patches for faac/x264/lame/xvid/... ?
[13:15] Action: Compn searches documentation
[13:15] <Compn> does ffmpeg have an option to disable index ? say for accessing large avi file over slow network ?
[13:16] <Compn> mplayer has -noidx
[13:16] <av500> you dont need to read the whole file to get the index :)
[13:18] <Compn> right, but the index is at the end
[13:18] <Compn> and i dont want it seeking back there
[13:19] <av500> does lavc read avi using the index?
[13:20] <Compn> i am pretty sure lavf reads the avi index.
[13:20] <Compn> altho i have not verified this 
[13:21] <Compn> since reading avi indexes is a standard thing for all players i've ever heard of
[13:21] <Compn> lavf obviously does not need an index to play, if one is not found however
[13:22] <av500> my avi player does not read the index for playing :)
[13:23] <av500> noidx might make sense for avi and asf
[13:23] <av500> for mp4 you need it
[13:23] <Compn> yes mp4 needs it
[13:23] <Compn> i cant remember mkv 
[13:23] <av500> not needed to play
[13:23] <Compn> right
[13:23] <Compn> anyways, thats in response to this bug > https://ffmpeg.org/trac/ffmpeg/ticket/986
[13:24] <av500> but that asks to disable all seeking
[13:26] <Compn> i dont think the user knows what he wants, so i guess what his problems are
[13:27] <Compn> and yes, he did hit the mp4 over pipe problem with index being at the end. his command line said ffmpeg -i out.mkv but his error said [mov,mp4,m4a,3gp,3g2,mj2 @ 01988da0] moov atom not found .
[13:27] <Compn> tricky users :)
[13:27] <Compn> make me re-read things
[13:27] Action: Compn goes afk
[14:06] <SvenL-> Yeah sure ProRes lives ! Where can I have find """"official"""" infos ? Looked at the the gmane ml archive, the git and some forum threads, is there any other place for infos ?
[14:21] <overflow_0f8b> hi
[14:21] <overflow_0f8b> gcc does not evaluate my 3x3 multiply 3x3 matrix by 3x3 matrix macro at compile time :(
[14:22] <overflow_0f8b> ( it does with 2x2 matrix rotation )
[14:24] <av500> #gcc?
[14:25] <Tjoppen> also: write a generator?
[14:26] <overflow_0f8b> av500<< chan is +R cant talk there
[14:27] <overflow_0f8b> so i'm skipping it
[15:55] <Compn> SvenL- : you mean prores internal info ?
[15:55] <Compn> or specs/ docs ?
[15:56] <SvenL-> prores with ffmpeg, just wondering if there is a place I missed for some info
[15:56] <Compn> there is the wiki
[15:57] <Compn> http://wiki.multimedia.cx/index.php?title=Apple_ProRes
[15:58] <Compn> is that what you are looking for? or something else? how to use prores decoder with ffmpeg ? its already in ffmpeg, if your ffmpeg does not decode prores, you may have to upgrade 
[15:58] <Compn> and if you are using latest git ffmpeg and still having problems with some prores file, you should probably file a bug and upload a sample so ffmpeg can fix it :)
[15:59] <SvenL-> okay, thanks a lot, that's exactly what I was looking for
[15:59] <SvenL-> I saw a link to this wiki on a thread about prores, but the address doesn't "sound" like ffmpeg, so i didn't dig this way haha
[15:59] <Compn> hehe
[16:00] <SvenL-> while I'm at it I checked for this : http://wiki.multimedia.cx/index.php?title=REDCode Interesting
[16:00] <Compn> wiki.multimedia.cx is run by a xine developer , who is also an ffmpeg developer :)
[16:00] <SvenL-> Oh okay
[16:01] <Compn> just its on his server instead of ours :)
[16:01] <SvenL-> so that's *the* place for ffmpeg doc ?
[16:02] <Compn> yes, theres some other pages too
[16:02] <Compn> http://wiki.multimedia.cx/index.php?title=Mirrored_Files
[16:02] <Compn> thats for official specs
[16:02] Action: Compn digs up other url
[16:02] <SvenL-> I may be able provide a small sample of .R3D footage of recent RED cameras
[16:02] <Compn> that would be nice , i think RED codec is on our list to reverse engineer 
[16:02] <SvenL-> uhm
[16:04] <Compn> michaelni keeps some of his documents here : http://www1.mplayerhq.hu/~michael/
[16:04] <SvenL-> so as is it can't playback ?
[16:05] <Compn> i havent tested red codec, let me see.
[16:06] <SvenL-> Ha the doc says it's a "special" JPEG2000 file
[16:07] <Compn> it gives error with my ffmpeg
[16:07] <SvenL-> Okay
[16:07] <Compn> but i dont have latest version
[16:07] <JEEBsv> just like DV is a "special" version of MJPEG?
[16:07] <Compn> hmm?
[16:08] <Compn> i think it means like rv40/wmv3 are special versions of mpeg4 asp :P
[16:08] <SvenL-> yes
[16:08] <SvenL-> I think too
[16:08] <Compn> modified, proprietary
[16:08] <JEEBsv> yeh
[16:08] Action: Compn looks for binary codec to add to mplayer
[16:08] <SvenL-> hence the need for reverse engineering ?
[16:09] <Compn> yes, we need specs or reverse engineer it :)
[16:09] <SvenL-> I get it, thx
[16:09] <Compn> heres our r3d samples > http://samples.mplayerhq.hu/r3d/
[16:09] <JEEBsv> then there's f.ex. Earthsoft's "super special" DV format, which is their own representation of "DV" :3
[16:09] <SvenL-> question : they are generally huge files that can't be played back at full resolution
[16:09] <Compn> one here too > http://samples.mplayerhq.hu/fate-suite/r3d/
[16:10] <Compn> JEEBsv : btw if you have samples of non-standard DV codecs... well i like to have them so i can add them to mplayer :)
[16:10] <SvenL-> what in the decoding process handles rendering at lower resolution ?
[16:10] <Compn> there is 'lowres' option, to decode at lower resolution
[16:11] <Compn> i am not sure if it works with jpeg2000 codec 
[16:11] <JEEBsv> I used to have a sample of Earthsoft DV uploaded, I could re-upload it if it's no longer there
[16:11] <SvenL-> RED, the company who develops it, provides a software that basically has a player where you can choose on the fly 1/2 res, 1/4 res etc...
[16:11] <iive> wasn't lowres removed recently?
[16:11] <Compn> iive : i seriously doubt it?
[16:11] <JEEBsv> there's already a patch for it for ffmpeg/libav IIRC tho. Haven't gotten to reading through it yet tho
[16:12] <iive> i think i've read that libav people were working on it.
[16:12] <Compn> i think libav's plan is to seperate all codecs into seperate projects
[16:12] <JEEBsv> I was told to take a look at the patch, pretty much IIRC
[16:12] <JEEBsv> as I brought it up
[16:12] <JEEBsv> I originally started checking out the specs for it, and wrote a parser for the index files
[16:13] <Compn> JEEBsv : for red or dv? 
[16:13] <JEEBsv> earthsoft dv
[16:13] <SvenL-> (background about me : I'm trying to build up a toolset for on-set footage playback, analysis, dailies exports and pre-colorgrade so I'm currently looking at what could be handled)
[16:13] <JEEBsv> http://x264.fushizen.eu/encodan/earthsoft_dv/ <- also the sample is still up
[16:13] <Compn> SvenL- : cool! i think its always fun when ffmpeg can do realtime movie work :)
[16:14] <SvenL-> yes !
[16:14] <Compn> JEEBsv : oh, i was hoping it was dv-in-avi :P
[16:14] <SvenL-> that's why i'm asking about resolution, what's my best pick for lower-res realtime playback ?
[16:14] <SvenL-> hahaha
[16:14] <JEEBsv> Compn: oh yes, they have their own format too :D
[16:15] <JEEBsv> you don't really need the index tho
[16:15] <JEEBsv> as the headers contain the same info
[16:15] <Compn> SvenL- : you mean which codec is best for low-res playback ?
[16:15] <Compn> good question, i dont know the answer...
[16:15] <JEEBsv> http://smdn.jp/softwares/libav/pv4/ <- the patch that someone then googled up after I had already used some time to parse the 'specs' Earthsoft had on their site
[16:16] <JEEBsv> (has patches for ffmpeg/libav)
[16:16] <SvenL-> this may be out of my knowledge already but as I understand it, for redcode at least, there is one """codec""", a decoder only in fact
[16:16] <SvenL-> Compn, I'm rather asking where I should be looking at for choosing resolution
[16:16] <overflow_0f8b> Tjoppen<< neat idea :), but i have made it work lol, i just don't use a tmp variable and let gcc evaluate it
[16:17] <overflow_0f8b> now     _printM33("z1 x z2:", MM33( z1, z2 ) ); prints constants evaluated at compile time ^^
[16:17] <SvenL-> I mean, at what level, ffmpeg options ? player application ? even lower level ?
[16:18] <SvenL-> ( http://ffmpeg-users.933282.n4.nabble.com/Cannot-decode-REDCODE-r3d-files-td3209331.html )
[16:19] <Compn> SvenL- : that ffmpeg-users post talks about old r3d files, not the newer ones. :(
[16:19] <SvenL-> yes
[16:19] <SvenL-> Hum...
[16:20] <SvenL-> and I wonder how old/new are the samples you mentionned
[16:22] <SvenL-> do you know about Cineform ?
[16:22] <Compn> i know ffmpeg doesnt have a decoder for that either :)
[16:23] <SvenL-> hehe, I know
[16:23] <SvenL-> right now they only provide a SDK
[16:27] <Compn> SvenL- : you might want to look at this project > http://code.google.com/p/ffmbc/
[16:27] <SvenL-> thank you, didn't know about it
[16:27] <SvenL-> *reading*
[16:31] <SvenL-> really interesting project, actually everything I need is here, awesome basically :)
[16:33] <SvenL-> Compn, may I bother you just a bit more with two other quetions ? :>
[16:36] <Compn> sure
[16:36] <Compn> the ffmbc project is on ffmpegs' website , on the projects page. in case you want to look for other projects . of course not all projects are listed there, we need help finding them :)
[16:37] <SvenL-> Okay, ffmbc is really interesting, I'll look into it. do you think it's possible to add filters ? vary basic curve filter
[16:37] <SvenL-> very*
[16:38] <SvenL-> (the footage from the cam is flat, just want to apply a nice S curve for convenience at editing)
[16:38] <Compn> video filters? yes, ffmpeg and ffmbc use the same filter api and you can add a filter to them
[16:38] <SvenL-> okay, great
[16:38] <Compn> i dunno if that filter exists yet 
[16:39] <SvenL-> hum, I'm pretty sure it does
[16:39] <Compn> i havent been paying attention :)
[16:40] <SvenL-> curves is one of the most basic filter along with brightness/contrast/saturation. Oh well anyway I will just read, never used ffmpeg with advanced goals
[16:41] <Compn> ah
[16:41] <Compn> i have no idea what 'flat' footage means, nor what an S curve would do to it :)
[16:42] <Compn> ffmpeg can also use mplayer's filters , not sure about ffmbc tho
[16:42] <SvenL-> sorry, flat means that it has not much contrast at all, dark areas of the video are in fact dark grey
[16:42] <Compn> oh :)
[16:43] <SvenL-> that's common for current cameras, you generally tweak and get _the_ picture in post-production
[16:45] <SvenL-> and an S-curve is an easy way to correct this, it bumps the contrast up so the image is "nice". It's just a quick fix so the film editor gets a nice looking video
[16:46] <Compn> right, gotcha
[16:49] <SvenL-> Compn, do you know what a video histogram is ? I apologize for such off-topic and basic question, /query me if you prefer
[16:49] <Compn> i do, i think i've seen one in a video editor once
[16:50] <Compn> dont think ffmpeg has one of those :)
[16:51] <SvenL-> I've had a hard time finding out, it's what you would find in a video editor exactly
[16:51] <SvenL-> it's both an analysis and effect-filter
[16:53] <Mista_D> Any advice on macroblock detection?
[16:53] <SvenL-> I don't know where I should ask for this, maybe in #ffmpeg first
[16:57] <Compn> yes, ask user questions there
[16:57] <SvenL-> yes, sorry :)
[16:58] <Compn> Mista_D : sorry, no idea. try ffmpeg.org 's libav-user list or ffmpeg-user ?
[18:04] <CIA-67> ffmpeg: 03Laurent 07master * r71d6b46849 10ffmpeg/configure: 
[18:04] <CIA-67> ffmpeg: configure: fix gcc asm tests for MinGW GCC 4.6.2
[18:04] <CIA-67> ffmpeg: Thanks to stackoverflow
[18:04] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:40] <CIA-67> ffmpeg: 03Stefano Sabatini 07master * rc28e7af70d 10ffmpeg/ (4 files in 2 dirs): lavu: add timestamp.h header with convenience timestamp utilities
[19:06] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * r677911ad81 10ffmpeg/ffmpeg.c: 
[19:06] <CIA-67> ffmpeg: ffmpeg: better dox for *pts.
[19:06] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:06] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * r491d8353e8 10ffmpeg/libavcodec/mpeg12.c: 
[19:06] <CIA-67> ffmpeg: mpeg12dec: print the value that is being ignored in load_matrix()
[19:06] <CIA-67> ffmpeg: error message by Reimar
[19:06] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:06] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * rdd7198b25f 10ffmpeg/libavformat/utils.c: 
[19:06] <CIA-67> ffmpeg: lavf: put av_new_stream() compatibility wrapper back
[19:06] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:06] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * rf480fcb1f9 10ffmpeg/libavcodec/bink.c: 
[19:06] <CIA-67> ffmpeg: bink: fix pointer type warnings.
[19:06] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:06] <CIA-67> ffmpeg: 03Michael Niedermayer 07master * r8a07429318 10ffmpeg/libavformat/utils.c: 
[19:06] <CIA-67> ffmpeg: lavf: put av_find_stream_info() compatibility wrapper back.
[19:06] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:22] <Snaggle> I'm trying to build ffmpeg-git w/ --disable-swresample and linking fffplay errors out with undefined symbols for _swr_free and other _swr symbols (and ld warns about "-Llibswresample" not being found).  Is this a bug in the library disabling code, or a mistake on my part?  Disabling swscale fails similarly, but postproc doesn't.
[22:23] <Snaggle> If this is a bug, I'll go ahead and file a trac item, but wanted to make sure I wasn't doing something incorrectly before going there
[22:25] <durandal_1707> Snaggle: how old is that ffmpeg, give version
[22:26] <Snaggle> fresh git pull.  not sure how to get git to spit out a 'version'
[22:27] <durandal_1707> ffmpeg -version
[22:27] <Snaggle> durandal_1707: program linking is failing, so I have no executable to ask for a version
[22:28] <durandal_1707> Snaggle: what git -describe says?
[22:28] <Snaggle> n0.8-6741-gdd7198b
[22:29] <durandal_1707> ffplay depends on swr so it should not be build
[22:30] <durandal_1707> fix is trivial
[22:37] <Snaggle> ah, didn't know swr was a hard dep on ffplay.  I see now that ffplay_deps in configure is missing swr
[23:05] <Snaggle> durandal_1707: thanks for the hint about where to look
[23:28] <CIA-67> ffmpeg: 03Carl Eugen Hoyos 07master * r5c2c6bbfba 10ffmpeg/ (libavcodec/raw.c libavformat/nut.c): Allow encoding rawvideo RGB0 and friends.
[23:28] <CIA-67> ffmpeg: 03Carl Eugen Hoyos 07master * r52719dae8a 10ffmpeg/libavcodec/ (ljpegenc.c mjpegenc.c mpegvideo_enc.c): Support encoding BGR24 and BGR0 in ljpeg.
[23:28] <CIA-67> ffmpeg: 03Carl Eugen Hoyos 07master * redf34c346e 10ffmpeg/ (libavcodec/raw.c libavformat/nut.c): Allow encoding rawvideo RGBA64 and friends.
[23:36] <Mista_D> where is histogram data stored while using -qphist? Tried with `...-an -vstats_file vstats.txt -stats -vstats -benchmark -qphist -loglevel verbose -psnr -y test.h264  2>2.txt 1>time.txt`
[23:57] <CIA-67> ffmpeg: 03Carl Eugen Hoyos 07master * r3b25d6193c 10ffmpeg/tests/ref/lavfi/ (5 files): Fix fate tests after 5c2c6bbf.
[23:58] <CIA-67> ffmpeg: 03Carl Eugen Hoyos 07master * rea4037162f 10ffmpeg/libavcodec/libstagefright.cpp: 
[23:58] <CIA-67> ffmpeg: Fix libstagefright compilation
[23:58] <CIA-67> ffmpeg: Comment-by-michael: iam commiting this as the code cannot work without it and likely works with it.
[23:58] <CIA-67> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Fri Feb 10 2012