[Ffmpeg-devel-irc] ffmpeg-devel.log.20130506

[burek]

[00:17] <kierank> 22:24:25 <"Daemon404> i still find it amusing that rff flags have to be hadled by an external filter. --?
[00:17] <kierank> why not
[00:23] <Daemon404> because applyign rff flags is the correct thing to do before returning video to users
[00:23] <Daemon404> part of 'decoding' imho
[00:23] <cehoyos> durandal_1707: I live in PAL-land and don't know much about sof-ttelecine, I believe there are several samples that use the feature on samples.ffmpeg.org
[00:23] <Daemon404> without making them jump through hoops
[00:23] <Daemon404> (and lavfi is a BIG hoop_)
[00:25] <durandal_1707> so what is done in filter should be instead done in decoder?
[01:27] <megaTherion> Is there an function which creates an flat duplicate of an AVFrame struct?
[01:29] <Plorkyeran> memcpy?
[02:24] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:dc35b77b28f4: ff_h264_decode_picture_parameter_set: use the correct sps
[02:24] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:ccdaa31d8647: ff_h264_decode_picture_parameter_set: fix return code
[02:24] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:26482ae716ba: h264_sei/decode_picture_timing: do not depend on h->sps being set
[02:24] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:edabbfbaf2c9: ff_h264_decode_seq_parameter_set: dont set h->sps
[02:24] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:5a9e3760495e: h264: check mb_width/height
[02:52] <michaelni> ubitux, why is the "--enable-shared + libs " fate test failing?
[02:53] <michaelni> gcc 4.8 ?
[02:53] <michaelni> iam asking cuz i couldnt reproduce with older gcc
[02:59] <cone-652> ffmpeg.git 03Paul B Mahol 07master:f5846dc98cee: id3v2: stop ignoring text encoding for chapter titles
[03:01] <michaelni> ubitux, seems the ones failing are 20130425 gcc and the ones passing are 20130411
[03:17] <ubitux> michaelni: i think that's simply because i added libaacplus
[03:18] <ubitux> and the fate test might not explicit the codec or something
[03:18] <ubitux> so it might select libaacplus instead of the native decoder
[03:18] <ubitux> (libaacplus and libfdk-aac)
[03:19] <durandal_1707> its shared build?
[03:21] <michaelni> ubitux, its failing in a vc1 test
[03:22] <ubitux> ah, then i dunno
[03:22] <ubitux> i'll try tomorrow
[03:22] <michaelni> also http://fate.ffmpeg.org/report.cgi?time=20130505101638&slot=x86_64-darwin-gcc-4.8
[03:22] <michaelni> fails similarly
[03:23] <michaelni> both 4.8
[03:23] <michaelni> in absence of other ideas a bisect might be interresting
[03:24] <ubitux> mmh im on a slow laptop right now, but i have 4.8, give me a few minutes
[03:24] <michaelni> no hurry, it can wait till tomorrow or when you have time
[03:27] <ubitux> 'got a gcc upgrade, possibly fixing that bug
[03:37] <ubitux> fate-vc1-ism passes here with 4.8.0 20130425
[03:38] <ubitux> i'll have a look later if it's still reproducible after the upgrade i did
[03:40] <ubitux> ah and i'm going to clean the ccache as well, just in case
[03:41] <durandal_1707> why you setup .0 compiler at first place?
[03:41] <ubitux> because i can
[04:00] <highgod> Hi, all, I want to ask a question, what does ffserver do? I just know it is a server, but I don't konw what the capability it is, thanks all
[04:01] <durandal_1707> there is documentation
[04:06] <highgod> @durandal_1707:oh thanks, is there any application use it?
[04:07] <durandal_1707> highgod: just durandal_1707 
[04:09] <durandal_1707> highgod: http://ffmpeg.org/ffserver.html
[04:09] <highgod> :)thanks
[04:09] <durandal_1707> michaelni: i have hard time understanding asm in swr
[04:10] <durandal_1707> is there any asm still to be written?
[04:12] <michaelni> theres always more and better asm that could be written
[04:15] <michaelni> also with every generation of cpus new instructions get added, so asm has to evolve too
[09:51] <BuxiNess> Morning
[09:52] <michaelni> morning BuxiNess 
[10:27] <cone-833> ffmpeg.git 03Diego Biurrun 07master:7abd35a1ffae: avformat: Add av_cold attributes to init functions missing them
[10:27] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:5c2ccf83df3b: Merge commit '7abd35a1ffaecfd79fa07b801621ee11ab595c43'
[10:34] <cone-833> ffmpeg.git 03Diego Biurrun 07master:bd8ac882140a: avcodec: Add av_cold attributes to end functions missing them
[10:34] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:b50ab3a4c7a1: Merge commit 'bd8ac882140a38868c33c000a430a1292a352533'
[10:42] <cone-833> ffmpeg.git 03Diego Biurrun 07master:2d5fd7886a29: bgmc: Do not mark ff_bgmc_decode_init() as av_cold
[10:42] <cone-833> ffmpeg.git 03Luca Barbato 07master:46c1917350f5: nut: use a define for the nut version
[10:42] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:77c85cbd9d4d: Merge commit '46c1917350f58dfab84e41919e6c02d43950db8c'
[10:55] <burek> is there a recommended way to add dynamic text to a video using ffmpeg? something like gps data, provided with external application and such things
[10:55] <durandal_1707> burek: drawtext supports process_command
[10:56] <burek> great :) will look into that :) thanks man :)
[10:56] <cone-833> ffmpeg.git 03Luca Barbato 07master:ac87eaf856e0: wav: Always seek to an even offset
[10:56] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:5cb9093aa848: Merge remote-tracking branch 'qatar/master'
[10:57] <durandal_1707> michaelni: when you gonna merge rsd/redspakr/.. ?
[11:08] <michaelni> durandal_1707, I dont remember anyone asking for it to be merged. But i can anytime, is it jamrial/rsd on github ?
[11:09] <durandal_1707> michaelni: i can merge it too...
[11:09] <durandal_1707> i prefer to not commit as merget though - cleaner log
[11:25] <durandal_1707> michaelni: so pick it: me or you
[11:42] <michaelni> durandal_1707, i can do it once the extradata issue is fixed
[11:44] <durandal_1707> michaelni: extradata is changed in each packet
[11:47] <michaelni> durandal_1707, the data is only changed in rsd_read_header() which is called just once per file
[11:47] <michaelni> for rsd
[11:48] <durandal_1707> indeed
[11:49] <michaelni> and same for redspark_read_header() and redspark
[11:49] <durandal_1707> but not for brstm...
[11:50] <michaelni> yes that needs to use something else than avctx.extradata
[12:03] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:09602dbe7a64: wav_seek_tag: fix integer overflow
[12:03] <cone-833> ffmpeg.git 03highgod0401 07master:9a09a9fc43e0: lavfi/unsharp: fix opencl crash on 64bit linux
[12:14] <kierank> 23:23:28 <"Daemon404> part of 'decoding' imho --> no it's not
[12:14] <kierank> applying rff is part of *display*
[12:15] <kierank> such things are well defined in mpeg
[12:17] <cone-833> ffmpeg.git 03highgod0401 07master:41aa19999858: lavu/opencl_internal: use av_opencl_errstr
[12:17] <cone-833> ffmpeg.git 03highgod0401 07master:e215cb5e9670: lavfi/unsharp_opencl: use av_opencl_errstr
[12:43] <cone-833> ffmpeg.git 03Carl Eugen Hoyos 07master:76f43cbe2648: Remove a superfluous include from libavcodec/options_table.h.
[12:43] <cone-833> ffmpeg.git 03Michael Niedermayer 07master:0b5f4fdc381f: Merge remote-tracking branch 'cehoyos/master'
[13:59] <xlinkz0> so there is a way to take a normal video and make it 3d?
[13:59] <av500> yes
[13:59] <av500> lots of TVs do that
[13:59] <av500> its crap of course
[14:00] <av500> a useless feature trying to sell 3D TVs
[14:05] <durandal_1707> you can't make normal video 3d
[14:08] <av500> you can fake it
[14:14] <michaelni> you can hire a director, actors, cameraman and so on and turn any 2d movie in a 3d one 
[14:16] <durandal_1707> and compare that to recording movie first time with right camera
[14:16] <iive> they actually do it with already filmed 2d movies, they cut out sections and displace them to create depth.
[14:17] <iive> the first few movies after avatar were done this way.
[14:17] <xlinkz0> so what is this stereoscopic thing in ffmpeg?
[14:18] <xlinkz0> a fake filter?
[14:22] Action: durandal_1707 omg! its fake!
[14:38] <durandal_1707> seriously, why would filter be fake?
[14:39] <av500> there is so much evil in the world these days
[14:40] <durandal_1707> what is evil in this specific case?
[14:40] <av500> no idea
[14:40] <av500> I dont even know what that filter does
[14:42] <xlinkz0> i meant to ask if it is similar to what tv's do and if it is any better
[14:45] <iive> xlinkz0: there are videos that provide 2 separate pictures for each eye, but there are a number of different ways to encode the, side by side, top/bottom pictures, interlace (fields) and whatever mvc provides. so there is need to convert from one format to the other.
[14:46] <iive> the, -> them,
[14:48] <xlinkz0> i see
[14:52] <Compn> hopefully 3d is just a fad
[14:52] <Compn> because idiot movie directors only use it to put things in your face, instead of using it to enhance the experience
[15:12] <ubitux> maybe xwininfo could be mentionned in that wiki page
[15:12] <ubitux> eventually with a small 3-lines shell script to extract window position & run a capturing ffmpeg session
[16:10] <durandal_1707> michaelni: working on gbrap input/output?
[16:31] <michaelni> durandal_1707, not ATM
[16:31] <michaelni> so you can work on it if you want ...
[16:33] <durandal_1707> michaelni: i could try, but i'm not very happy what become of RGBA64 output patch
[16:33] <durandal_1707> *already trying
[16:34] <durandal_1707> what i need to do to read alpha? nothing?
[16:36] <cone-125> ffmpeg.git 03highgod0401 07master:b02f073ad492: lavfi/deshake_opencl: use ff_opencl_set_parameter
[16:40] <michaelni> iam not aware of anything special being needed for alpha, we already support packed rgba and planar yuva
[16:43] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:d2bf8aecd35a: j2kdec: frame multithreading support
[16:47] <durandal_1707> michaelni: looks like i got output working for rgba->gbrap, but it crashes for rgb24->gbrap case
[16:52] <durandal_1707> there is code that fills one plane with 255
[16:59] <michaelni> thats for the case where input has no alpha but output has IIRC
[17:00] <durandal_1707> yes, i have still that amount of brain to conclude that
[17:00] <durandal_1707> thing is i get black for alpha plane, when i should get white
[17:05] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:f4341c64be19: avfilter: add av_cold to uninit()
[17:05] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:8cdf95d9b8f9: snow: add av_cold where appropriate
[17:07] <durandal_1707> fixed segv, now only black issue remains
[17:27] <durandal_1707> hmm gbrap output looks good and conversion from gbrap to 444p is ok, but to subsampled its wrong
[17:39] <durandal_1707> huh gbrp16 sets output func
[17:40] <durandal_1707> michaelni: sent patch, awaiting your review
[18:15] <michaelni> durandal_1707, make fate fails with the patch
[18:16] <michaelni> also when you update the checksums please look at if the newly added entries "look" ok
[19:38] <Compn> http://gynvael.vexillium.org/ext/ffmpeg_log.txt
[19:40] <Compn> they've been working hard :)
[19:54] <saste> Compn, 666, the beast inside ffmpeg
[19:55] <saste> trac is ill
[20:52] <cone-588> ffmpeg.git 03Diego Biurrun 07master:9e5e76ef9ea8: x86: More specific ifdefs for dsputil/hpeldsp init functions
[20:52] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:32cc7dacde6d: Merge commit '9e5e76ef9ea803432ef2782a3f528c3f5bab621e'
[20:58] <cone-588> ffmpeg.git 03Diego Biurrun 07master:dc1b328d0df6: x86: hpeldsp: Only compile MMX hpeldsp code if MMX is enabled
[20:58] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:a6e782434a21: Merge commit 'dc1b328d0df6e5ad5ff0ca4ae031e08466624f9c'
[21:06] <cone-588> ffmpeg.git 03Diego Biurrun 07master:110796739ab3: x86: hpeldsp: Move avg_pixels8_x2_mmx() out of hpeldsp_rnd_template.c
[21:06] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:108e2ae8299a: Merge commit '110796739ab32854dc0b6b0a1c95e6ae98889062'
[21:38] <cone-588> ffmpeg.git 03Nicolas Bertrand 07master:28816f9db877: jpeg2000: Fix uninitialized data errors pointed out by valgrind/memcheck
[21:38] <cone-588> ffmpeg.git 03Martin Storsjö 07master:6c5eac9a603a: swscale: Remove unnecessary backslashes
[21:38] <cone-588> ffmpeg.git 03Nicolas Bertrand 07master:a06f0cef5bd4: jpeg2000: fix for uninitialized data errors in valgrind/memcheck
[21:38] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:050b48f39783: Merge commit '6c5eac9a603a91a707d030d96ba97e2ef6b3ce50'
[21:39] <michaelni> BuxiNess, your incomplete valgrind patch was applied to libav, so i applied&merged it too
[21:39] <michaelni> ill post a patch that fixes the length==1 case in a moment
[21:46] <michaelni> BuxiNess, patch posted, please review
[21:52] <cone-588> ffmpeg.git 03Martin Storsjö 07master:6e293d111fca: swscale: Use alpha from the right row in yuva2rgba_c
[21:52] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:b3e9fd1fe578: Merge remote-tracking branch 'qatar/master'
[22:11] <michaelni> BuxiNess, posted a second patch that fixes the memleaks, please review
[22:31] <cone-588> ffmpeg.git 03Rainer Hochecker 07master:06d8fdb660fd: Fix vdpau vc1 interlace modes also when using decoder vc1_vdpau.
[22:31] <cone-588> ffmpeg.git 03Carl Eugen Hoyos 07master:8b8fa09b654e: vc1dec: Call ff_vdpau_vc1_decode_picture() once for each field.
[22:31] <cone-588> ffmpeg.git 03Carl Eugen Hoyos 07master:905e1e4b521b: Cosmetics: Fix indentation and add braces after last commit.
[22:54] <cone-588> ffmpeg.git 03Clément BSsch 07master:a82cc54fb5ce: lavf/mpsub: remove sscanf from probing.
[22:56] <cone-588> ffmpeg.git 03Michael Niedermayer 07master:974420a701c7: jpeg2000dec: fix memleaks in error cases in jpeg2000_decode_frame()
[00:00] --- Tue May  7 2013