[Ffmpeg-devel-irc] ffmpeg-devel.log.20121009

burek burek021 at gmail.com
Wed Oct 10 02:05:02 CEST 2012


[00:00] <ubitux> given a ./ffprobe -f lavfi movie=... -show_frames
[00:01] <saste> filter buffer -> avframe
[00:01] <saste> what am i missing?
[00:02] <ubitux> the lavfi device is a classic format filling AVPackets
[00:02] <ubitux> so you av_read_frame() and just get a AVPacket
[00:02] <ubitux> how can the tool (in this case ffprobe, but it's the same problem with the other) can get extra info?
[00:03] <ubitux> i may be missing something as well :p
[00:05] <ubitux> saste: you don't see the problem?
[00:06] <ubitux> in the case of a lavfi device, the buffersink is only in the scope of that device/format
[00:08] <ubitux> well anyway, the other way would be to make a sink printer like you proposed, with the probe writers :p
[00:09] <ubitux> also, there is still side_data :-°
[00:13] <saste> well yes lavfi is like a rawvideo demuxer, so can't store metadata info
[00:13] <saste> right now only tiff supports that
[00:13] <saste> yes side data is a possibility
[00:14] <ubitux> what should i put into the side_data?
[00:14] <saste> metadata?
[00:14] <ubitux> badly worded sorry
[00:14] Action: saste don't have a clue about side_data
[00:14] <ubitux> how am i supposed to store it? :)
[00:14] <ubitux> i guess i can't store a pointer
[00:15] <saste> you can copy the textual data
[00:15] <ubitux> klv? :p
[00:15] <cone-907> ffmpeg.git 3Tim Nicholson 7a02762995bc5: movenc.c: Force correct value for "Samples per packet" for pcm audio * 3http://tinyurl.com/9h85j6l3
[00:15] <saste> but again, that's beyond my current knowledge
[00:15] <ubitux> [len][key][len][value] ? :p
[00:16] <ubitux> or well, just key\0value\0 should do the trick.
[00:16] <ubitux> ok, let's try this.
[00:28] Action: beastd is going to sleep
[00:28] <beastd> bye...
[00:29] <ubitux> i think i'll do the same.. :)
[00:29] Action: ubitux &
[00:29] <Daemon404> fg
[00:33] <ubitux> no :(
[01:35] <burek> <nevcairiel> why are some hashes pink? :d
[01:35] <burek> you noticed it well :)
[01:35] <burek> <cone-907> ffmpeg.git 3Anton Khirnov 7716d413c1398: Replac
[01:35] <burek> this should have been
[01:36] <burek> <cone-907> ffmpeg.git 3Anton Khirnov 7 716d413c1398: Replac
[01:36] <burek> (if you look at the url)
[01:36] <burek> the problem is the bot uses mirc color codes
[01:36] <burek> ctrl+K+7
[01:36] <burek> and when another digit is added, it becomes ctrl+K+77
[01:39] <burek> just using double digits, like 07, should work
[01:40] <burek> btw, who maintains cone-907 ?
[02:01] <cone-907> ffmpeg.git 3Carl Eugen Hoyos 7b49d94e4f25e: Support decoding of targa files with 32bit palette. * 3http://tinyurl.com/8qd3zzv3
[02:03] <llogan> burek: it's an irker bot I think. maybe j-b or thresh?
[02:05] <burek> oh, ok, they are both here, so they can read the logs
[02:59] <RT|AO> so where are the separators in bot's commit log? separating with color doesn't work in plain text logs.
[03:14] <Compn> hah, got firefox 15 working on win2k
[03:18] <Compn> and most other msvc2010 compiled apps i hope
[03:18] Action: Compn flaunts to Daemon404
[03:18] <Compn> :P
[03:21] <Compn> only took a bunch of hacked beta winxp dlls :P
[03:21] <Compn> a completely unofficial service pack :D
[03:36] Action: llogan sees no use for win2k
[03:39] <ohsix> it's an operating system :]
[03:44] <Compn> it kinda was an os
[03:44] <Compn> 12 years ago
[03:45] <ohsix> and xp 11 years ago, heh
[03:47] <Compn> kinda makes me feel better about win2k :P
[03:48] <TheRyuu> that sounds retarded
[03:49] <ohsix> differently abled?
[03:49] <ohsix> xp is just as obsolete as win2k, i'm still using xp
[03:50] <ohsix> there's not much practical consequence
[03:50] <ohsix> software that doesn't work on win2k or xp you just wouldn't use, hasn't really happened yet
[03:52] <TheRyuu> I was under the impression a lot of stuff doesn't work for 2k
[03:52] <TheRyuu> basically everything should work with xp (sp3) though
[03:53] <ohsix> about the only thing that may not work is hardware
[03:53] <ohsix> and things that use XP, Vista, windows 7 features that define the software (surprisingly almost none of this exists)
[03:54] <ohsix> it's always been common practice to dynamically load or detect new features, too; as part of being able to run on different versions of windows, you may get degraded functionality but stuff just failing to start is very unusual
[03:54] <ohsix> even the way microsoft distributes runtimes comports with this usage
[04:24] <cone-907> ffmpeg.git 3Michael Niedermayer 78da7907a4a89: adpcmenc: switch to av_assert() * 3http://tinyurl.com/8fr4kbx3
[04:24] <cone-907> ffmpeg.git 3Michael Niedermayer 772c2d8a34692: dsputil: convert asserts to av_asserts * 3http://tinyurl.com/8gmeafw3
[05:27] <cone-907> ffmpeg.git 3Michael Niedermayer 7d07940b76d03: motion_est: switch asserts to av_asserts * 3http://tinyurl.com/8kuq7su3
[05:54] <cone-907> ffmpeg.git 3Bobby Bingham 73d9cdfdce76a: targa: use named constants for flag values * 3http://tinyurl.com/8ms9c5y3
[05:54] <cone-907> ffmpeg.git 3Bobby Bingham 7c2eec3df8996: targa: support 2-way and 4-way interleaved files * 3http://tinyurl.com/952gzr83
[05:54] <cone-907> ffmpeg.git 3Bobby Bingham 750787fe350c2: targa: remove unused context members * 3http://tinyurl.com/9qgs3jg3
[05:54] <cone-907> ffmpeg.git 3Bobby Bingham 7b56f94cc363c: targa: cosmetics - add some whitespace * 3http://tinyurl.com/9777r6m3
[06:19] <cone-907> ffmpeg.git 3Michael Niedermayer 7e73bac484f71: configure: add support to nicely enable ftrapv * 3http://tinyurl.com/9xvwzfc3
[06:19] <cone-907> ffmpeg.git 3Michael Niedermayer 703f5043f5d13: eval: Fix eval test with ftrapv * 3http://tinyurl.com/8j5u6sl3
[06:19] <cone-907> ffmpeg.git 3Michael Niedermayer 71e83e6ad7a8b: ra144: fix code with ftrapv. * 3http://tinyurl.com/8nfcnkg3
[06:34] <koorogi> Who do I talk to about being able to close tickets I've fixed?
[07:17] <thresh> burek: it is a bug in irker, was fixed this night
[09:53] <burek> thresh great :)
[10:07] <ubitux> huh, is AVFrame->metadata unused at the moment?
[10:09] <durandal_1707> it should be used by GeoTIFF
[10:11] <ubitux> ah right, indeed
[10:11] <ubitux> thx
[10:12] <durandal_1707> also png could make use of it...
[10:15] <cone-932> ffmpeg.git 3Carl Eugen Hoyos 761a9f099b7da: Write 32bit palette to Targa files. * 3http://tinyurl.com/8crp2et3
[10:32] <durandal_1707> hmm, why fate reports memory leak for tak that have single key in ape2 tags?
[10:35] <ubitux> durandal_1707: -f s16le -f crc -
[10:35] <ubitux> -f re-dup
[10:35] <durandal_1707> ?
[10:35] <ubitux> in the fate test
[10:35] <ubitux> remove -f s16le
[10:36] <ubitux> it will fix the leak
[10:36] <durandal_1707> but other tests had it too
[10:36] <ubitux> i don't think so
[10:38] <durandal_1707> fate-lossless-monkeysaudio: CMD = md5 -i $(SAMPLES)/lossless-audio/luckynight-partial.ape -f s16le 
[10:39] <ubitux> md5
[10:39] <ubitux> ` crc
[10:39] <ubitux> md5 doesn't add a -f md5 or something
[10:40] <ubitux> crc will add a -f crc
[10:40] <ubitux> and then, format=av_strdup("s16le"); [...] format=av_strdup("crc")
[10:41] <durandal_1707> leak should not happen
[10:41] <ubitux> see the av_freep() commented in the appropriate place
[10:41] <ubitux> michael removed it because it was causing problems
[10:42] <ubitux> durandal_1707: see a1bcc76e6036e78f25cbb7323c145056cfca9d93
[10:42] <ubitux> but it was reverted
[10:43] <ubitux> durandal_1707: the leak should not happen, but the -f should not appear two times in the fate test anyway
[10:43] <ubitux> you can remove the -f s16le, it doesn't affect your test
[10:44] <ubitux> related: 6a0dfe3b9dc6b9420b25bfd089c655583cf045d6
[10:51] <cone-932> ffmpeg.git 3Paul B Mahol 73a2d3df0e02d: fate-lossless-tak: remove unneeded -f s16le * 3http://tinyurl.com/9rzldn63
[10:53] <thresh> I've updated the irker instance on videolan git, the colors should be fixed now :)
[11:02] <xxthink> how to profile ffmpeg?
[11:02] <xxthink> there are no configure options now
[11:02] <cone-932> ffmpeg.git 03Paul B Mahol 07238e904df398: DTS-HD demuxer * 03http://tinyurl.com/9xfmf4k03
[11:02] <burek> it works good now thresh :)
[11:03] Action: ubitux still doesn't like tinyurl :(
[11:04] <j-b> this SUXX
[11:04] <thresh> what
[11:10] <j-b> What is a DTSHD demuxer ? a demuxer for .dtshd files? How is that different from .dts? samples?
[11:11] <durandal_1707> j-b: you do not have samples for .dtshd ?
[11:12] <j-b> durandal_1707: unmuxed? no.
[11:13] <durandal_1707> what?
[11:13] <durandal_1707> you have files with dtshd extension or not?
[11:13] <ubitux> xxthink: mmh, valgrind/cachegrind, START_TIMER, ... ?
[11:14] <ubitux> callgrind*
[11:14] <xxthink> ok
[11:14] <j-b> durandal_1707: no.
[11:14] <xxthink> ubitux: I use oprofile
[11:14] <xxthink> add the -pg options to the CFLAGS
[11:15] <ubitux> if it makes sense, maybe add a --enable-oprofile :p
[11:19] <durandal_1707> j-b: http://dotwhat.net/dtshd/10632/
[11:24] <av500> j-b: I read on internet your vlc is borken
[11:27] <j-b> av500: how is it?
[11:27] <burek> ubitux, there could be a simple redirect, created within .htaccess on videolan that would translate all these: http://git.videolan.org/?p=ffmpeg.git;a=commit;h=61a9f099b7da38722f758db38d9c89d1b39f3a87 into something shorter like this: http://url.videolan.org/ffmpeg/61a9f099b7da
[11:28] <burek> that way you'd avoid tinyurl or any other external dependency
[11:28] <av500> j-b: that avengers dvd or sp
[11:28] <av500> so
[11:28] <j-b> av500: indeed. ARCCOS DRM
[11:28] <ubitux> burek: if you trim the hash to 7 char it's short enough
[11:29] <burek> I just followed cone's output :)
[11:29] <burek> but ok, even better with 7 chars only
[11:37] <j-b> durandal_1707: that does not answer the question.
[11:37] <j-b> ubitux: burek: I tried to register v.lc domain... I failed
[11:38] <ubitux> :(
[11:38] <ubitux> what about g.it?
[11:45] <burek> j-b, v.lc is already taken, but the owner is: Nic LC Admin
[11:45] <burek> so you might ask for just a DNS A record maybe
[11:45] <j-b> burek: nic is the registar of .lc
[11:45] <burek> yes i know
[11:46] <j-b> I asked them :)
[11:46] <burek> :)
[11:46] <j-b> they told me: "do a correct business proposal"
[11:47] <durandal_1707> j-b: http://www.atsc.org/cms/index.php/standards/standards/296-a1032012-non-real-time-content-delivery
[11:50] <burek> j-b how about videol.an
[11:51] <j-b> .an exists?
[11:51] <ubitux> batman ananananaan
[11:51] <burek> http://www.controloye.com/whois/an.php
[11:51] <av500> j-b: yes
[11:51] <burek> anyway domain whois shows it's not been taken
[11:51] <j-b> ok, that's because our name registar does not
[11:52] <durandal_1707> video.xxx
[11:52] <av500> +1
[11:52] <av500> hotvlc.xxx
[11:53] <j-b> probably already taken
[11:53] <av500> bringing you the best of variable length coding...
[11:53] <ubitux> Le domaine videol.an existe déjà, et ne peut pas être créé.
[11:53] <ubitux> according to ovh
[11:53] <j-b> getv.lc
[11:54] <av500> j.bk
[11:55] <j-b> unlikely
[11:55] <j-b> if I cannot get v.lc with a popular software, I doubt I can get a single-letter domain for me
[11:56] <durandal_1707> what is correct business proposal?
[11:57] <j-b> that is a good question
[11:57] <j-b> but > 10k probably
[11:57] <durandal_1707> per year?
[11:58] <j-b> per acquisition, I guess
[12:31] <burek> btw, ffmp.eg is free if there is an interest for such thing.. http://lookup.egregistry.eg/english.aspx
[12:35] <av500> people dont remember domain names anyway
[12:35] <av500> they google "yahoo"
[12:35] <av500> I stopped typing domains long ago
[12:35] <av500> I just mash a few matching letters into el goog
[12:35] <av500> fmfge :
[12:35] <av500> :)
[12:44] <divVerent> ffmpeg's -vf scale is great
[12:44] <divVerent> expression evaluator...
[12:57] <durandal_1707> what extensions libdca have?
[13:10] <cone-932> ffmpeg.git 03Janne Grunau 07ea14a655f7b8: avutil: skip old_pix_fmts.h since it is just a list * 03http://tinyurl.com/8qgjj4p03
[13:10] <cone-932> ffmpeg.git 03Mans Rullgard 0727a310e3813c: doc: allow building with old texi2html versions * 03http://tinyurl.com/8uq22tk03
[13:10] <cone-932> ffmpeg.git 03Martin Storsjö 0766d652cbf38e: rtpdec_vp8: Make the depacketizer implement the latest spec draft * 03http://tinyurl.com/95rahqn03
[13:10] <cone-932> ffmpeg.git 03Martin Storsjö 07c9b10cc4dbb6: rtpenc_vp8: Update the packetizer to the latest spec version * 03http://tinyurl.com/9fmn3hs03
[13:10] <cone-932> ffmpeg.git 03Mans Rullgard 078f23907f3e4c: build: add -Mdse to PGI optimisation flags * 03http://tinyurl.com/8m3jbsq03
[13:10] <cone-932> ffmpeg.git 03Mans Rullgard 07643933f51d1c: build: add LTO support for PGI compiler * 03http://tinyurl.com/8cf975t03
[13:10] <cone-932> ffmpeg.git 03Mans Rullgard 07f79364b2c30a: ppc: fix Altivec build with old compilers * 03http://tinyurl.com/9javb5w03
[13:10] <cone-932> ffmpeg.git 03Yusuke Nakamura 07e04826c34e9b: file: Set the return value type for lseek to int64_t. * 03http://tinyurl.com/93ycxaz03
[13:10] <cone-932> ffmpeg.git 03Martin Storsjö 07c136a813d77e: rtp: Support packetization/depacketization of opus * 03http://tinyurl.com/9zza5m303
[13:10] <cone-932> ffmpeg.git 03Diego Biurrun 07f75f4194d1ea: Restructure av_log_missing_feature message * 03http://tinyurl.com/8n8kxn303
[13:10] <cone-932> ffmpeg.git 03Diego Biurrun 07a75b9a180476: mingw/cygwin: Stop adding -fno-common to gcc CFLAGS * 03http://tinyurl.com/8zhwlsp03
[13:11] <cone-932> ffmpeg.git 03Michael Niedermayer 07ef9fe5bedd19: Merge remote-tracking branch 'qatar/master' * 03http://tinyurl.com/8casapo03
[14:09] <maker> FYI: Firefox 15 signals https://ffmpeg.org as insecure
[16:29] <burek> <maker> FYI: Firefox 15 signals https://ffmpeg.org as insecure
[16:29] <burek> https://support.google.com/chrome/bin/answer.py?hl=en&answer=95617&p=ui_security_indicator
[16:29] <burek> it shows me: The site uses SSL, but Google Chrome has detected insecure content on the page.
[16:30] <burek> which means there are some src="http://..." content inside html
[16:30] <burek> i.e. not all content is from https://ffmpeg.org
[16:31] <burek> for example this image http://americancensorship.org/images/stop-censorship-small.png is included above the text "FFmpeg supports the fight against American Internet censorship."
[16:31] <burek> and that's considered an insecure content
[16:40] <maker> burek: ah uh
[16:55] <cone-932> ffmpeg.git 03Anuj Mittal 07ce19aec15b42: Check resync marker only when enabled. * 03http://tinyurl.com/8p5bgkx03
[16:55] <cone-932> ffmpeg.git 03Anuj Mittal 078d2e0e2c7058: Check for resync marker based on vop coding type- vop_fcode_forward and vop_fcode_backward L * 03http://tinyurl.com/9j5k3bz03
[17:18] <ubitux> saste: any reason you're not storing the the g_XX as bitfields?
[17:19] <saste> ubitux: i just ported the original code
[17:19] <saste> i didn't feel like i wanted to process the table with an hand-written program
[17:20] <saste> ah you say to change uint8_t with uint8_t:1?
[17:20] <saste> yes that's possible
[17:20] <saste> what I'm mostly concerned is the integer implementation, which is rather tricky
[17:20] <saste> *concerned with
[17:21] <ubitux> i don't think you can do a uint8_t:1
[17:21] <ubitux> you'll need to change the code :p
[17:21] <saste> whatever syntax it is
[17:21] <ubitux> i don't think the syntax will allow it in this context
[17:22] <saste> anyway reply by email or i'll discard it in the next iteration (which may well be after a few months)
[17:22] <ubitux> i was just asking a random question :)
[17:22] <cone-932> ffmpeg.git 03KO Myung-Hun 073a45688abc5c: lavc: include os2threads.h in vp8.h if HAVE_OS2THREADS is enabled * 03http://tinyurl.com/9yv3bxu03
[17:22] <cone-932> ffmpeg.git 03Andrey Utkin 07d2b18c8f5b08: Introduce ff_network_wait_fd_timeout() * 03http://tinyurl.com/9ncesq503
[17:22] <cone-932> ffmpeg.git 03Andrey Utkin 07949acefc11b4: tcp: use timeout option consistently * 03http://tinyurl.com/9ma2ojv03
[17:23] <ubitux> saste: anyway, please at least commit the '-' thing for the help output :(
[17:40] <saste> ubitux, i wonder if it is a good idea after all...
[17:40] <ubitux> why?
[17:40] <ubitux> i'm talking about removing it only for the filters, right?
[17:40] <saste> the output looks less structured
[17:41] <saste> no you can't do that only for filters
[17:41] <saste> all or nothing
[17:41] <ubitux> mmh.
[17:42] <saste> some reindentation may help
[17:42] <saste> but that's kinda subjective, also that would need some changes in opt.c code
[17:42] <ubitux> that was one of the other patch
[17:42] <ubitux> i would personally add 2 spaces index
[17:42] <ubitux> indent*
[17:43] <ubitux> (and more padding for the other options btw)
[17:44] <saste> what about XML output?
[17:45] <saste> uhm no i never reindented
[18:26] <cone-932> ffmpeg.git 03Andrey Utkin 072e009c6042bd: tcp: accept params through avio_open2() options * 03http://tinyurl.com/95zg2bo03
[18:26] <cone-932> ffmpeg.git 03Andrey Utkin 07b6f435fbc87c: http: add 'timeout' AVOption * 03http://tinyurl.com/9k4fwu703
[20:12] <michaelni> if someone wants to setup a ffmpeg IOC fate box, it seems there are now also binaries on the IOC page (http://embed.cs.utah.edu/ioc/)
[20:33] <Compn> michaelni : guess its good that ffmpeg supports clang :)
[20:33] <Compn> ehe
[20:33] <Compn> since ioc is based on clang
[20:46] <Daemon404> libav has had an ioc fate instance for ages
[20:46] <Daemon404> should be trivial to set up
[20:47] <Daemon404> on another note
[20:47] Action: Daemon404 still waiting for baptiste
[20:48] <nevcairiel> he answered the same day for me when i setup mine
[20:48] <nevcairiel> you must've hit him at a bad time =p
[20:58] <burek> one question, what does 'start' represent in input's info: Duration: 00:07:35.04, start: 0.481967, bitrate: 23837 kb/s
[20:58] <burek> or this, if more convenient: http://pastie.org/5025652
[21:45] <burek> it seems that -map option is broken
[21:46] <burek> http://pastie.org/5025843
[21:46] <burek> http://pastie.org/5025918
[21:51] <michaelni> burek, is that a -map O or -map 0 ?
[21:53] <michaelni> btw, if the 2 l00k the same t0 y0u then y0u maybe sh0uld try an0ther f0nt
[21:53] <burek> geez .. he typed letter 'O'
[21:53] <burek> thanks :D
[22:20] <ubitux> -        [X264_LOG_INFO]    = AV_LOG_INFO,
[22:20] <ubitux> +        [X264_LOG_INFO]    = AV_LOG_VERBOSE,
[22:20] <ubitux> hehe
[22:20] <cone-932> ffmpeg.git 03Duncan Salerno 07bd2f8e8f79e9: Fix second use of AVOptions in HLS * 03http://tinyurl.com/9oy5s9903
[22:20] <ubitux> reminds me something :))
[22:27] <nevcairiel> INFO = INFO, i fail to see why everyone wants to change it =p
[22:27] <ubitux> :)
[22:31] <cone-932> ffmpeg.git 03Clément BSsch 076ac5e3fe9d92: lavfi/silencedetect: add av_opt_free() call. * 03http://tinyurl.com/8bfsj4w03
[22:41] <Daemon404> nevcairiel, inorite
[22:41] <Daemon404> because it offends them or something
[22:41] <Daemon404> how DARE a lib have its own output
[22:42] <ubitux> TAG:lavfi.silence_start=3.40735
[22:42] <ubitux> yay.
[22:43] <Daemon404> >_>
[22:43] <Daemon404> outputting via tags...
[22:43] <Daemon404> really?
[22:44] <ubitux> ?
[22:45] <ubitux> lavfi will use the metadata for communication
[22:45] <Daemon404> i see
[22:45] <ubitux> at the end of the chain ffprobe is able to dump them :)
[22:45] <Daemon404> thats seems majorly bad for anything with lots
[22:45] <Daemon404> like, i was thinking, scene change info
[22:45] <ubitux> so you get the xml/json/etc benefits
[22:45] <Daemon404> this will be really ugly...
[22:45] <ubitux> Daemon404: yes that's what i was planing to do
[22:45] <ubitux> (see my mail)
[22:45] <ubitux> very easy to do now that i did the hard part
[22:57] <cone-932> ffmpeg.git 03Duncan Salerno 07f3f35f7430f1: crypto should allow passing of options to the underlying protocol via the url_open2 interface * 03http://tinyurl.com/9m3hngr03
[00:00] --- Wed Oct 10 2012


More information about the Ffmpeg-devel-irc mailing list