[Ffmpeg-devel-irc] ffmpeg-devel.log.20130312

burek burek021 at gmail.com
Wed Mar 13 02:05:02 CET 2013


[00:00] <michaelni> (i still have some deadlocks left that i need to debug before it can be pushed ...)
[03:04] <michaelni> what happened to cone ?!
[03:12] <Compn> cone vlc ?
[03:17] <michaelni> Compn, yes
[03:18] <michaelni> i pushed a commit and cone did not report it ...
[03:31] <Compn> gasp
[03:31] <Compn> gsoc project : design new logo for videolan...
[03:31] <Compn> ehe
[03:33] <michaelni> google summer of cone
[03:35] <BBB> michaelni: hah now you can push my awesome hpeldsp patchset
[03:35] <BBB> (or maybe I should rebase it)
[03:36] <BBB> is fate functional again?
[03:38] <michaelni> fate should always have been fine or do i misunderstand something ?
[03:38] <Compn> fate looks ok here ...
[03:38] Action: michaelni realized that he messed up the config of some clients 
[03:39] <michaelni> iconv and qemu 
[03:45] <michaelni> BBB, Daemon404  if there are any important fixes i should cherry pick please tell me the git hash
[03:46] <michaelni> we are sadly still 4 days behind the merging thanks to the evil plan
[03:47] <michaelni> j-b, is cone healthy ? it didnt report the last 9 ffmpeg commits
[04:04] <cone-277> ffmpeg.git 03Anton Khirnov 07master:3b199d29cd59: lavc decoders: properly initialize AVFrame.
[04:04] <cone-277> ffmpeg.git 03Michael Niedermayer 07master:b5e5ed8ea707: Merge commit '3b199d29cd597a3518136d78860e172060b9e83d'
[04:12] Action: michaelni hugs cone
[04:13] <michaelni> you are back :)
[04:17] <BBB> Program received signal EXC_BAD_ACCESS, Could not access memory.
[04:17] <BBB> Reason: 13 at address: 0x0000000000000000
[04:17] <BBB> 0x0000000100556123 in ff_bswap32_buf_ssse3 ()
[04:17] <BBB> ?
[04:17] <BBB> I get a lot of fate problems
[04:18] <BBB> do I need to make cleancleanall or so?
[04:21] <michaelni> make distclean is sure worth a try
[04:21] <michaelni> it could be a bug too though, the evil plan patches where quite big and nasty and sure introduced bugs
[07:52] <nevcairiel> michaelni: i think i found the reason for the memleak with the movie source, but i'm not sure if its a bug in lavfi or the avdevice for lavfi input .. it calls av_buffersink_get_buffer_ref with the PEEK flag, which causes lavfi to increase the refcount on a buffer, and return it. but then avdevice doesnt free that buffer. I couldn't really find in the docs if you're supposed to, but i think its only logical to assume you are
[07:52] <nevcairiel> apparently before refcounting this worked without memleaks
[07:54] <nevcairiel> i would put my money on bug in lavd
[08:00] <nevcairiel> michaelni: also, these two hashes will fix mingw builds: e460aa3282962aa3e89f19e207ec13b501187949 d6d369bf1370999896500ae7eb5b9447ab635a3d if you want to cherrypick
[09:38] <j-b> michaelni: sorry, a small temporary issue
[10:32] <ubitux> why is mpegvideo calling avcodec_default_get_buffer2() instead of ff_get_buffer?
[10:34] <nevcairiel> sounds like a bad idea to do that
[10:46] Action: ubitux doesn't understand the return ret in vp56 decode function causing crashes
[10:46] <ubitux> ...where return -1 works
[10:49] <michaelni> nevcairiel, cherry picked the 2
[10:50] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:c536253854ea: snow: set frame defaults
[10:50] <cone-892> ffmpeg.git 03Martin Storsjö 07master:c48828f76367: atomic: Check for __sync_val_compare_and_swap instead of __sync_synchronize
[10:50] <cone-892> ffmpeg.git 03Hendrik Leppkes 07master:b91459e565ec: atomic: prefer gcc builtins over win32 atomics, if available.
[10:50] <ubitux> oh, got it, i'm stupid.
[11:10] Action: ubitux just rechecked every single change of the {re,}get_buffer commit
[11:10] <ubitux> michaelni: can you have a look to the avpacket patch?
[11:11] <nevcairiel> why does vp56 have issues with returning error codes other then -1?
[11:11] <ubitux> it was a stupid mistake
[11:11] <ubitux> i was re-using res
[11:11] <ubitux> which was needed afterward
[11:11] <nevcairiel> oh
[11:12] <ubitux> i've checked if i didn't make the same mistake in another place, but it seems ok
[11:14] <nevcairiel> ubitux: since you seem to know lavfi quite well as well, i have a question about the old user api, especially av_buffersink_get_buffer_ref .. if i call that with the peek flag, i'm still supposed to unref the frame it gives me, right?
[11:15] <ubitux> i don't know :)
[11:16] <nevcairiel> apparently the old api was fine when you didnt unref it, but the new one insists on it, so right now it leaks when you use the lavd lavfi device
[11:16] <nevcairiel> wonder who could tell me..
[11:16] <ubitux> Nicolas certainly can :
[11:16] <ubitux> :p
[11:16] <ubitux> he's the one maintaining the buffersink/src thing
[11:16] <nevcairiel> does he come here?
[11:17] <ubitux> he came once this week end
[11:17] <ubitux> i don't think we'll see him again soon
[11:17] <nevcairiel> then i'll mail him instead =p
[11:41] <ubitux> -rw-r--r-- 1 ubitux ubitux 119926440 Mar 12 11:40 libavcodec/libavcodec.a.new
[11:41] <ubitux> -rw-r--r-- 1 ubitux ubitux 119981632 Mar 12 11:40 libavcodec/libavcodec.a.old
[11:42] <ubitux> yepee
[11:45] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:a76ef998eb24: mpegvideo: move ff_print_debug_info2 to where ff_print_debug_info is
[11:45] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:6ae43725323b: h264/mpegvideo: fix ff_print_debug_info2 arguments
[11:45] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:53fd4f559429: avcodec: fix motion vector vissualization
[11:45] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:1ad542f11f47: get_video_buffer: add vertical padding
[11:48] <ubitux> michaelni: the thing you just fixed is what Anton said to be broken since one year?
[11:49] <nevcairiel> he also removed it, not that i think its really that useful =p
[11:50] <ubitux> iirc i used it a while ago and found that pretty fun
[11:51] <ubitux> i wonder if it was indeed broken since a year in libav, or if anton didn't want to admit he broke it :p
[11:53] <nevcairiel> sometimes source based linux distributions are a bit annoying .... rebuilding glibc with debug infos so valgrind can work properly .... zzzzz
[11:57] <michaelni> ubitux, vismv worked fine before the buffer references i used it from time to time so i would have noticed a major problem with it
[11:58] <nevcairiel> oh neat msvc 64-bit builds are fixed as well, wonder which actually fixed it
[12:06] <ubitux> it's a very fun option :p
[12:13] <cone-892> ffmpeg.git 03Clément BSsch 07master:80661e0f5586: buffer: use the atomic get instead of the add and fetch variant.
[12:22] <cone-892> ffmpeg.git 03Anton Khirnov 07master:15ec0450b4ae: lavc: allow decoders to override frame parameters.
[12:22] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:64d11cb615a9: mss1: set frame defaults
[12:22] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:36685c3c4b03: Merge commit '15ec0450b4ae891f3e6ababa03c777a4443b94ca'
[12:26] <ubitux> seems there is someone in panic on mplayer-user because of missing vf uspp filter (he made a mistake, it doesn't look dropped)
[12:26] <ubitux> he seems to be talking about snow as well
[12:26] <ubitux> so there are some snow users ;)
[12:26] <ubitux> (and pp+)
[12:28] <cone-892> ffmpeg.git 03Anton Khirnov 07master:192f1984b1a9: lavc: limit maximum number of channels to 63
[12:28] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:a7f9d92a6d54: Merge commit '192f1984b1a93aa08af053b8f9ab4950f307bd5d'
[12:55] <cone-892> ffmpeg.git 03Anton Khirnov 07master:2eba9087f303: lavc: make up a fake frame channel layout when there is no real one.
[12:55] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:f9fd6f983b97: Merge commit '2eba9087f3031c6050f8dcd996225490be6c2410'
[12:55] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:a303dce28c9b: utils: Dont sent fake channel layouts, 0 layout is better when its not known
[13:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:37045e422903: mpegvideo: drop vismv code
[13:03] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:3c1326411d6f: Merge commit '37045e422903695e610cca6ecb753df643ab9380'
[13:28] <Bor0> hi, regarding the FFmpeg LGPL license. is ffplay.c also LGPL?
[13:28] <nevcairiel> yes it is
[13:28] <Bor0> I'm intending on releasing libav* as dynamic libraries, however I have converted ffplay.c to libffplay.c. is it okay releasing it as libffplay dynamic library without releasing the source code?
[13:29] <av500> no
[13:29] <av500> why would you not release the source code?
[13:30] <Bor0> it is for a commercial project at work. I guess I'll have to talk to my boss about that
[13:31] <nevcairiel> LGPL still requires you to publish any changes you make
[13:31] <Bor0> I see, alright. thank you
[14:26] <cone-892> ffmpeg.git 03Anton Khirnov 07master:9b2dc29534b3: avconv: convert to new refcounted AVFrame API
[14:26] <cone-892> ffmpeg.git 03Anton Khirnov 07master:354468fc12dc: avplay: switch to new refcounted frames API
[14:26] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:32fdfdfbdac6: Merge commit '354468fc12dcf93e7fb4af552e18906771913273'
[14:33] <cone-892> ffmpeg.git 03Anton Khirnov 07master:e7bbfc4242ae: avserver: don't set deprecated options.
[14:33] <cone-892> ffmpeg.git 03Anton Khirnov 07master:ddcca4ba074a: lavc: stop setting AVFrame.motion_subsample_log2
[14:33] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:dac87995295e: Merge commit 'ddcca4ba074aa537c3059c76d69eb4ad6316bfe9'
[14:54] <cone-892> ffmpeg.git 03Anton Khirnov 07master:1296b1f6c063: AVFrame: deprecate all now unused fields
[14:54] <cone-892> ffmpeg.git 03Anton Khirnov 07master:8df23e938b40: lavc: postpone the removal of request_channels API.
[14:54] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:b2cd0d709abd: Merge commit '8df23e938b4022d6e6e9e1180ea6418abae74fda'
[15:30] <BBB> rebase finally done - patches on ML or https://github.com/rbultje/ffmpeg/commits/hpeldsp
[15:30] <BBB> again, this allows chrome to drop mpegvideo and dsputil, which would be huge in terms of binary size
[15:31] <BBB> (there's more stuff that can be dropped, but this was the most complex part so I wanted to start with this)
[15:41] <cone-892> ffmpeg.git 03Anton Khirnov 07master:d8b31be6caeb: Add the bumps and APIchanges entries for reference counted buffers changes.
[15:41] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:ad78b9ed7009: avcodec: postpone removial of parts still used
[15:41] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:9cf788eca8ba: avformat/avdevice: add missing time.h includes
[15:41] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:c292e340ed08: avformat: keep r_frame_rate
[15:41] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:a75f01d7e0bf: Merge commit 'd8b31be6caebd8d1321ecb754b6e7daaf81dc111'
[15:49] <cone-892> ffmpeg.git 03Anton Khirnov 07master:542b83fc90c1: Replace some forgotten instances of PIX_FMT_* with AV_PIX_FMT_*.
[15:49] <cone-892> ffmpeg.git 03Anton Khirnov 07master:0651e892e16d: Replace remaining includes of audioconvert.h with channel_layout.h
[15:49] <cone-892> ffmpeg.git 03Anton Khirnov 07master:7b89cd20d844: eamad: allocate a dummy reference frame when the real one is missing
[15:49] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:0cfbfaaa12e1: Merge commit '7b89cd20d844cbe763ca34e63e99d110043cf241'
[15:55] <nevcairiel> michaelni: does it really make sense to keep the old codec id names? its a rather big mess many people would be happy to get rid of
[15:57] <cone-892> ffmpeg.git 03Anton Khirnov 07master:d243896987b8: ffv1: fix calculating slice dimensions for version 2
[15:57] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:f506ae19ffb5: Merge commit 'd243896987b8b2062d1faba4d8d6f0c62d2dbee9'
[16:02] <michaelni> nevcairiel, ive no real oppinion on that, i just had to fix compilation, and keeping them likely breaks fewer apps
[16:02] <michaelni> if you want to remove them ive no objections
[16:03] <nevcairiel> i think nicolas just send a patch to fix compilation
[16:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:bde48aa92d33: FATE: enable multiple slices in the ffv1 vsynth test
[16:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:6c7d339afcd1: tty: set avg_frame_rate.
[16:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:8a49d2bcbe75: xxan: fix invalid memory access in xan_decode_frame_type0()
[16:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:3e2f200237af: roqvideodec: fix a potential infinite loop in roqvideo_decode_frame().
[16:03] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:b36095ea0f66: Merge commit '3e2f200237af977b9253b0aff121eee27bcedb44'
[16:05] <nevcairiel> the problem with the old codec id compat is the evil #define for CodecID, which is a generic name which can easily affect other apps too, and has in the past, getting rid of it would make this problem go away, even if some applications have to add a few AV_'s 
[16:06] <nevcairiel> I'll send a patch for other peoples opinions on the ML then
[16:19] <Daemon404> humm
[16:19] <Daemon404> is there a way to figure out the transport packet size?
[16:19] <Daemon404> i.e. 188 or 192
[16:19] Action: Daemon404 thinks nevcairiel may know
[16:19] <nevcairiel> you mean from the mpegts demuxer?
[16:19] <nevcairiel> or from a bitstream?
[16:20] <Daemon404> demuxer
[16:20] <Daemon404> ideally from ffprobe...
[16:20] <nevcairiel> its not exposed
[16:20] <Daemon404> meh
[16:20] <cone-892> ffmpeg.git 03Anton Khirnov 07master:3ba40ebb6cc5: ivi_common: do not call MC for intra frames when dc_transform is unset
[16:20] <nevcairiel> because its hardly useful =p
[16:20] <cone-892> ffmpeg.git 03Anton Khirnov 07master:ea1136baafb1: wmadec: require block_align to be set.
[16:20] <cone-892> ffmpeg.git 03Anton Khirnov 07master:cacad1c058f6: wmaprodec: require block_align to be set.
[16:20] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:6ec0447d3d22: xface: remove unused stuff, fix related warnings
[16:20] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:7787a6c7caa0: Merge commit 'cacad1c058f66558ec727faac3b277d2dee264d4'
[16:20] <Daemon404> nevcairiel, i know
[16:22] <Daemon404> k
[16:23] <cone-892> ffmpeg.git 03Nicolas George 07master:b52421940f5c: lavf: add AV_ to two remaining CODEC_ID_*.
[16:35] <cone-892> ffmpeg.git 03Anton Khirnov 07master:4c0080b7e7d5: wmaprodec: return an error, not 0, when the input is too small.
[16:35] <cone-892> ffmpeg.git 03Anton Khirnov 07master:f86d66bcfa48: vmdaudio: fix invalid reads when packet size is not a multiple of chunk size
[16:35] <cone-892> ffmpeg.git 03Yusuke Nakamura 07master:19dd4017ab6d: libopencore-amr: Add the missing 3rd argument of ff_get_buffer()
[16:35] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:bf780b7ea8d9: Merge commit '19dd4017ab6dac11c77d797acebee4f60ad63a6f'
[16:41] <cone-892> ffmpeg.git 03Anton Khirnov 07master:7c45087b8441: lavc: update the fallback versions of ff_thread_*
[16:41] <cone-892> ffmpeg.git 03Martin Storsjö 07master:2240e92f0529: atomic: Add include guards to the implementation headers
[16:41] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:7093b7534fd3: Merge commit '2240e92f052960693de55cf4924e80426dfaa965'
[16:51] <Daemon404> is there documentation of the "reference" values anywhere?
[16:51] <Daemon404> in AVFrame
[16:52] <Daemon404> or is that field even valid anymore...
[16:52] <nevcairiel> its not
[16:52] <Daemon404> im trying to figure out in a GOP is open or not
[16:52] <Daemon404> based on ffprobe output
[16:53] <Daemon404> i guess its impossible?
[16:53] <Daemon404> (-show_frames or -show_packets)
[16:55] <Daemon404> so... :/
[17:28] <ubitux> Daemon404: there was some documentation
[17:28] <ubitux> it got removed while moving it
[17:28] <ubitux> let me find it again...
[17:29] <Daemon404> doesnt matter
[17:29] <Daemon404> not what i wanted
[17:29] <ubitux> 7ecc2d403ce5c7b6ea3b1f368dccefd105209c7e
[17:29] <ubitux> in case you're still interested
[17:29] <Daemon404> turns out what i wanted isnt exported at all anyway
[17:29] <Daemon404> janne is looking into adding it
[17:35] <cone-892> ffmpeg.git 03Giorgio Vazzana 07master:ee4a6586952b: lavd/v4l2: fix bug in init_convert_timestamp()
[18:12] <burek> -strict is missing from online docs
[18:14] <michaelni> doc/examples/filtering_audio.c needs an update i think
[18:17] <cone-892> ffmpeg.git 03Martin Storsjö 07master:f1af3d19a792: output-example: Update to use encode_video2 instead of the now dropped encode_video
[18:17] <cone-892> ffmpeg.git 03Martin Storsjö 07master:e460aa328296: atomic: Check for __sync_val_compare_and_swap instead of __sync_synchronize
[18:17] <cone-892> ffmpeg.git 03Anton Khirnov 07master:9d3009c6c4b9: avconv: print an error on applying options of the wrong type.
[18:17] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:6b6b0e9daed3: pthread: fix unused variable warning
[18:17] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:5743095ca956: Merge commit '9d3009c6c4b9b6734f07df7c88f6a42ded6cdf38'
[18:31] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:bd93f3c5bdb8: ffmpeg: fix reinit_filter type
[18:31] <cone-892> ffmpeg.git 03Hendrik Leppkes 07master:f91730a16b40: jv: initialize AVFrame
[18:57] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:dc7ba0c1f4e9: j2kdec: remove unused variable and empty function
[18:57] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:19688e87e5f8: h264: show recovery SEIs in debug output
[19:02] <burek> do we guys have a twitter/fb page?
[19:02] <burek> where we can announce new stuff
[19:02] <burek> that's being developed?
[19:03] <michaelni> burek, we have a g+ page at least
[19:04] <burek> that's good too
[19:09] <llogan> burek: do you think we should have a twitter page?
[19:10] <burek> I think it would be cool
[19:10] <burek> it's not difficult to maintain
[19:10] <burek> just you have to agree on access rights, who will announce things, to avoid problems :)
[19:13] <llogan> FFmpeg user name is already taken
[19:14] <llogan> oh, it's that guy: https://twitter.com/FFmpeg
[19:14] <llogan> he shouldn't be using that name
[19:14] <burek> you can contact him or if he doesn't have any understanding about the issue
[19:15] <burek> then we can contact fb/twitter directly
[19:15] <llogan> i'll contact him. i think he'll understand
[19:15] <burek> they usually prefer well known organizations using their native usernames over some random users doing so
[19:16] Action: llogan has never used twitter or facebook
[19:23] <burek> :)
[19:55] <michaelni> burek, about "strict" or anything else missing, ask saste
[19:55] <burek> ok, i will
[19:55] <burek> thx
[19:56] <michaelni> IIRC he disabled the inclusion of the auto generated AVOption lists
[19:56] <michaelni> or its broken dunno
[20:08] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:b1a9a2c4ffb8: ffmpeg: remove unused variable
[20:08] <cone-892> ffmpeg.git 03Ronald S. Bultje 07master:6a701306db16: dsputil: make selectable.
[20:12] <cone-892> ffmpeg.git 03Hendrik Leppkes 07master:713ac2109755: avutil/os_support: use 64-bit fstat/lseek variants for MSVC as well
[20:20] <burek> if one wants a precise -maxrate not to exceed in any circumstances, not even a bit over it, how can he/she specify it with ffmpeg? is it enough to just type: -b Xk -maxrate Yk -bufsize Xk
[20:21] <relaxed> isn't that the whole point -maxrate?
[20:21] <burek> well, it goes over even when some Y value has been specified, that's why I'm asking
[20:23] <michaelni> burek, if someone says max rate is ignored then 99% of the times the person has no clue what VBV, maxrate and bufsize mean, the remaining 1% are actual bugs
[20:24] <burek> I agree, I've read a lot of posts on the internet on this issue, and I would like to write a wiki page that explains it a little bit with examples
[20:24] <burek> so that we have a kind of "official" instructions how to properly use those options
[20:24] <burek> because it's a common problem in streaming
[20:28] <kierank> meh, streaming doesn't care much for vbv
[20:29] <burek> well, i for example like using -crf, but i also like to have some -maxrate, just in case it goes crazy :)
[20:30] <relaxed> It should work fine with -crf
[20:40] <burek> is there any good article on vbv and maxrate, that i could use as a starting point for the wiki article?
[20:48] <michaelni> burek, the mpeg spec ;)
[20:57] <cone-892> ffmpeg.git 03Hendrik Leppkes 07master:cde1e7db3e67: vorbis: unref the skipped first frame.
[21:03] <cone-892> ffmpeg.git 03Anton Khirnov 07master:1da54e908eba: avconv: use a local variable for OutputFile in open_output_file().
[21:03] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:49e87819d09e: Merge commit '1da54e908eba73eb86f5d614c51f06811c85c3ea'
[21:15] <cone-892> ffmpeg.git 03Anton Khirnov 07master:41d2008b15c2: avconv: use a local variable for InputFile in open_input_file().
[21:15] <cone-892> ffmpeg.git 03Anton Khirnov 07master:6deae8e0530a: FATE: remove the -dct option from some decoding tests.
[21:15] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:771919e11fcd: Merge commit '6deae8e0530a885b76f149841ed4899cb438ec23'
[21:16] <burek> michaelni right :) do you have any url perhaps? ^^
[21:16] <burek> or should i ask google? :)
[21:17] <michaelni> google 
[21:21] <cone-892> ffmpeg.git 03Anton Khirnov 07master:64bfc3e20588: FATE: replace -vb with -b:v
[21:21] <cone-892> ffmpeg.git 03Anton Khirnov 07master:db2d65ce1ed1: avconv: fix a typo
[21:21] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:c51fcdf0f62e: Merge commit 'db2d65ce1ed13ea9a530b1934963a16c3e8fa460'
[21:39] <cone-892> ffmpeg.git 03Anton Khirnov 07master:e7553f4c7817: avconv: do not silently ignore unused codec AVOptions.
[21:39] <cone-892> ffmpeg.git 03Martin Storsjö 07master:666fe5da47d1: atomic: Exclude the unsupported implementation headers from checkheaders
[21:39] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:212b89f8b426: avfilter: remove unused variable
[21:39] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:d64b8540751b: Merge commit '666fe5da47d127074be7f0e2bac93db6af8b4a30'
[21:46] <cone-892> ffmpeg.git 03Justin Ruggles 07master:e984f4787325: libmp3lame: use the correct remaining buffer size when flushing
[21:46] <cone-892> ffmpeg.git 03Justin Ruggles 07master:486f0b0cfc80: png: use av_mallocz_array() for the zlib zalloc function
[21:46] <cone-892> ffmpeg.git 03Anton Khirnov 07master:6327c1070292: atomic: fix CAS with armcc.
[21:46] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:0d0f76ea5697: Merge commit '6327c10702922eabcb1c6170abd3f03d23ce4c51'
[22:11] <burek> just to mention -b appears to be missing from the online docs too (saste is not online, but will probably read the irc logs, i hope) :)
[22:16] <cone-892> ffmpeg.git 03Anton Khirnov 07master:5d606863c353: lavc: remove disabled FF_API_OLD_DECODE_AUDIO cruft
[22:16] <cone-892> ffmpeg.git 03Anton Khirnov 07master:f073b1500e3b: lavc: remove disabled FF_API_OLD_ENCODE_AUDIO cruft
[22:16] <cone-892> ffmpeg.git 03Anton Khirnov 07master:0f24a3ca999a: lavc: remove disabled FF_API_OLD_ENCODE_VIDEO cruft
[22:16] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:2ec30267fe21: Changelog: fix project name
[22:16] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:eba6a04e1c9c: vp6: remove unused variable
[22:16] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:e052f06531c4: Merge commit '0f24a3ca999a702f83af9307f9f47b6fdeb546a5'
[22:26] <cone-892> ffmpeg.git 03Anton Khirnov 07master:3b0f586f07bc: lavc: remove disabled FF_API_MPV_GLOBAL_OPTS cruft
[22:26] <cone-892> ffmpeg.git 03Anton Khirnov 07master:9d6cf81f02d3: lavc: remove disabled FF_API_COLOR_TABLE_ID cruft
[22:26] <cone-892> ffmpeg.git 03Anton Khirnov 07master:2a113695238e: lavc: remove disabled FF_API_INTER_THRESHOLD cruft
[22:26] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:5beddbc4cf81: Merge commit '2a113695238e7675ab75c5fb9a5f59fc92f2ddd9'
[22:35] <burek> can i suggest that you guys add at least a hint if people encode yuv444p (or any non yuv420p) with x264, to just hint the user that the output might not work properly for some players or something similar
[22:35] <burek> i suggested on the tracker to change the default output to yuv420p but due to back compatibility it wasnt accepted
[22:36] <nevcairiel> changing the default to downconvert to 420 by default would be very bad
[22:37] <burek> yes, I realized that later, but at least a hint wouldnt hurt i think
[22:38] <nevcairiel> there are various players with various limitations, you want it to warn about levels above 4.1 as well, and stuff like that? :p imho it opens a can of worms better left untouched
[22:39] <burek> but this worm is really frequent :/
[22:39] <bcoudurier> burek, I agree
[22:40] <nevcairiel> personally i dont see ffmpeg as a tool for the extremely dumb users, they can use a nice clicky frontend which has presets optimized for their silly iDevices
[22:40] <bcoudurier> default when unspecified should be yuv420p for h264
[22:40] <bcoudurier> nevcairiel, dumb != novice
[22:40] <nevcairiel> it should use whatever fits the input the best, everything else is just wrong
[22:40] <burek> nevcairiel, with dumb, you mean people who are newbees? :)
[22:41] <nevcairiel> most people learn best through mistakes :)
[22:42] <burek> trust me, many of them just drop ffmpeg and go for other tools
[22:42] <bcoudurier> nevcairiel, I don't think that's a good idea for h264
[22:42] <burek> they think it just doesnt do the job correctly and they dont bother learning what is wrong
[22:42] <bcoudurier> most people want to do 420 so it should default to that unless requested
[22:43] <iive> why do people encode to yuv444p? because this is the default yuv mode picked up after rgb->yuv ?
[22:43] <nevcairiel> i dont believe in catering to what "most" want, but do what is best
[22:43] <bcoudurier> what is best is what most want/will want
[22:43] <bcoudurier> nobody wants 444 h264 for sure
[22:43] <nevcairiel> if my source is 444, i want h264 to preserve that, unless i tell it not to
[22:44] <llogan> nevcairiel: if you've answered as many yuv420p quesitons as i (or burek) have answered you'd understand how frequently this occurs
[22:44] <bcoudurier> nevcairiel, you are the exception here
[22:45] <burek> well, he's got the point
[22:45] <burek> but if you are re-encoding, what is the point of preserving anything
[22:45] <llogan> i don't know if the default should be yuv420p, but i think an informative message at least would be useful
[22:47] <cone-892> ffmpeg.git 03Anton Khirnov 07master:d03a94e2b752: lavc: remove disabled FF_API_SUB_ID cruft
[22:47] <cone-892> ffmpeg.git 03Anton Khirnov 07master:845017105f24: lavc: remove disabled FF_API_DSP_MASK cruft
[22:47] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:db594f65ec4e: Merge commit '845017105f240ac5348aad1f8a5256b86d49db76'
[22:48] <relaxed> llogan: users don't read, you know this.
[22:48] <nevcairiel> the problem with such a fixed default is that its annoying to handle in automated things. Say i have a list of videos in various color spaces which i all want to encode to h264 as untouched as possible. If it auto-picks the best matching, i have to do nothing, but if it defaults to 420, i have to specify the correct one for every video, which means i have to figure out the correct one for every video in the first place
[22:48] <burek> yes, but it would be the hint to us, who read their pastebins :)
[22:48] <relaxed> anything but yuv420p is a hint to us.
[22:49] <michaelni> just split the libx264 AVCodec, one that lists all pixfmts as supported and one just with 420
[22:49] <llogan> it's easier/more efficient to simply point to the message
[22:50] <llogan> relaxed: ^
[22:51] <burek> honestly I think the hint would avoid this arguing about if it's ok to make it default or not
[22:51] <burek> it would help those who need a clue what's wrong and for those who dont read, it would be the same anyway
[22:51] <llogan> i think the message is the best compromise and probably fairly easy to implement.
[22:54] <llogan> burek: making a feature request with an example message would be a good start
[22:54] <burek> im afraid of cehoyos
[22:54] <burek> ^^
[22:55] <relaxed> man up
[22:56] <llogan> just blame me if he bites
[23:00] <burek> btw, i think this is some kind of a bug http://pastebin.com/Zr3iPR1y (note the "No program" string).. a guy says it loads correctly in vlc and shows all the programs correctly
[23:00] <burek> brb
[23:01] <iive> wasn't preset created to resolve such things?
[23:03] <cone-892> ffmpeg.git 03Ronald S. Bultje 07master:d85c9b036e65: vp3/x86: use full transpose for all IDCTs.
[23:12] <Compn> did anyone do a before / after buffer plan benchmark ?
[23:12] Action: Compn does not volunteer either
[23:24] <BBB-work> michaelni, ty for those two - any idea if you'll have time to review the hpeldsp stuff?
[23:25] <cone-892> ffmpeg.git 03Anton Khirnov 07master:7d42fd6bb9df: lavc: remove disabled FF_API_FIND_BEST_PIX_FMT cruft
[23:25] <cone-892> ffmpeg.git 03Michael Niedermayer 07master:f7b041b11b0f: Merge commit '7d42fd6bb9df99a20cf2d0d0e3b166a83a649d08'
[23:29] <michaelni> BBB-work, ill try to do that ASAP
[23:32] <BBB-work> lol is that my commit?
[23:34] <BBB-work> hm no it's something called 'jv'
[23:35] <ubitux> note: i'll apply "lavc: factorize ff_{thread_,re,}get_buffer error messages." in two days if i see no comment
[23:35] <ubitux> also note i've added some forgotten entries; if anyone is interested into a new version of the patch, please raise a hand
[23:41] <michaelni> btw, if someone else wants to help review the hpel patches, that IS welcome
[23:45] <michaelni> BBB-work, do you have some git branch that i can checlout/merge for testing the whole patchset ?
[23:46] <BBB-work> michaelni, https://github.com/rbultje/ffmpeg/commits/hpeldsp
[23:47] <cone-892> ffmpeg.git 03Nicolas George 07master:9a2688826dd4: lavfi/buffersink: rename filter structures.
[23:47] <cone-892> ffmpeg.git 03Nicolas George 07master:ceac5c54dd38: Remove references to the "ff" variant of buffersink.
[23:47] <cone-892> ffmpeg.git 03Nicolas George 07master:c14a8c834c0f: lavfi/buffersink: schedule removing the "ff" variants.
[23:47] <cone-892> ffmpeg.git 03Nicolas George 07master:684a264638b6: configure: remove selection of ff(a)buffersink.
[00:00] --- Wed Mar 13 2013


More information about the Ffmpeg-devel-irc mailing list