[Ffmpeg-devel-irc] ffmpeg-devel.log.20140718

[burek]

[00:08] <jamrial> and dsputil is no more
[00:11] <michaelni> and the ac3 encoder now needs the motion estimation context hooray
[00:13] <kierank> why
[00:14] <kierank> oh sad
[00:20] <jamrial> and alpha is broken again
[00:20] <jamrial> meh, super easy to fix. i'll send a patch later tonight if nobody fixes it by then
[00:23] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:162cffca802d: avcodec/me_cmp: restore author attribution and copyrights
[00:46] <wm4> is there code anywhere in ffmpeg that rewrites timestamps of a packet stream with timestamp resets to continuous timestamps? I suppose at least ffmpeg.c with remuxing must have something like this
[00:58] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:c527c14d6323: avcodec/dvenc: clear all dsp contexts
[01:00] <michaelni> wm4, yes there is discontinuity removial code in ffmpeg.c
[01:01] <wm4> aaaaand I'm spotting ffmpeg.c using libavformat internals
[01:01] <wm4> it accesses pts_wrap_bits, which is not part of the public api
[01:02] <j-b> and once again proves my point that ffmpeg should be splitted from the libav* repo
[01:02] <wm4> yeah, that would be a good idea
[01:02] <wm4> also, remove all non-public fields from the public header files
[01:03] <j-b> eat your own dog food
[01:06] <wm4> from what I can see, ffmpeg.c considers the timestamp differences between all streams to guess whether there is a reset
[01:06] <wm4> doesn't this fail hard if the stream is not perfectly interleaved?
[01:25] <Timothy_Gu> saste: someone donated $1000 in April?
[01:25] <Timothy_Gu> sweet!
[01:29] <wm4> (and the default is 10 seconds timestamp difference, now that's approximate...)
[01:30] <wm4> it would all be so simple if libdvdnav and libbluray wouldn't stink to hell
[02:31] <michaelni> wm4, the code should work fine if its crazily mis-interleaved
[02:42] <cone-149> ffmpeg.git 03Ben Avison 07master:5c22e8e4ad08: armv6: Accelerate ff_imdct_half for general case (mdct_bits != 6)
[02:42] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:4c735ee0e3d4: Merge commit '5c22e8e4ad0852d61d5c4ba8d67d33fd72339497'
[03:12] <cone-149> ffmpeg.git 03Ben Avison 07master:87552d54d333: armv6: Accelerate ff_fft_calc for general case (nbits != 4)
[03:12] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:c27adb37ef5b: Merge commit '87552d54d3337c3241e8a9e1a05df16eaa821496'
[03:43] <cone-149> ffmpeg.git 03Ben Avison 07master:5edad2c4a1f4: armv6: Accelerate vector_fmul_window
[03:44] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:7809c2600612: Merge commit '5edad2c4a1f46bcc56be755af86ab355c2f1b37f'
[03:50] <cone-149> ffmpeg.git 03Ben Avison 07master:5a272190a046: armv6: Accelerate butterflies_float
[03:50] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:c67cf84628bc: Merge commit '5a272190a04666f0fe41be767396b30712638c21'
[04:53] <cone-149> ffmpeg.git 03Timothy Gu 07master:b64808e9ddc8: kerndeint: remove dead assignments
[05:14] <cone-149> ffmpeg.git 03Timothy Gu 07master:cd3c85ac5cb2: a64multienc: remove dead assignment
[05:14] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:d5ba19e611b0: avcodec/a64multienc use av_freep()
[05:42] <cone-149> ffmpeg.git 03James Almer 07master:95a2d705cd98: alpha/idctdsp: move disabled cruft out of dsputil
[05:42] <cone-149> ffmpeg.git 03James Almer 07master:277e5ca6175e: alpha/me_cmp: move code out of dsputil
[11:28] <ubitux> ePirat: speaking of this, no other way of pushing to icecast?
[11:28] <ubitux> like, icecast isn't supporting some http pushing?
[11:28] <ePirat> http PUT is what should be used
[11:28] <ePirat> at least with 2.4+
[11:28] <ubitux> and it doesn't work?
[11:29] <ePirat> it works but ffmpeg can't do it
[11:29] <ubitux> huh
[11:29] <ePirat> it only does POST
[11:29] <ubitux> ok
[11:30] <ubitux> iirc i was able with chunked_post to push to some kind of icecast-like
[11:30] <ubitux> but it wasn't icecast itself
[11:32] <ubitux> it's kind of surprising if it doesn't support this
[12:53] <cone-719> ffmpeg.git 03Martin Storsjö 07master:fe67f3fbb5f9: arm: cosmetics: Fix a misaligned asm operand
[12:53] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:cd4497d8c5c4: Merge commit 'fe67f3fbb5f9f6a6b60f837f6bc5e087ac11f3bf'
[13:31] <cone-719> ffmpeg.git 03Martin Storsjö 07master:7e18a727d2c2: arm: cosmetics: Consistently use lowercase for shift operators
[13:31] <cone-719> ffmpeg.git 03Michael Niedermayer 07master:d13effb0b4d8: Merge commit '7e18a727d2c2a19f22fcf68875d1b05fd2eafcef'
[13:35] <cone-719> ffmpeg.git 03Timothy Gu 07master:0035783488cc: smacker: remove dead code
[14:24] <cone-719> ffmpeg.git 03Andrey Utkin 07master:cb8cb85da791: lavfi/drawtext: add alias "expr_int_format" to expansion function "eif"
[16:13] <berzerka> hi there
[16:14] <berzerka> i was looking at the git log to see when the avcodec_alloc_frame deprecation took place
[16:14] <berzerka> the commit was: 2013-12-11 - 29c83d2 / b9fb59d,409a143 / 9431356,44967ab / d7b3ee9 - lavc 55.45.101 / 55.28.1 - avcodec.h
[16:14] <berzerka> what do the two version numbers tell me? in which version was the change introduced, former, latter, both?
[16:16] <berzerka> and how would i properly compile-time test for this in another library?
[16:22] <Daemon404> one is the libav # and one is the ffmpeg #
[16:22] <Daemon404> the .101 is the ffmpeg one and the version you'd check for ffmpeg
[16:25] <berzerka> ok so i would check against >=libavcodec-55.28
[16:26] <berzerka> what's the proper #if macro to do so?
[16:27] <berzerka> and how do i interpret the commits where there's only one version shown? does it mean the commit does not affect ffmpeg, yet?
[16:29] <berzerka> oh looks like this project is including ffmpeg and not libavcodec after all
[16:29] <berzerka> so how would i check against >=ffmpeg-55.45.101?
[16:31] <wm4> ffmpeg versions start from x.y.100
[16:31] <wm4> while libav starts from x.y.0
[16:32] <wm4> it's stupid, but this is how it is
[16:32] <berzerka> ok so i would only check against >=55.45
[16:33] <wm4> no, ffmpeg and libav version numbers are not aligned
[16:33] <wm4> so you have to check for them separately
[16:34] <berzerka> the project i'm dealing with with only uses the libavcodec version which comes which ffmpeg, so i think i only need to check for the ffmpeg version?
[16:34] <berzerka> which comes with*
[16:35] <berzerka> otoh, avcodec.h is part of libavcodec, so i would need to check only for that one? i'm confused.
[16:37] <wm4> that should be sufficient
[16:38] <berzerka> ok thanks! still, how do i interpret the commits which only show one version number. e.g.: 2013-07-23 - abc8110 - lavc 55.19.100 - avcodec.h
[16:39] <berzerka> from the version number i guess the commit only affects only ffmpeg. however, avcodec.h was changeg, which is clearly part of libavcodec?
[16:39] <berzerka> my god s/only//
[16:39] <berzerka> changed*
[17:15] <berzerka> whatever, have a nice weekend everyone!
[18:34] <jamrial> libav is planning a release in a couple weeks. what are we going to do?
[18:34] <jamrial> rush an ffmpeg 2.4/3.0 out of the door, break protocol and add non-abi-breaking features to 2.3.1 so it's in sync with libav11, or do nothing at all?
[18:41] <ubitux> a couple of weeks = probably about 3 months, that will probably match 2.4 schedule
[18:41] <jamrial> lol
[18:54] <nevcairiel> Yeah they aren't exactly fast at releasing stuff
[19:05] <jamrial> well, lu did mention he was targeting august, which is why i asked
[19:06] <cone-444> ffmpeg.git 03Carl Eugen Hoyos 07master:5dcd43cd8666: Fix make checkheaders for libavcodec/proresdsp.h.
[19:06] <nevcairiel> They will have a beta again and it'll all take months
[19:08] <jamrial> Ah true
[19:19] <ubitux> jamrial: he also said "but I'd start the first working prototype of avscale this week" in march
[19:19] <Daemon404> lol avscale
[19:19] <Daemon404> lol.
[19:20] <kierank> good luck with that
[19:20] <iive> ubitux: well, he said "start" not "finish"
[19:46] <cone-444> ffmpeg.git 03Peter B 07master:0f7013108998: fate: Add test for wav Peak Envelope Chunk encoder (levl chunk).
[21:29] <michaelni> mraulet, smarter, is "0718 17:41 Christophe Gisq (3.4K) [FFmpeg-devel] [PATCH] [RFC] hevc: report more precise progress" ok
[21:29] <michaelni> ?
[21:33] <mraulet> I am looking into it right now
[21:36] <cone-444> ffmpeg.git 03Diego Biurrun 07master:cb44b21da1f5: dct-test: Move cpu_flags variable out of global scope
[21:36] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:2dbda9378c1e: Merge commit 'cb44b21da1f59923be577f08c267ec270529be97'
[21:43] <cone-444> ffmpeg.git 03Diego Biurrun 07master:746ad4e0df7f: dct-test: Improve CPU flags struct member name
[21:43] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:29119aff6cc7: Merge commit '746ad4e0df7faf93329804e412ec53c1d929a75b'
[21:48] <cone-444> ffmpeg.git 03Diego Biurrun 07master:24af1aa0f703: fft: Convert FFT/MDCT permutation type #defines to enums
[21:48] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:cff61d595e45: Merge commit '24af1aa0f70362a66cda04c9d7cd012e019f5572'
[21:53] <cone-444> ffmpeg.git 03Diego Biurrun 07master:c0de9159a7ba: avdevice: Give names to anonymously typedeffed structs
[21:53] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:f6ed5df9c00c: Merge commit 'c0de9159a7ba5707aa0a5c2bc73ae78b7b87ec46'
[22:14] <cone-444> ffmpeg.git 03Diego Biurrun 07master:b4987f72197e: idct: Convert IDCT permutation #defines to an enum
[22:14] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:42d326353c17: Merge commit 'b4987f72197e0c62cf2633bf835a9c32d2a445ae'
[22:16] <mraulet> michaelni: patch partially dropped after a discussion with the author
[22:17] <michaelni> mraulet, ok, if theres something i should apply/cherry pick ...
[22:18] <mraulet> yes the sao stuff
[22:18] <mraulet> 10fc29fc19a12c4d8168fbe1a954b76386db12d0
[22:24] <cone-444> ffmpeg.git 03Diego Biurrun 07master:8b0dd4942aac: idctdsp: prettyprinting cosmetics
[22:24] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:521f5697348d: Merge commit '8b0dd4942aac320d1ca3c40fa7ea1be342c71273'
[22:25] <Timothy_Gu> michaelni: when cherry-picking please put the "(cherry-picked from" a separate line
[22:26] <mraulet> michaelni: bitstreams for fate-hevc
[22:26] <mraulet> https://dl.dropboxusercontent.com/u/588600/rext_biststreams.zip
[22:26] <mraulet> for rext
[22:26] <michaelni> Timothy_Gu, ok will do
[22:27] <mraulet> let me know where they are in I will update the tests
[22:28] <kurosu> 10fc29fc19a12c4d8168fbe1a954b76386db12d0 <- related to sao, but not to the patch I submited
[22:28] <mraulet> yes
[22:32] <cone-444> ffmpeg.git 03Nidhi Makhijani 07master:b4edbe9a0eb4: libgsm: Split decoder and encoder
[22:32] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:4b0fa45b842c: Merge commit 'b4edbe9a0eb48423fbf249b0ca7fb62838bdb7f5'
[22:39] <cone-444> ffmpeg.git 03Diego Biurrun 07master:9e0b29911f1f: x86: dnxhdenc: Eliminate some unnecessary ifdefs
[22:39] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:097bf834bacd: Merge commit '9e0b29911f1f167381a7dbdfca68bf417b8c767b'
[22:44] <michaelni> mraulet, rext uploaded 
[22:44] <mraulet> ok
[22:45] <mraulet> I have them
[22:45] <cone-444> ffmpeg.git 03Diego Biurrun 07master:913fa85a2c1f: dct-test: Skip indirection for MMX IDCT permutation
[22:45] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:2897447ddf80: Merge commit '913fa85a2c1fcd71802f7decdef5f355e0d9c757'
[23:04] <cone-444> ffmpeg.git 03Mickaël Raulet 07master:d249e6828e8a: hevc/sao: optimze sao implementation
[23:04] <cone-444> ffmpeg.git 03Michael Niedermayer 07master:ffa0b510cd15: avcodec/hevc: fix pict_type
[23:04] <mraulet> michaelni: for rext: with fate I will have to put pixfmt for each bitstream?
[23:05] <michaelni> fate must pass on all archs, big endian and little, not sure but i would guess that means choosing one endianness and forcing it
[23:06] <mraulet> why do we need it for 
[23:06] <mraulet> ok
[23:06] <mraulet> forget my last comment
[23:06] <mraulet> a bit painfull for rext then
[23:40] <mraulet> michaelni: tests for hevc rext https://github.com/OpenHEVC/FFmpeg/compare/10fc29fc19...e6ee89e6fa
[23:40] <cone-444> ffmpeg.git 03Carl Eugen Hoyos 07master:0e70266bbfad: Write valid free atom when -moov_size is specified.
[23:40] <cone-444> ffmpeg.git 03Carl Eugen Hoyos 07master:ee88a2080ef0: Parse dri when receiving jpg via rtp.
[00:00] --- Sat Jul 19 2014