[Ffmpeg-devel-irc] ffmpeg-devel.log.20150821

burek burek021 at gmail.com
Sat Aug 22 02:05:02 CEST 2015


[00:59:59 CEST] <cone-497> ffmpeg 03Paul B Mahol 07master:4e8963fa385b: avfilter: add video vectorscope filter
[02:04:39 CEST] <DrBenway> hi, is there a public windows build that includes x264?
[02:05:31 CEST] <klaxa> hi DrBenway, wrong channel, but yes: http://ffmpeg.zeranoe.com/builds/
[02:05:48 CEST] <klaxa> oh i see you asked in the right channel
[02:06:19 CEST] <DrBenway> hmmm i thought this build idnd't include x264
[02:06:46 CEST] <klaxa> let's move this to #ffmpeg
[02:07:03 CEST] <DrBenway> ok
[02:09:54 CEST] <klaxa> args, i've written too many programs in languages with automatic memory management, should i av_free() every string/variable i allocate with av_strdup()/av_opt_get() ?
[02:12:11 CEST] <klaxa> i'm thinking "yes"
[02:15:56 CEST] <DrBenway> strdup normally assumes you will delete the memory yourself
[02:17:29 CEST] <DrBenway> (I'm new to ffmpeg and i had a lot of questions about "should i free this piece of memory". I cleared all my leaks using clang asan
[02:17:35 CEST] <DrBenway> worked great
[02:17:56 CEST] <DrBenway> except that now it crashes on android while doing an av_free :(
[02:17:58 CEST] <klaxa> i just ran my code with valgrind, it told me where i'm leaking memory
[02:19:39 CEST] <DrBenway> im glad that worked out for you
[02:19:47 CEST] <DrBenway> (i've had bad experiences with valgrind)
[02:20:01 CEST] <klaxa> i'm probably better off copying the data into stacked strings
[02:47:02 CEST] <rcombs> klaxa: it sometimes depends what you do with the result
[02:47:38 CEST] <rcombs> klaxa: in some cases, allocated strings will later be freed by the libraries, if they were used as arguments to certain functions
[02:47:47 CEST] <rcombs> e.g. some things around AVOptions
[02:48:01 CEST] <klaxa> ok
[02:48:17 CEST] <rcombs> but in general, you should free what you allocate
[02:48:38 CEST] <klaxa> the strings extract are just av_opt_get() and are never passed anywhere and only used within the function
[02:48:54 CEST] <klaxa> so i probably want to free them
[02:49:26 CEST] <rcombs> the docs for that function are quite clear: https://www.ffmpeg.org/doxygen/trunk/group__opt__get__funcs.html#gaf31144e60f9ce89dbe8cbea57a0b232c
[02:50:01 CEST] <klaxa> regarding that, i can't seem to find what i'm doing wrong: i have a char* variable = NULL, do av_opt_get(...), but when i call av_free(variable) i get invalid free()s
[02:50:10 CEST] <klaxa> i'll post code, one sec
[02:53:03 CEST] <klaxa> https://gist.github.com/klaxa/dd33e5c5d67d24f8374e i get errors in av_free(resource)
[02:53:25 CEST] <klaxa> oh wait... i think i found out what i did wrong
[02:53:29 CEST] <klaxa> nevermind
[02:53:41 CEST] <klaxa> i'm doing resource++ obviously that messes with av_free()
[02:57:32 CEST] <rcombs> yeah, that'd do it
[02:58:19 CEST] <klaxa> i should still be able to declare such a pointer as const, right?
[02:59:02 CEST] <klaxa> doesn't crash, so probably yes :)
[03:02:09 CEST] <rcombs> incorrect operations around constness generally result in compiler warnings, not crashes
[03:02:36 CEST] <klaxa> should these be defined as const though?
[03:03:48 CEST] <klaxa> and indeed, i get compiler warnings
[03:38:54 CEST] <cone-497> ffmpeg 03Michael Niedermayer 07master:15ff3f3fdfc7: ffmpeg: check avpicture_fill() return value
[03:38:55 CEST] <cone-497> ffmpeg 03Michael Niedermayer 07master:c8890941d63d: ffmpeg: Check for RAWVIDEO and do not relay only on AVFMT_RAWPICTURE
[03:38:56 CEST] <cone-497> ffmpeg 03Michael Niedermayer 07master:2bb54b82b509: avcodec/h264_mp4toannexb_bsf: Reorder operations in nal_size check
[03:38:57 CEST] <cone-497> ffmpeg 03Michael Niedermayer 07master:ac0ba6f23369: ffmpeg: Check av_parser_change() for failure
[03:38:58 CEST] <cone-497> ffmpeg 03Michael Niedermayer 07master:45f3d4e63e78: ffmpeg: Use correct codec_id for av_parser_change() check
[03:48:40 CEST] <atomnuker> the aac MIPS encoder is like a patch and it's not possible to distinguish between the two
[03:49:01 CEST] <atomnuker> but the FATE tests for it will keep failing because it's lagging behind the normal encoder
[03:49:30 CEST] <atomnuker> anyone knows if there's a way to create a separate test for it? at least
[03:50:39 CEST] <jamrial> wouldn't it be better to poke the mips maintainer to fix/update the encoder?
[03:53:47 CEST] <atomnuker> the last commit to the MIPS encoder was in 2013, when it was originally added
[03:54:08 CEST] <atomnuker> (there are only 2 more commits to that file, both of them one-liners fixing conflicts)
[03:56:26 CEST] <atomnuker> correction, the last commit was huge in order to fix compilations on mips64 back in Feb 2015
[03:58:04 CEST] <atomnuker> (and was not done by the original guy who wrote it)
[04:06:06 CEST] <jamrial> Nedeljko Babic is listed as mips maintainer, and he's active
[05:01:45 CEST] <klaxa> with the new http-serving capabilities ffmpeg could potentially be a stand-alone hls server, would this be desirable? i just want to know if i'm wasting time thinking about it
[05:12:41 CEST] <jamrial> doubt anyone will be against it. not many here care about ffserver :p
[05:15:21 CEST] <klaxa> i've actually come to like it after working on it for the past weeks
[05:15:42 CEST] <klaxa> some sort of hate-love or so
[06:53:33 CEST] <cone-525> ffmpeg 03Vesselin Bontchev 07master:e32a99264be6: Add support for Audible AA files
[06:59:03 CEST] <Timothy_Gu> michaelni: why didn't you use av_err2str in 15ff3f3fdfc788c0e4e584badd7ec300abfbd716 ?
[07:26:03 CEST] <cone-525> ffmpeg 03James Almer 07master:7a806c68a62c: avcodec/hevcdsp: rename sao_band_filter c functions
[10:07:19 CEST] <cone-525> ffmpeg 03Paul B Mahol 07master:5e1d8637602a: avfilter/vf_vectorscope: support yuv(a)420p and yuv410p as input for common case
[10:15:07 CEST] <durandal_1707>  michaelni: alphablend give bunch of warnings
[11:38:48 CEST] <michaelni> Timothy_Gu, no reason beyond copy and paste
[12:23:50 CEST] <cone-525> ffmpeg 03Michael Niedermayer 07master:b3d2035ec3b8: ffmpeg: use av_err2str()
[12:23:51 CEST] <cone-525> ffmpeg 03Michael Niedermayer 07master:7c72a4bbd3a0: swscale/alphablend: Fix pointer type warnings
[12:57:03 CEST] <kierank> ubitux: hmm the soylent is a bit dusty
[12:58:31 CEST] <nevcairiel> that was my thought exactly
[12:58:41 CEST] <nevcairiel> people recommended to leave it in the fridge over night
[12:58:45 CEST] <nevcairiel> so it can "soak"
[12:58:48 CEST] <nevcairiel> but never tried that
[12:59:06 CEST] <kierank> I wonder if I'll get hungry
[12:59:09 CEST] <kierank> lglinskih: hi
[13:00:42 CEST] <lglinskih> kierank: hi! I hope that my audio decode test will work today)
[13:00:47 CEST] <kierank> ok
[13:04:10 CEST] <Compn> was thinking of buying soylent and then eating it raw...
[13:04:18 CEST] <Compn> mixing with water? bah! 
[13:08:00 CEST] <Compn> vdd dietary requirement (o)ther [soylent]
[13:14:01 CEST] <kierank> I've tried the powder alone as well
[13:17:00 CEST] <cone-525> ffmpeg 03Ganesh Ajjanagadde 07master:36f7a378950b: avcodec/jpeg2000: comment out unused variable
[13:36:10 CEST] <iive> Compn: like this : https://www.youtube.com/watch?v=UXkGtJUP0WE ?
[14:19:56 CEST] <cone-525> ffmpeg 03Michael Niedermayer 07master:5176443b2b2e: avfilter/vf_vectorscope: Fix ;;
[14:23:04 CEST] <BBB> libav never answered my question why they wouldnt use the openhevc optimizations :(
[14:23:33 CEST] <nevcairiel> of course not
[14:24:56 CEST] <BBB> ..
[14:54:17 CEST] <wm4> if I could fix the libavformat seek API, I'd totally add seeking streams independently (trivial for mp4, other formats can it implement by skipping packets, which would also help)
[15:03:33 CEST] <nevcairiel> why would you ever want to seek to different points in different streams
[15:03:39 CEST] <nevcairiel> most formats are interleaved for a reason
[15:04:12 CEST] <wm4> for example properly compensating A/V delay, or mp4 edit lists
[15:30:47 CEST] <cone-525> ffmpeg 03Ganesh Ajjanagadde 07master:5edf8b118d9f: avformat/hls: correct comment for ensure_playlist()
[15:30:55 CEST] <ubitux> wm4: the api supports that already
[15:31:02 CEST] <ubitux> you can specify the stream already
[15:31:19 CEST] <ubitux> kierank: yes :)
[15:31:46 CEST] <wm4> ubitux: but that doesn't tell you what happens with the other streams (hint: nothing that makes sense)
[15:31:57 CEST] <ubitux> :(
[15:32:01 CEST] <ubitux> ok
[16:54:19 CEST] <BBB> nevcairiel: so & since youre doing merges & whats the hevc mc plan?
[16:54:29 CEST] <BBB> nevcairiel: should we test if it is actually faster?
[16:54:33 CEST] <BBB> or just enocare?
[17:01:49 CEST] <nevcairiel> i'll just not merge it, because a straight merge would be extremely messy, anyone that knows the hevc simd stuff and is interested could see which parts would be worth using
[17:09:39 CEST] <nevcairiel> personally I think it might be beneficial to investigate this and see if we can get improvements from that, however I do not necessarily have the expertise to do that on my own
[17:18:16 CEST] <BBB> j-b: for vdd, you will book hotels right?
[17:19:16 CEST] <j-b> BBB: yes.
[17:27:10 CEST] <cone-525> ffmpeg 03Ganesh Ajjanagadde 07master:92b1a0fa9e7b: ffserver: cast PID to int64_t before printing
[17:27:11 CEST] <cone-525> ffmpeg 03Ganesh Ajjanagadde 07master:907373ea9d23: avcodec/x86/v210-init: fix unused variable warning
[17:39:30 CEST] <j45> j-b: have you picked a hotel yet?
[17:40:37 CEST] <j-b> j45: we're on it, but it's a bit complex, because we are quite numerous
[17:42:08 CEST] <j45> i would like to reserve a room for myself at or near the hotel(s) you choose.  so please let me know when it is figured out.
[20:47:03 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:32be264cea54: aacenc: coding style changes
[20:47:04 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:e6c9f3a166ad: aacenc: reset special bands in the main frame encoding function
[20:47:05 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:23e786be61c1: aacenc: populate the sce->ics.swb_offset table pointer
[20:47:06 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:b47a1e5c5f58: aacenc: create and initialize an LTP context
[20:47:07 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:43b378a0d321: aaccoder: move the quantization functions to a separate file
[20:47:08 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:860dbe0275e5: aaccoder_mips: update function definitions
[20:47:09 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:d1ca7142ac93: aaccoder: move the Intensity Stereo implementation out
[20:47:10 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:eab12d072e65: aacenc: do not reject AAC-Main profile
[20:47:11 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:a1c487e9215c: aacenc_tns: implement temporal noise shaping
[20:47:12 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:76b81b10d907: aacenc: implement the complete AAC-Main profile
[20:47:13 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:58cd5386e8b3: MAINTAINERS: add myself as an AAC encoder maintainer
[20:48:56 CEST] <atomnuker> finally
[20:49:07 CEST] <atomnuker> that 400+ comment ticket has its days numbered
[20:49:48 CEST] <JEEB> aren't those completely different patches?
[20:50:20 CEST] <nevcairiel> well it improves quality nevertheless
[20:50:37 CEST] <atomnuker> kinda, the rtz got merged which was the important bit
[20:50:56 CEST] <JEEB> nice
[21:05:16 CEST] <nevcairiel> can we enable IS by default soon? :d
[21:06:20 CEST] <BtbN> I'm looking forward to ffmpeg beating fdk soon
[21:06:28 CEST] <nevcairiel> although for compat reasons i need to encode mpeg-2 aac anyway
[21:10:02 CEST] <nevcairiel> once some of these feautres get enabled by default, i should lobby for a profile configuration to select mpeg-2 aac instead of mpeg-4
[21:10:06 CEST] <nevcairiel> :d
[21:45:11 CEST] <jamrial> atomnuker: /ffmpeg/src/libavcodec/aacenc_tns.c:54:12: error: implicit declaration of function 'ffs' [-Werror=implicit-function-declaration]
[21:45:19 CEST] <jamrial> this is with mingw-w64
[21:48:55 CEST] <jamrial> using __builtin_ffs seems to fix it, but that's a gcc builtin, so if msvc doesn't have ffs...
[21:49:55 CEST] <atomnuker> looking for alternatives
[21:51:29 CEST] <nevcairiel> msvc has _BitScanReverse
[21:51:35 CEST] <nevcairiel> but of course also "special"
[21:52:26 CEST] <nevcairiel> ffs is technically POSIX, so using it with a special macro for msvc might be acceptable
[21:53:17 CEST] <jamrial> yeah, __builtin_ffs for mingw/icc since they both support gcc builtins, then a custom implemtation for msvc
[21:58:18 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:fb0c295cc3d2: aacenc_tns: temporarily disable coefficient compression
[21:58:28 CEST] <atomnuker> this should fix things for now
[21:59:03 CEST] <nevcairiel> dont we maybe even have some avutil function for that
[21:59:43 CEST] <atomnuker> coefficient compression only saves a few bits per frame at most, I'll fix it later
[22:00:40 CEST] <nevcairiel> we have ctz, cant you use that?
[22:01:46 CEST] <atomnuker> I need to get the first two most significant bits
[22:02:14 CEST] <nevcairiel> ffs should always be ctz+1, shouldnt it
[22:03:16 CEST] <nevcairiel> or does ffs count the other way around than i'm thinking
[22:03:29 CEST] <atomnuker> yeah, I could use ctz
[22:03:34 CEST] <atomnuker> is there an av_ctz?
[22:03:42 CEST] <jamrial> ff_ctz
[22:03:58 CEST] <nevcairiel> ff_ctz is in an inline function in libavutil/intmath.h
[22:04:43 CEST] <jamrial> you could add one for ffs using the builtin and ctz+1 as well
[22:11:59 CEST] <jamrial> also, as Andreas said, the patches broke both aacenc tests. assuming the changes are intended you most likely forgot to update the ref files :p
[22:13:54 CEST] <jamrial> [aac @ 0000000001ad0200] Unsupported profile -99 ; Error while opening encoder for output stream #0:0
[22:13:56 CEST] <jamrial> or not...
[22:19:34 CEST] <atomnuker> huh
[22:21:12 CEST] <atomnuker> damnit
[22:22:07 CEST] <atomnuker> is profile -99 used as a default value?
[22:22:20 CEST] <atomnuker> when -profile:a isn't specified
[22:25:02 CEST] <atomnuker> yep, profile_unknown
[22:31:13 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:88a5f93f629a: aacenc: treat unknown profile as AAC-LC
[22:37:24 CEST] <cone-525> ffmpeg 03Rostislav Pehlivanov 07master:5df166e43054: aacenc_tns: re-enable coefficient compression
[22:38:29 CEST] <atomnuker> alright, that takes care of those two bugs
[23:42:30 CEST] <nevcairiel> atomnuker: there is another one from msvc; aacenc_tns.c(28) : fatal error C1083: Cannot open include file: 'strings.h': No such file or directory
[23:43:38 CEST] <nevcairiel> i guess that may be a leftover from ffs?
[23:44:25 CEST] <nevcairiel> (even though it would be in string.h, not strings.h)
[23:46:51 CEST] <jamrial> seems to compile fine without that line, so probably safe to remove
[00:00:00 CEST] --- Sat Aug 22 2015


More information about the Ffmpeg-devel-irc mailing list