[Ffmpeg-devel-irc] ffmpeg-devel.log.20160411

burek burek021 at gmail.com
Tue Apr 12 02:05:04 CEST 2016


[00:05:19 CEST] <kierank> should I demand a vote for this faac thing
[00:23:28 CEST] <durandal_170> kierank: you can, but voting may never happen as next meeting is not going to happen, it appears..
[00:32:22 CEST] <atomnuker> pointless, by then we'll have a fast enough aac coder
[00:38:30 CEST] <jamrial> durandal_170: i'm fairly sure voting can be done on the ml
[00:38:53 CEST] <Daemon404> i have yet to see anything go to a vote
[00:41:53 CEST] <atomnuker> no need to make a fuss out of it, it's not that much work to make a fast coder with an RC
[00:42:41 CEST] <nevcairiel> how much performance does one gain by disabling all the feautres like MS, PNS, TNS?
[00:45:21 CEST] <atomnuker> not much, most actually increase the performance (1/2 coeffs for IS, no coeffs for PNS)
[00:45:52 CEST] <atomnuker> though figuring out when to use PNS overweights the performance benefit there
[00:45:53 CEST] <nevcairiel> i thought it might do some extra analysis to see if its worth using that might be skipped and therefor faster
[00:46:21 CEST] <atomnuker> quantization isn't particularly fast
[01:35:12 CEST] <cone-277> ffmpeg 03Michael Niedermayer 07master:02d08da81fd8: avdevice/caca: switch to codecpar
[04:35:25 CEST] <cone-854> ffmpeg 03Michael Niedermayer 07master:4104f1835851: avformat/segment: Pass flags to child context
[04:35:25 CEST] <cone-854> ffmpeg 03Michael Niedermayer 07master:0c9ad94e97ba: avformat/dashenc: Pass flags to child context
[04:35:25 CEST] <cone-854> ffmpeg 03Michael Niedermayer 07master:15fa01786ce6: avformat/hdsenc: Pass flags to child context
[04:44:07 CEST] <jamrial> 25k fps, lol
[06:24:05 CEST] <rcombs> jamrial: SMOOOOOOOOOOTH
[11:36:16 CEST] <durandal_1707> michaelni: your fix for shorten is wrong for valid files
[11:36:41 CEST] <durandal_1707> bitshift can be 32
[12:33:54 CEST] <ubitux> oh, codecpar merged
[12:34:00 CEST] <ubitux> thanks everyone
[12:34:20 CEST] <ubitux> Daemon404: so do you need help now with monkey merging?
[12:35:10 CEST] <ubitux> btw, unrelated but i have all kind of off by ones error in fate-rv20-1239 with aarch64 (under qemu at least)
[12:35:11 CEST] <durandal_1707> its valid to shift int by 32? 
[12:35:19 CEST] <ubitux> but the instance on fate doesn't seem to have that issue
[12:35:41 CEST] <ubitux> durandal_1707: which direction?
[12:35:55 CEST] <durandal_1707> left
[12:36:21 CEST] <ubitux> if the int is negative, left shift is undefined
[12:37:23 CEST] <nevcairiel> which means it still works everywhere we know of
[12:38:15 CEST] <ubitux> we need more aarch64 fate instances :(
[12:38:30 CEST] <durandal_1707> it gives 0?
[12:38:44 CEST] <jkqxz> Left shifting int by 32 is undefined because some architectures mask to only the relevant bits (so it's << 0) and some don't.
[12:43:24 CEST] <rcombs> strictly that depends on the size of int :P
[12:44:02 CEST] Action: rcombs defines an ABI where int is 8-byte, just because we needed more complexity in our lives
[12:44:30 CEST] <rcombs> but hey at least shifting int left by 32 bits would be well-defined there
[12:44:31 CEST] Action: nevcairiel hires a guy for some good old fashioned "wet work"
[12:44:47 CEST] <iive> rcombs: if you want more complexity, try 16bit char 
[12:45:04 CEST] <rcombs> iive: so 16-bit bytes?
[12:45:14 CEST] <nevcairiel> no, only chars
[12:45:18 CEST] <iive> i didn't say that :P
[12:45:22 CEST] <rcombs> iirc C defines sizeof(char) == 1
[12:45:34 CEST] <rcombs> but doesn't define how many bits in a char
[12:46:38 CEST] <JEEB> yeah
[12:47:02 CEST] <rcombs> `CHAR_BIT`
[12:48:03 CEST] <rcombs> oh apparently it's guaranteed to be _at least_ 8 bits, since `signed char` must be "Capable of containing at least the [127, +127] range"
[12:48:12 CEST] <JEEB> aj
[12:48:13 CEST] <JEEB> ah
[13:08:08 CEST] <Daemon404> ubitux, yea sure. lets see how far i get today.
[13:22:03 CEST] <cone-985> ffmpeg 03Paul B Mahol 07master:b62ed56e25c8: avcodec/shorten: properly handle bitshift > 31
[13:44:08 CEST] <ubitux> the fuck is wrong with this fate-rv20-1239 test :(
[14:04:16 CEST] <nevcairiel> lowres be evil
[14:07:10 CEST] <wm4> ubitux: ?
[14:07:32 CEST] <ubitux> it fails on aarch64 here, dunno why
[14:07:38 CEST] <ubitux> all kind of off-by-one
[14:07:54 CEST] <ubitux> it predates codecpar merge afaict
[14:18:29 CEST] <Daemon404> ubitux, well at least thats good to know :)
[14:19:12 CEST] <wm4> lowres for realvideo sure is important
[14:20:47 CEST] <ubitux> it indeed looks related to lowres
[14:21:14 CEST] <ubitux> checksum matches without
[14:32:33 CEST] <Compn> rv20 probably never used for hd anyway 
[14:32:36 CEST] <Compn> rv40 yes ;P
[14:33:10 CEST] <Daemon404> i dont think rv40 has loweres.
[14:34:27 CEST] <ubitux> michaelni: fate-rv20-1239 is failing for me on aarch64 since its addition in 1d64a9d9 (off-by-ones). the issue mismatch seems related to -lowres 
[14:35:41 CEST] <wm4> this lowres crap is such a waste of time and energy
[14:35:41 CEST] <michaelni> ubitux, maybe something with bitexact MC and arm
[14:36:19 CEST] <ubitux> i tried adding sws bitexact/average_rnd just in case but didn't change anything
[14:36:28 CEST] <ubitux> any flag in mind i can try?
[14:38:04 CEST] <Daemon404> oh fyi: marton said he'd work on ffplay codecpar
[14:42:39 CEST] <michaelni> ubitux, dunno, but a bit shooting in the dark, is the difference also there with disabled asm ? and its only happening with lowres or the file without lowres differs too ?
[14:44:12 CEST] <michaelni> ubitux, also "aarch64 	linux 	gcc 4.8 (Ubuntu/Linaro 4.8.2-13ubuntu1)" is passing on http://fatebeta.ffmpeg.org/
[14:44:29 CEST] <Daemon404> maybe it is compiler dependent?
[15:07:02 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:9897d9f4e074: examples/output: convert to codecpar
[15:07:03 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:d76972b58156: Merge commit '9897d9f4e074cdc6c7f2409885ddefe300f18dc7'
[15:08:00 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:a9e1f2cc61cb: examples/qsvdec: convert to codecpar
[15:08:01 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:030e69b4dcb9: Merge commit 'a9e1f2cc61cbd5606a087a60565e87923c39de5a'
[15:13:39 CEST] <Daemon404> that patch from nablet is just gross
[15:15:47 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:ac6d53589f36: examples/transcode_aac: convert to codecpar
[15:15:48 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:6bc5ef37cbf6: Merge commit 'ac6d53589f3631ae08467c784fb371a15c957f01'
[15:29:04 CEST] <Daemon404> do we have any tests that test src_movie?
[15:29:07 CEST] <Daemon404> ^ durandal_1707 
[15:32:20 CEST] <Daemon404> TEST    filter-testsrc2-yuv420p
[15:32:20 CEST] <Daemon404> TEST    filter-testsrc2-yuv444p
[15:32:22 CEST] <Daemon404> ah
[15:32:31 CEST] <Daemon404> oh wait, no.
[15:34:49 CEST] <Daemon404> aside: er... is ffmpeg.org down?
[15:35:27 CEST] <kierank> yes
[15:35:45 CEST] <kierank> oh a bit slow
[15:36:01 CEST] <Daemon404> a bit? it wont load at all for me
[15:36:11 CEST] <Daemon404> oh finally. a few minutes later.
[15:36:14 CEST] <Daemon404> DoS?
[15:39:40 CEST] <durandal_1707> this ape decoder is not lossless here
[15:40:20 CEST] <durandal_1707> borked but numerous security additions
[15:40:31 CEST] <durandal_1707> *by
[15:42:48 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:1138eb5509d3: vsrc_movie: convert to codecpar
[15:42:49 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:8cae006c5646: Merge commit '1138eb5509d3db7f6d565cb45f137a786d22beb9'
[15:43:01 CEST] <Daemon404> durandal_1707, ported and tested src_movie
[15:44:04 CEST] <durandal_1707> people should not be allowed to code....
[15:46:43 CEST] <cone-985> ffmpeg 03Maxym Dmytrychenko 07master:a1335149fd61: qsvenc: store the sync point in heap memory
[15:46:44 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:acc155ac55ba: Merge commit 'a1335149fd610b16459d9281b611282cac51c950'
[15:56:34 CEST] <durandal_1707> so can I get someone to doublecheck ape decoder for me, by encoding with latest monkey encoder? 
[15:57:03 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:3c53627ac17f: qsvdec: store the sync point in heap memory
[15:57:04 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:d30cf57a7b20: Merge commit '3c53627ac17fc6bdea5029be57da1e03b32d265d'
[15:58:54 CEST] <cone-985> ffmpeg 03Diego Biurrun 07master:d6e49096c0c3: idct: Only build prores IDCT if ProRes decoder is enabled
[15:58:55 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:8a6cc30b0428: Merge commit 'd6e49096c0c3c10ffb176761b0da150c93bedbf6'
[15:59:45 CEST] <durandal_1707> anyone?
[16:00:18 CEST] <cone-985> ffmpeg 03Vittorio Giovara 07master:fa55addd23c2: img2: Drop av_ prefix for a static function
[16:00:19 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:8a1c79024da4: Merge commit 'fa55addd23c2f168163175aee17adb125c2c0710'
[16:00:28 CEST] <Daemon404> durandal_1707, you cant use wine?
[16:01:16 CEST] <cone-985> ffmpeg 03Vittorio Giovara 07master:35b1cd343cd7: vc1dec: Drop commented out cruft
[16:01:17 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:e06f5c6f9483: Merge commit '35b1cd343cd703c1b0fc926dc43a92141a357380'
[16:02:00 CEST] <cone-985> ffmpeg 03Vittorio Giovara 07master:f91d94bdfc3f: vc1dec: Properly call deinit function on error
[16:02:01 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:015ca20030be: Merge commit 'f91d94bdfc3f5f83ff0be4d19d10d0a35697386f'
[16:02:30 CEST] <durandal_1707> I used, need confirmation
[16:03:24 CEST] <durandal_1707> By someone on windows
[16:03:40 CEST] <cone-985> ffmpeg 03Vittorio Giovara 07master:01f0e6a0c927: vc1dec: Fix leak on error for array allocations
[16:03:41 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:8cd1323103d3: Merge commit '01f0e6a0c9270f1d5bef08459a6f167cf55e0596'
[16:03:56 CEST] <Daemon404> durandal_1707, do you have a smaple for me to encode
[16:04:31 CEST] <durandal_1707> for the single sample tried it was off by one in bunch of samples
[16:05:16 CEST] <Daemon404> what do you need from me then
[16:05:21 CEST] <Daemon404> a .wav and matching .ape?
[16:05:21 CEST] <Compn> i'm on windows
[16:05:23 CEST] <Compn> whats up
[16:05:28 CEST] <Daemon404> oh hey Compn can do it
[16:05:29 CEST] Action: Daemon404 runs
[16:05:30 CEST] <durandal_1707> I'm in the city now, you could try any sample with silence parts
[16:05:43 CEST] <Compn> oh i'd need latest git ffmpeg...
[16:05:45 CEST] <Compn> cant compile here
[16:05:58 CEST] <durandal_1707> why?
[16:05:59 CEST] <Compn> Daemon404 could do me a crosscompile
[16:06:07 CEST] <Compn> mingw install always corrupts my hd
[16:06:23 CEST] Action: Compn afk brb
[16:06:35 CEST] <durandal_1707> looks like am alone in quest
[16:06:40 CEST] <cone-985> ffmpeg 03Vittorio Giovara 07master:e66fa35392cd: vc1dec: Check group allocations separatedly
[16:06:41 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:3dec7ed3cf04: Merge commit 'e66fa35392cd45d0a80774cd057fb765d60def43'
[16:07:07 CEST] <Daemon404> durandal_1707, i can provide a .ape and matching decoded .wav if you want
[16:07:23 CEST] <Daemon404> using the official encoder
[16:08:01 CEST] <durandal_1707> you just need to see if md5 is same
[16:09:05 CEST] <durandal_1707> I used latest monkey encoder with wine
[16:09:14 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:3e8fd93b6ab2: lavf: add a missing bump and APIchanges for the codecpar switch
[16:09:15 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:6586f62d1a7c: Merge commit '3e8fd93b6ab219221e17fa2b6243cc72cf2d69dc'
[16:09:26 CEST] <Daemon404> durandal_1707, i can do this in an hour or so
[16:09:53 CEST] <durandal_1707> That one couldn't properly decode 24 bit sample we have
[16:10:05 CEST] <durandal_1707> so its one big mess
[16:10:17 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:dc4983d78af2: APIchanges: add missing hashes and dates
[16:10:18 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:0dfbca73bbb3: Merge commit 'dc4983d78af2a666461654067d2e5d45b835358a'
[16:10:37 CEST] <durandal_1707> Daemon404: that would be fine
[16:10:58 CEST] <Daemon404> cool
[16:16:32 CEST] <ubitux> michaelni: -cpuflags none doesn't make a difference. if i remove -lowres from the fate test, regenerate the ref with the x86 code, and try again on aarch64, the test doesn't fail anymore
[16:16:48 CEST] <ubitux> michaelni: and yeah, i saw that on fate.f.o it doesn't fail for some reason
[16:19:43 CEST] <ubitux> i'm using gcc 4.9 though
[16:25:17 CEST] <ubitux> btw, mem leak in fate-yop
[16:27:38 CEST] <cone-985> ffmpeg 03Clément BSsch 07master:c921f4f68797: sws/aarch64: add ff_yuv2planeX_8_neon
[16:31:17 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:e7188a1a8481: avprobe: remove a pointless condition and a dead branch
[16:31:18 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:80195236c85d: Merge commit 'e7188a1a84817b8d4337340c21c552ad0b6cb2fd'
[16:33:20 CEST] <BtbN> i guess master is in an unstable state right now?
[16:33:29 CEST] <BtbN> running into this right now: https://bpaste.net/show/1ab5bcd5ab5c
[16:33:56 CEST] <nevcairiel> its not meant to be broken, feel free to send fixes
[16:34:51 CEST] <Daemon404> BtbN, concatdec is very "special" (stupid)
[16:34:54 CEST] <Daemon404> it may have been broke
[16:34:55 CEST] <Daemon404> n
[16:35:07 CEST] <BtbN> a concat list with more than two files, no matter what is in them, crashes
[16:35:25 CEST] <BtbN> (Still valid video files, not fuzzing)
[16:35:31 CEST] <Daemon404> what does safe 0 do
[16:35:46 CEST] <Daemon404> ah ... allows files with absolute paths
[16:35:49 CEST] <BtbN> disables the strange safe-file rule, so i can use a concat list with absolute paths
[16:35:57 CEST] <Daemon404> its not very strange
[16:36:38 CEST] <durandal_1707> have backtrace?
[16:37:14 CEST] <BtbN> gdb is still installing
[16:38:52 CEST] <BtbN> Is there anything better than the concat demuxer?
[16:40:02 CEST] <Daemon404> wq
[16:40:12 CEST] <wm4>     x4->params.i_fps_num = avctx->time_base.den;
[16:40:12 CEST] <wm4>     x4->params.i_fps_den = avctx->time_base.num * avctx->ticks_per_frame;
[16:40:21 CEST] <wm4> is this how encoders are supposed to get the framerate?
[16:40:47 CEST] <Daemon404> that sort of code has always been derpy
[16:41:27 CEST] <nevcairiel> there is a new framerate field in avctx which is supposed to do this, but previously it was that yes
[16:41:53 CEST] <wm4> well this is still in master
[16:42:55 CEST] <BtbN> not sure what to make of that backtrace, it looks like it failed to find the debug symbols? https://bpaste.net/show/771002fcc7c0
[16:44:16 CEST] <jkqxz> AVCodecContext.framerate: "   * - encoding: unused".  That one?
[16:44:35 CEST] <durandal_1707> using static or dynamic build?
[16:45:53 CEST] <nevcairiel> suppose its still time_base for encoders then
[16:48:43 CEST] <BtbN> seems like something is messed up with those input files as well, strange. They were created by passing an rtmp stream through the segment muxer
[16:50:44 CEST] <BtbN> durandal_1707, https://bpaste.net/show/979126636425
[16:52:04 CEST] <durandal_1707> that's ffprobe?
[16:52:19 CEST] <Daemon404> BtbN, i have a patch to test
[16:52:22 CEST] <Daemon404> for you
[16:52:23 CEST] <Daemon404> see if it fixes it
[16:52:46 CEST] <BtbN> durandal_1707, yes. But ffmpeg -i .. crashes the same
[16:53:43 CEST] <Daemon404> http://pastie.org/10793601
[16:53:47 CEST] <Daemon404> try this
[16:55:13 CEST] <nevcairiel> why would this start appearing now
[16:55:27 CEST] <Daemon404> nevcairiel, because concatdec used to just leak memory
[16:55:29 CEST] <Daemon404> i added frees
[16:55:42 CEST] <nevcairiel> how evil of you
[16:56:20 CEST] <BtbN> i also wonder what the hell the segment muxer did to these .ts segments. Something went horribly wrong there.
[16:56:40 CEST] <BtbN> Daemon404, src/libavformat/concatdec.c:369:21: error: used struct type value where scalar is required
[16:57:39 CEST] <Daemon404> running fate
[16:57:43 CEST] <Daemon404> im sure its just a silly thing
[16:57:47 CEST] <Daemon404> give a sec
[16:58:02 CEST] <nevcairiel> you check a struct in a boolean thing which is not a pointer
[16:58:22 CEST] <Daemon404> ah
[16:58:30 CEST] <Daemon404> just remove that if then i suppose
[16:58:52 CEST] <nevcairiel> should probably iterate over file->nb_streams, as that matches the actually allocated streams, instead of avf->nb_streams
[16:59:11 CEST] <Daemon404> ah that could be why yes
[16:59:18 CEST] <Daemon404> that should probably be the correct fix
[16:59:19 CEST] <nevcairiel> should be the same, except in an error case
[16:59:41 CEST] <Daemon404> BtbN, ignore the patch i pasted you and just change what nevcairiel said
[16:59:54 CEST] <Daemon404> checking if streams exists doesnt hurt though
[17:00:42 CEST] <BtbN> Heading home now, will test in an hour or so
[17:01:48 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:168a443d43b1: avprobe: print information from the codec descriptor
[17:01:49 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:6614214ece70: Merge commit '168a443d43b10ef6a3545d64b2f8bc90144ce4b7'
[17:02:37 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:c80344d01015: mpegvideo_enc: use avcodec_free_context() instead of av_free()
[17:02:38 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:95348174ef87: Merge commit 'c80344d0101558098a6cd2ed5082ff5fda7ca18b'
[17:14:54 CEST] <durandal_1707> Daemon404: don't forget
[17:19:33 CEST] <Daemon404> itll get done today... i have work still though
[17:33:28 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:c9478410c68c: avprobe: add local per-file state
[17:33:29 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:9d48aa6d7df6: Merge commit 'c9478410c68c04261f9cfcd80474607e50bd1852'
[17:46:53 CEST] <cone-985> ffmpeg 03Anton Khirnov 07master:567d6d5f9d14: avprobe: add local per-stream state
[17:46:54 CEST] <cone-985> ffmpeg 03Derek Buitenhuis 07master:48fb5294d03f: Merge commit '567d6d5f9d1400f00445183b3477391f58979aa3'
[18:07:38 CEST] <durandal_1707> is this ebu-stl subtitle xml thing?
[18:08:18 CEST] <kierank> no
[18:08:21 CEST] <kierank> that's ebu-tt
[18:08:27 CEST] <kierank> stl is the one from the 80s
[18:08:30 CEST] <michaelni> Daemon404, ive a small fix for examples/muxing, should i push it or wait? dont want to cause problems for any in progress mere
[18:08:40 CEST] <Daemon404> michaelni, theres no more merges for examples
[18:08:54 CEST] <Daemon404> it is safe to push
[18:10:10 CEST] <cone-985> ffmpeg 03Michael Niedermayer 07master:0c79c96cf20f: doc/examples/muxing: Add support to pass flags to muxer as since codecpar the codec flags are not available to the muxer anymore
[18:15:46 CEST] <cone-985> ffmpeg 03Matthieu Bouron 07master:4c2244127631: swscale/arm: add yuv2planeX_8_neon
[19:18:23 CEST] <BtbN> nevcairiel, Daemon404, durandal_1707: That fixed the crash
[19:18:40 CEST] <BtbN> https://gist.github.com/3c5f8c5cb437530706b6c40cb050bd17
[19:19:47 CEST] <BtbN> Should I just push that, or send it to the ML?
[19:30:35 CEST] <durandal_1707> Imho just push, isn't that file under direct maintership?
[19:31:07 CEST] <BtbN> grep for concat in MAINTAINERS yields no results
[19:34:18 CEST] <durandal_1707> then just push IMO
[20:09:34 CEST] <wm4> #define DEFAULT 0 //should be NAN but it does not work as it is not a constant in glibc as required by ANSI/ISO C
[20:09:37 CEST] <wm4> loving this
[20:11:35 CEST] <TD-Linux> how old is that line
[20:14:28 CEST] <wm4> 2005
[20:15:12 CEST] <wm4> "AVOption first try"
[20:29:54 CEST] <cone-985> ffmpeg 03Lou Logan 07master:03d8fee912a4: doc/filters: document testsrc2 source filter
[22:25:50 CEST] <cone-985> ffmpeg 03Timo Rothenpieler 07master:901b0f1a219e: avformat/concatdec: Use correct stream count on close
[22:39:10 CEST] <durandal_170> Daemon404: looks like I found bug and not regression
[22:40:16 CEST] <Daemon404> ok... cause i didnr get to testing :x
[22:47:29 CEST] <omerjerk> hey what's the best way to allocate a 2d array in ffmpeg ?
[22:48:15 CEST] <omerjerk> for e.g. float** array = av_malloc_array(5, sizeof(float*));
[22:48:59 CEST] <omerjerk> and then iterate over array[i], and do array[i] = av_malloc_array(5, sizeof(float));
[22:49:13 CEST] <omerjerk> is this okay ?
[22:50:21 CEST] <BtbN> why not just allocate a normal array and calculate the position?
[22:50:24 CEST] <ubitux> if it's small you can also do a 1d alloc and access y*N+x
[22:51:12 CEST] <omerjerk> I can do that. But just wondering what should I go with.
[22:51:45 CEST] <ubitux> if fixed, you can add a[5][5] into the local context, it will be heap allocated automatically
[22:51:51 CEST] <omerjerk> the array's dimensions will be number of channels * frame size
[22:52:11 CEST] <omerjerk> it's not fixed.
[22:52:34 CEST] <ubitux> then your initial solution makes sense to me
[22:52:55 CEST] <nevcairiel> sounds like you want storage for your audio samples, you probably want get_buffer that
[22:53:36 CEST] <omerjerk> yes. I want to store the samples in a 2d array inside the context. 
[22:53:43 CEST] <nevcairiel> what for
[22:53:57 CEST] <omerjerk> to write to the file later. 
[23:18:03 CEST] <rcombs> wat
[23:18:20 CEST] <rcombs> omerjerk: you seem to be describing an AVFrame
[23:24:53 CEST] <omerjerk> rcombs: I'm adding a 2d raw_float array here - https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/alsdec.c#L228
[23:27:35 CEST] <nevcairiel> there appears to already be one for int32 samples, so why don't you do exactly what it does
[00:00:00 CEST] --- Tue Apr 12 2016


More information about the Ffmpeg-devel-irc mailing list