[Ffmpeg-devel-irc] ffmpeg-devel.log.20140804

burek burek021 at gmail.com
Tue Aug 5 02:05:02 CEST 2014


[02:44] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:8f6d36b3cb8e: avutil/mem: document NULL related saftey of av_freep()
[03:12] <cone-293> ffmpeg.git 03Justin Ruggles 07master:fbc0b8659967: lavr: Do not change the sample format for mono audio
[03:12] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:c62fea75747d: Merge commit 'fbc0b8659967ea54a8472b5f795270d38bb085dd'
[03:12] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:05ff1a2c05d3: swresample/swresample: Treat mono as planar
[03:17] <jamrial> michaelni: you forgot to remove the old spp dependencies from configure
[03:18] <jamrial> wait, scratch that. they are needed for avdct
[03:49] <cone-293> ffmpeg.git 03Luca Barbato 07master:6a463e7fb4f0: http: Refactor http_open_cnx
[03:49] <cone-293> ffmpeg.git 03Rodger Combs 07master:c2a170ac0ddd: avformat/http: fix tls/tcp protocol after a 302 move in https
[03:49] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:b4d4f8f8266e: Merge commit '6a463e7fb4f028c52d2e2d054f9483f4fff492bc'
[04:05] <cone-293> ffmpeg.git 03Janne Grunau 07master:07d8fa58121b: fate: add informative cpu test
[04:05] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:01fc6533aa41: Merge commit '07d8fa58121be8fe315bd51ab760547fe209a745'
[04:13] <cone-293> ffmpeg.git 03Janne Grunau 07master:caf5ef852bf7: cpu-test: test av_cpu_count
[04:13] <cone-293> ffmpeg.git 03Janne Grunau 07master:ac6b95dbc0b5: aarch64: add ',' between assembler macro arguments where missing
[04:13] <cone-293> ffmpeg.git 03Janne Grunau 07master:a238b83b1364: aarch64: use MACH-O const data asm directive in const macro
[04:13] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:7fd7aa6b2b84: Merge commit 'caf5ef852bf71984d3322bbeaf48cfb04ac8255f'
[04:14] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:7fd60d1e7ad0: Merge commit 'ac6b95dbc0b53b3ea461bd5e5e7f7f31d2983733'
[04:14] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:66eacd55807c: Merge commit 'a238b83b13640e3192d7d4aaad2242f13a9a84a1'
[04:21] <cone-293> ffmpeg.git 03Anton Khirnov 07master:f89d76c10355: mpeg4video: Initialize xvididct for all threads
[04:21] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:2aeeb3ba1f4f: Merge commit 'f89d76c10355242c39b08f253c1d1524f45ef778'
[04:40] <cone-293> ffmpeg.git 03Diego Biurrun 07master:c065f4a0c652: huffyuv: K&R formatting cosmetics
[04:40] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:c1e7fb60d1ea: Merge commit 'c065f4a0c652cc569c927c95148111b3aa2962dc'
[04:48] <cone-293> ffmpeg.git 03Diego Biurrun 07master:b7616f5716dc: huffyuv: Eliminate some pointless casts
[04:48] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:e9cd8f4d20cf: Merge commit 'b7616f5716dce68c749d5dfeab19cf595a11807c'
[04:58] <cone-293> ffmpeg.git 03Diego Biurrun 07master:3160bdc7f7bc: huffyuv: Use avpriv_report_missing_feature() where appropriate
[04:58] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:2b9044a6f0b0: Merge commit '3160bdc7f7bc27bb67561270b4e730cd2d844afd'
[05:04] <cone-293> ffmpeg.git 03Diego Biurrun 07master:6234058148d6: huffyuv: Return proper error codes
[05:04] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:d5fde13bd226: Merge commit '6234058148d6e1ee765913b678a22385a87bffac'
[05:31] <cone-293> ffmpeg.git 03Diego Biurrun 07master:d0393d79bc3d: huffyuv: Check and propagate function return values
[05:31] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:2793b218bdd5: Merge commit 'd0393d79bc3d61c9f2ff832c0e273b7774ff0269'
[06:12] <cone-293> ffmpeg.git 03Kieran Kunhya 07master:1ef9e8376466: avcodec: Deprecate dtg_active_format field in favor of avframe side-data
[06:12] <cone-293> ffmpeg.git 03Kieran Kunhya 07master:2a3c36e920d9: Deprecate AFD field and add AFD as side-data
[06:12] <cone-293> ffmpeg.git 03Michael Niedermayer 07master:71d008ebe4f9: Merge commit '1ef9e8376466bb1e2c147e47554b94cab9c8b04a'
[11:36] <zenderz> hi all. running ffmpeg 2.3 on a mips32 platform and getting a segmentation fault. ran gdb and by the backtrace it looks like av_fast_realloc is being called recursively just before the seg fault. any ideas?
[11:37] <ubitux> maybe a realloc of a malloc'ed area
[11:47] <zenderz> okay will add some debug to see exactly what occurs
[12:14] <zenderz> can someone explain to me the concept behind MEMALIGN_HACK
[12:18] <ubitux> zenderz: it's for plateforms that don't support posix_memalign()
[12:18] <ubitux> so in order to get an aligned buffer, a normal malloc() is done
[12:18] <zenderz> thanks
[12:18] <ubitux> and shifted by the necessary amount to get an aligned address
[12:19] <ubitux> and there is a backpointer thing stored for free
[12:42] <zenderz> hmm. still getting the seg fault with that enabled. seems like i was on the wrong track
[12:49] <cone-125> ffmpeg.git 03Diego Biurrun 07master:bb41115d5693: imgutils: Do not declare avpriv_set_systematic_pal2 in the public header
[12:49] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:22a0249debb1: Merge commit 'bb41115d56930b9f5d59e79dca254d1201246967'
[13:26] <zenderz> seems to be happening in ff_av_fast_realloc
[13:26] <zenderz> Program received signal SIGSEGV, Segmentation fault. 0x7711e764 in ff_av_fast_realloc (ptr=0x0, size=0x0, min_size=0)     at libavcodec/utils.c:130
[13:26] <ubitux> what's the full backtrace?
[13:28] <zenderz> http://pastebin.com/jUQY1ppe
[13:28] <ubitux> lol
[13:28] <ubitux> how do you trigger that?
[13:28] <zenderz> lol. just ran ffmpeg -i testinput.mpg
[13:29] <ubitux> what's ff_av_fast_malloc?
[13:29] <zenderz> what do you mean?
[13:29] <ubitux> what is that function defined?
[13:29] <zenderz> let me check
[13:30] <ubitux> s/what/where/
[13:32] <ubitux> ff_av_fast_realloc doesn't exist (and never did) in ffmpeg afaict
[13:33] <zenderz> in utils.c
[13:33] <zenderz> http://pastebin.com/AJcBsGu2
[13:34] <nevcairiel> might be a weird result of symbol versioning
[13:34] <zenderz> remember i am xcompiling for mips
[13:34] <nevcairiel> sounds like symbol versioning went wrong though, as it calls itself again
[13:34] <zenderz> ah so you think i cant trust the gdb output
[13:36] <zenderz> strange because the seg fault happens almost immediately
[13:37] <zenderz> http://pastebin.com/VrQ1pspG
[13:39] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:7d03732f7e60: ffmpeg: Do not fail if a demuxer and decoder use the same option identifer
[13:39] <cone-125> ffmpeg.git 03ePirat 07master:e3dc2c86fc41: libavformat: Add Icecast protocol
[13:47] <zenderz> any pointers on how i can debug this further?
[13:48] <ubitux> well you probably want to debug the FF_SYMVER thing
[13:56] <zenderz> is it a mapping for function pointers?
[13:57] <nevcairiel> isnt there a configure option to disable symver, might try and see what happens
[13:59] <zenderz> okay i just build with shared disabled and i dont get the segmentation fault anymore
[14:02] <zenderz> can someone explain to me how FF_SYMVER is supposed to work
[14:05] <ubitux> it was introduced in http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=b462d13262f1a1d88c08981a742355f7c99c1959 it seems
[14:15] <kurosu> zenderz: shared failing? sorry to ask the obvious, but are you sure the proper, newest libraries were loaded? eg not some older/unversioned one (from system)
[14:16] <zenderz> pretty sure. but since shared is failing and non-shared is working, im guessing I could have made a mistake along the way. I'm going to retry and will give feedback
[14:18] <kurosu> the fact that these symbols were decorated by versioning looks strange
[14:22] <cone-125> ffmpeg.git 03Marvin Scholz 07master:eb9244f20210: Add Icecast protocol
[14:22] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:491c52d3b077: Merge commit 'eb9244f20210fd420fb9b3c98126f9cae525d1cc'
[14:25] <kurosu> *were not
[14:32] <cone-125> ffmpeg.git 03Janne Grunau 07master:23dfa00b88fc: fate: explicitly set the default THREADS value
[14:32] <cone-125> ffmpeg.git 03Michael Niedermayer 07master:fac292b8cd4b: Merge commit '23dfa00b88fc927d4c1854ab4fc60f5c6398f3ac'
[14:46] <cone-125> ffmpeg.git 03James Almer 07master:b1a44e6bf5e2: x86/hevc_mc: remove an unnecessary pxor
[15:03] <cone-125> ffmpeg.git 03James Almer 07master:b7863c972c93: x86/hevc_mc: use fewer instructions in hevc_put_hevc_{uni, bi}_w[24]_{8, 10, 12}
[17:10] <ubitux> michaelni: the current factorization is supposed to be quite stable numerically, at least that's what the paper says :p
[17:11] <ubitux> it doesn't cover the final scaling, so i did integrate it in the most stupid way; i'll try to merge it automatically in the appropriate place
[17:11] <ubitux> and then apply
[17:33] <ubitux> BBB: ah fuck, replied to you directly
[17:33] <ubitux> :p
[17:43] <michaelni> ubitux, sure, ok, just dont wait for month just because it can be optimized further
[17:54] <cone-891> ffmpeg.git 03Christophe Gisquet 07master:e8c003edd259: x86: hevc_deblock: remove unnecessary masking
[18:23] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:4e855c11b50b: avformat/util: change av_find_default_stream_index() to use a score based system
[18:53] <kurosu> BBB: should xmm6+ be callee-saved under macos x ?
[19:01] <kurosu> BBB: nevermind, it's identical to AMD64/linux
[19:05] <rcombs> ubitux: ninja'd me on that reply
[19:06] <rcombs> kurosu: I enjoy how x86inc.asm lets me not worry about that :)
[19:10] <rcombs> ubitux: query: if a complete index is unavailable and packets ` frames, could ffmpeg parse-but-not-decode the video stream for a speed boost?
[19:11] <wm4> rcombs: uh well, ffms2 is a lib that can index a file
[19:11] <wm4> and normally, number of frames is equivalent to the number of packets for video
[19:12] <wm4> and parsing would require reading the whole file anyway
[19:12] <wm4> but it'd be faster perhaps
[19:12] <rcombs> wm4: yeah, but reading is generally faster than decoding
[19:12] <wm4> but certainly should be an optional feature
[19:12] <rcombs> clearly
[19:13] <wm4> actually it would help greatly with all these semi-broken containerless audio formats (mp3, tta...)
[21:02] <Daemon404> oh damn my icl instances have been locked for liek 2 weeks
[21:02] <Daemon404> due to power outage which left a lock file aorund
[21:02] <Daemon404> nevcairiel, ah fuck... strip.exe segfaults
[21:03] <Daemon404> with my 4.9.1 chain
[21:03] <Daemon404> wtf?
[21:08] <nevcairiel> Mine works
[21:09] <Daemon404> it only degfaults during oen configure check
[21:09] <Daemon404> if you have dialogues off you may have simply not noticed
[21:09] <nevcairiel> But then again its built with 4.8
[21:09] <Daemon404> true
[21:09] <Daemon404> mine is 4.9.1 all the ay
[21:23] <cone-891> ffmpeg.git 03Kostya Shishkov 07master:5420099cab1e: swscale: correctly pad destination buffer in rgb conversion
[21:23] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:8862c49661fb: Merge commit '5420099cab1e915b191cceccec4364f54cec6e52'
[21:33] <cone-891> ffmpeg.git 03Luca Barbato 07master:d2962e9f89cc: pixdesc: Support pixelformat aliases
[21:33] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:073c074447c2: Merge commit 'd2962e9f89cca6ff40f0c9d5ffc9c4397b8b1b26'
[21:38] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:e0d73fda3617: avutil: add alias names for gray 8/16 colour spaces
[21:38] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:0b25261d29c6: Merge commit 'e0d73fda361729d207e51fa2afd8ba6102e58d98'
[22:06] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:e96c3b81cadd: avutil: rename AV_PIX_FMT_Y400A to AV_PIX_FMT_YA8
[22:06] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:4c8bc6fdee95: Merge commit 'e96c3b81cadd0ba84d43b1f3a54980df3785d9a5'
[22:13] <ubitux> michaelni: so, merging the scaling with the previous op:
[22:13] <ubitux> 21880 decicycles in before, 4194020 runs, 284 skips
[22:13] <ubitux> 20836 decicycles in after, 4194086 runs, 218 skips
[22:13] <ubitux> so yeah that's kinda better
[22:13] <ubitux> (http://pastie.org/9445297)
[22:13] <ubitux> assuming that's what you meant?
[22:15] <michaelni> ubitux, yes
[22:19] <ubitux> michaelni: do you want other changes/optim before i apply this version?
[22:23] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:e9abafca278f: avutil: add AV_PIX_FMT_YA16 pixel format
[22:23] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:6017c9803607: Merge commit 'e9abafca278f87e1ecc6d50091d99b73ff63c9a6'
[22:24] <michaelni> ubitux, id say its better to apply and if you want obfuscate/optimize it more after that, that would also allow us to bisect if some numerical stability issues end up being introduced
[22:24] <ubitux> ok thanks
[22:31] <ubitux> Timothy_Gu: ping
[22:33] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:f84a1b597c29: swscale: support AV_PIX_FMT_YA16 as input
[22:33] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:2f5bf2f7f20f: Merge commit 'f84a1b597c29dc035b8d5529ef88c2d7ff057820'
[22:37] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:2257165bff24: png: disable broken MMX/SIMD code for bpp <= 2
[22:37] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:e1795bc38089: Merge commit '2257165bff243534982f4ddabae4e65e2a35f2ab'
[22:45] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:e64f0bf2d2b1: png: support reading gray+alpha at 16 bits
[22:45] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:59412e3a0703: Merge commit 'e64f0bf2d2b1347ec9461f0e82852a62e8c6ffbe'
[22:50] <cone-891> ffmpeg.git 03Vittorio Giovara 07master:bcc5f69b33e2: tiff: support reading gray+alpha at 16 bits
[22:50] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:47080b6d1703: Merge commit 'bcc5f69b33e27b5e11aaea8304ee02f8d895cdab'
[22:57] <cone-891> ffmpeg.git 03Carl Eugen Hoyos 07master:60cbd6ad84de: tiff: support reading gray+alpha at 8 bits
[22:57] <cone-891> ffmpeg.git 03Michael Niedermayer 07master:8f83fb99912f: Merge commit '60cbd6ad84de0931314030b81df87b69d6196587'
[00:00] --- Tue Aug  5 2014


More information about the Ffmpeg-devel-irc mailing list