[Ffmpeg-devel-irc] ffmpeg-devel.log.20170221

[burek]

[00:42:55 CET] <cone-123> ffmpeg 03Michael Niedermayer 07master:ed69cb83f893: fate/source: Check for cases that could use av_clip_uintp2() and av_clip_intp2()
[00:42:56 CET] <cone-123> ffmpeg 03Steinar H. Gunderson 07master:e3c14eaa54c8: speedhq: fix decoding artifacts
[00:42:57 CET] <cone-123> ffmpeg 03Michael Niedermayer 07master:4614bf2caf67: Factorize CHECK/SUINT code
[00:42:58 CET] <cone-123> ffmpeg 03Michael Niedermayer 07master:e04108dfa6d1: avcodec/dca_xll: signed integer overflow: 255251 * 32768 cannot be represented in type 'int'
[00:42:59 CET] <cone-123> ffmpeg 03Michael Niedermayer 07master:0a65dae9d0c0: avcodec/flacdec: reduce limit for golomb so that the max value does not overflow
[00:43:00 CET] <cone-123> ffmpeg 03Michael Niedermayer 07master:e8a3498f2452: avcodec/dca_xll: Fix runtime error: signed integer overflow: -1073741824 * 32768 cannot be represented in type 'int'
[02:08:14 CET] <Chloe> What needs to be done for sub->sub filters?
[02:15:38 CET] <cone-123> ffmpeg 03Carl Eugen Hoyos 07master:a5c1c7a8b3d1: lavf/mpeg: Initialize a stack variable used by memcmp().
[02:16:32 CET] <cone-123> ffmpeg 03Carl Eugen Hoyos 07release/2.8:518158693ec3: lavf/mpeg: Initialize a stack variable used by memcmp().
[02:16:33 CET] <cone-123> ffmpeg 03Carl Eugen Hoyos 07release/3.0:dffd455b9b6a: lavf/mpeg: Initialize a stack variable used by memcmp().
[02:16:34 CET] <cone-123> ffmpeg 03Carl Eugen Hoyos 07release/3.1:007cf1786c8b: lavf/mpeg: Initialize a stack variable used by memcmp().
[02:16:35 CET] <cone-123> ffmpeg 03Carl Eugen Hoyos 07release/3.2:e93e215b36ab: lavf/mpeg: Initialize a stack variable used by memcmp().
[10:17:07 CET] <cone-751> ffmpeg 03Paul B Mahol 07master:74267333a10e: avformat/mpl2dec: skip BOM when probing
[10:34:13 CET] <mateo`> michaelni: Hello, I've ported the simple idct (neon variant) to aarch64 but I just noticed that it is marked as experimental (it is used by default when neon is available on arm). Is it still experimental ? Did you remember what does it mean in terms of quality ?
[10:47:47 CET] <mateo`> michaelni: sorry, looks like the original author is Mans (based on simple idct that was written by you)
[10:48:18 CET] <mateo`> anyway if you have any ideas of why it is still experimental
[10:48:26 CET] <mateo`> that would be helpful
[10:50:08 CET] <mateo`> also, would it be acceptable to have the aarch64 asm code be bitexact with its arm counterpart (and not with the other implementation ? ie: C and x86)
[12:22:07 CET] <cone-751> ffmpeg 03Paul B Mahol 07master:f4777d1b89c6: avcodec/qdrw: add support for decoding rgb555
[13:56:06 CET] <stevenliu> Hi
[13:56:11 CET] <stevenliu> Paul here
[13:57:10 CET] <JEEB> durandal_1707 is him I think
[13:57:48 CET] <durandal_1707> yes?
[13:58:53 CET] <stevenliu> I saw ffmpeg have no AEC(AudioEchoCancellation),is there have plan about it?
[14:02:02 CET] <durandal_1707> stevenliu: how that works?
[14:13:52 CET] <stevenliu> @durandal_1707  look similar with clean noise,but that clean the afilter aecho
[14:14:27 CET] <stevenliu> for example,be used in talk between two people by live streaming
[14:18:04 CET] <durandal_1707> stevenliu: which algorithm
[14:23:06 CET] <stevenliu> ah,let me learn about it :)
[14:24:06 CET] <durandal_1707> stevenliu: speex have some aec
[14:26:49 CET] <stevenliu> I saw lots guys use opus now, there have lots people use WebRTC......
[14:27:40 CET] <stevenliu> Ok,Thanks , Maybe try speex is a good choice
[14:28:25 CET] <durandal_1707> stevenliu: if aec is part of opus use that
[14:30:27 CET] <durandal_1707> ok looks like webrtc one is best oss
[14:32:23 CET] <stevenliu> haha FFmpeg is best one, I trust ffmpeg can fix every thing :D
[15:32:14 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:956472a3236c: avcodec/rv40: Fix runtime error: left shift of negative value
[15:32:15 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:28dc6e729137: avcodec/simple_idct: Fix runtime error: left shift of negative value -6395
[15:32:16 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:cbd622be997e: avcodec/h264_ps: Check delta scale for validity
[15:49:38 CET] <durandal_1707> michaelni: what we told about useless error messages?
[15:51:02 CET] <BBB> how do I tee a libavfilter link?
[15:51:43 CET] <BBB> like, say I had a yuv420p stream but I want to convert it to 422p before I calculate the psnr and ssim
[15:52:49 CET] <BBB> something like ffmpeg -i file -i ref -lavfi [0:v]format=yuv420p[a];[1:v]format=yuv420p[b];[a][b]psnr=-[c];[c][b]ssim=0 -f null - doesnt work becuase [b] is used twice, but without the format it works fine if I use [1:v] twice
[15:53:43 CET] <jkqxz> split
[15:54:04 CET] <BBB> aha
[15:54:19 CET] <BBB> ty!
[15:54:30 CET] <BBB> terminology :-p gstreamer called it tee so I expected it to be called tee
[16:02:03 CET] <Chloe> Anyone know much about dvbsubs (not teletext)? Having an issue with a shitty STB and wondering if it could be solved at all from the encoding end. A large black box flashes after a subtitle finishes (but only after some, and it's not constant either). Can't seem to detect it in decoding as an empty teletext page via a bunch of different methods (checking
[16:02:03 CET] <Chloe> pixel data, text eq. etc).
[16:39:08 CET] <cone-751> ffmpeg 03Steven Liu 07master:0c0aef1caf0c: avformat/hlsenc: fix cid 1401346  Dereferencing pointer error
[17:14:43 CET] <iive> RiCON: what happened with your mingw&opus related problem?
[17:15:00 CET] <RiCON> nothing new from atomnuker
[17:15:47 CET] <RiCON> ruled out that it's msys2 build of gcc and that it's surely opus_pvq_search since libopus' doesn't crash
[17:17:19 CET] <iive> have you tried the -O0 compilation and/or any of the -mno-sse ?
[17:25:16 CET] <RiCON> yeah, same crash
[17:25:53 CET] <RiCON> https://i.fsbn.eu/tC4q.txt
[17:27:01 CET] <cone-751> ffmpeg 03Rostislav Pehlivanov 07master:42959044ac7d: lavfi/buffersrc: fix directly setting channel layout
[17:28:10 CET] <atomnuker> michaelni: how was it that one applied master patches to release branches? cherry-pick?
[17:28:36 CET] <nevcairiel> yes
[17:29:46 CET] <michaelni> atomnuker, yes
[17:31:46 CET] <iive> RiCON: could you add the mno-sse/mmx/avx ?
[17:32:34 CET] <iive> --disable-sse disabled only the hand written assembly. the gcc could use sse for float point.
[17:32:35 CET] <RiCON> isn't that done with --disable-{sse,mmx,avx}?
[17:33:02 CET] <iive> aka, instead of using fpu instructions it uses sse ones.
[17:33:25 CET] <iive> i actually don't remember if it would do this with -O0...
[17:33:51 CET] <RiCON> there's --opts-flags=-O0 --disable-{{inline-,}asm,mmx,sse{,2,3,4,42},avx}
[17:34:25 CET] <RiCON> if there's still some asm there, i don't know what's turning it on
[17:34:29 CET] <nevcairiel> using sse for fpmath is not an "optimization", its just a choice
[17:34:48 CET] <iive> i think -ffast-math allows automatic generation of sse codes for float.
[17:35:06 CET] <nevcairiel> for example a x86_64 build will always use sse fpmath
[17:35:14 CET] <nevcairiel> only 32-bit even  has that choice, really
[17:35:35 CET] <iive> ok, so -fpmath=387 
[17:35:39 CET] <iive> is what i do want.
[17:36:07 CET] <iive> nevcairiel: yes, that' what i am trying to explain/check. Thank you.
[17:44:25 CET] <iive> RiCON: it's the gcc itself that generates sse code on x64, because all x64 cpu do have sse and it is faster than 387 fpu. In other words, if you have code that uses 'float x,y;'  variables it will use sse for it.
[18:09:04 CET] <RiCON> ok, some difference finally, 32-bit stops at the assert and 64-bit keeps going and segfaults
[18:11:02 CET] <RiCON> but i can't gdb the 32-bit since gdb crashes
[18:56:39 CET] <michaelni> atana, do you have time to continue working on peakpoints2 ?
[18:56:55 CET] <atana> yes I am working on it
[19:00:35 CET] <iive> RiCON: so it still segfaults with -fpmath=386 on 64 bit?
[19:00:44 CET] <iive> 387//
[19:05:44 CET] <RiCON> yeah
[19:06:25 CET] <cone-751> ffmpeg 03Nicolas Roy-Renaud 07master:4ec07e943144: avformat/sierravmd: Support for Shivers 2 stereo tracks
[19:06:26 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:0c42d0add37c: avcodec/bmp: Fix runtime error: negation of -2147483648 cannot be represented in type 'int'; cast to an unsigned type to negate this value to itself
[19:06:27 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:a59505ca7671: avcodec/gsmdec_template: Fix runtime error: signed integer overflow: -22527 * 99113 cannot be represented in type 'int'
[19:06:28 CET] <cone-751> ffmpeg 03Michael Niedermayer 07master:631f7484918a: avcodec/ituh263dec: Fix runtime error: left shift of negative value -22
[19:07:10 CET] <RiCON> iive: https://i.fsbn.eu/D2hT.txt
[19:12:30 CET] <kierank> durandal_1707: i don't understand why michaelni just ignore everything we say
[19:14:08 CET] <durandal_1707> kierank: we can go to libav or fork again
[19:18:46 CET] <cone-751> ffmpeg 03Paul B Mahol 07master:4e6b44559a29: avcodec/qdrw: fix writing past end of row
[19:28:14 CET] <michaelni> kierank, iam not intending to ignore anyone, i just want to work, ... a discussion about error messages is something that would keep me from working so i avoided replying. People interrested in such a discussion can discuss it
[19:29:06 CET] <kierank> it's a discussion about you pushing patches with error messages without review
[19:29:30 CET] <kierank> clearly that is a lost argument
[19:29:57 CET] <michaelni> why do you always spread hatread ?
[19:30:32 CET] <michaelni> everything can be discussed, i just have a long todo and id like to work on that, others can discuss
[19:33:38 CET] <kierank> I am not spreading hatred
[19:38:56 CET] <baptiste> drama
[19:43:45 CET] <wm4> what was ignored this time?
[19:47:42 CET] <cone-751> ffmpeg 03Paul B Mahol 07master:770ac75ae9cb: avcodec/qdrw: add support for 2bpp and 4bpp packed pallette format
[19:52:25 CET] <durandal_1707> why i dont have long todo like michaelni does?
[19:52:49 CET] <durandal_1707> can you share some tasks?
[19:57:59 CET] <michaelni> durandal_1707, its not a physical list but sure i can share some, "fix coverity issues", "fix tickets on track which are marked regressions", "go over patches on patchwork make sure everything is reviewed, applied, rejected or supersded", there are many more
[19:59:33 CET] <wm4> durandal_1707: merge libav commits
[20:03:22 CET] <kierank> durandal_1707: ask michaelni to add you to oss-fuzz
[20:03:26 CET] <kierank> And fix bugs 
[20:04:24 CET] <durandal_1707> how much stuff is on oss-fuzz?
[20:04:31 CET] <kierank> Who knows
[20:04:51 CET] <cone-751> ffmpeg 03Rostislav Pehlivanov 07release/3.2:5546294f63e4: lavfi/buffersrc: fix directly setting channel layout
[20:04:56 CET] <cone-751> ffmpeg 03Rostislav Pehlivanov 07release/3.1:e1ed2291ecba: lavfi/buffersrc: fix directly setting channel layout
[20:06:03 CET] <michaelni> i can handle oss-fuzz, its the other things i cannot handle
[20:15:51 CET] <atana> michaelni, updated repo. at line 855 it segfaults. I haven't freed p->mi not sure how it is null
[20:20:52 CET] <durandal_1707> atana repo link?
[20:21:21 CET] <atomnuker> durandal_1707: https://github.com/atana1/ffmpeg
[20:23:33 CET] <michaelni> atana, dunno about the segfault yet, but in the loop you have to compare songid and also matchtime, you only compare songid
[20:28:14 CET] <michaelni> atana, p->mi is likely null as it has not been allocated
[20:28:33 CET] <michaelni> its only allocated if theres a entry added
[20:29:49 CET] <michaelni> the whole sort and find best song id part should be skiped if nothing was found (or a file was added instead of one being lookedup)
[20:32:19 CET] <atana> oh
[20:32:36 CET] <atana> durandal_1707, https://github.com/atana1/FFmpeg/
[20:34:37 CET] <atana> michaelni, matchtime should also be identical?
[20:35:09 CET] <michaelni> yes
[00:00:00 CET] --- Wed Feb 22 2017