[Ffmpeg-devel-irc] ffmpeg-devel.log.20140623

[burek]

[00:00] <Timothy_Gu> debianuser: yep, it's the ogg's problem
[00:01] <Timothy_Gu> It is actually 0.02 seconds long, not 3.43
[00:01] <Timothy_Gu> try ffmpeg -i 2.ogg -f null -
[00:01] <Timothy_Gu> Also, there is a warning: [ogg @ 0xafbf3c0] Estimating duration from bitrate, this may be inaccurate
[00:02] <Timothy_Gu> that explains the inaccuracy of the duration of 2.ogg
[00:02] <Timothy_Gu> All in all, the output file from concat filter is completely fine
[00:02] <debianuser> If it's 0.02 seconds long then how could I hear it?
[00:02] <debianuser> ;)
[00:04] <nevcairiel> obviously your ogg file is broken
[00:06] <Timothy_Gu> Hey also nevcairiel: mind if you send a patch disabling all those stupid anonymous struct/union warnings for MSVC?
[00:07] <nevcairiel> i'm not at home for the next two weeks, i'm not going to send anything
[00:07] <Timothy_Gu> warning C4201: nonstandard extension used : nameless struct/union
[00:07] <Timothy_Gu> oh, ok
[00:07] <nevcairiel> also, some people would say that you better name them
[00:08] <Timothy_Gu> in C11 anonymous unions are supported
[00:08] <nevcairiel> but we use C99
[00:09] <Timothy_Gu> i know, but the flood of this kind of warnings really masks the warnings we should care about
[00:11] <nevcairiel> flood? its maybe 30, and 20 of those are all collected in dxva2api.h (which is a C++ header, where this is technically allowed)
[00:12] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:c69f7299473d: ffmpeg: fix reinitializing with threads while flushing
[00:13] <michaelni> ubitux, your hqx filter tests are evil ;)
[00:14] <debianuser> Timothy_Gu: It seems you're right. It's still a bug in ffmpeg, but in a different place: `ffmpeg -i 1.wav -i 2.ogg -i 3.wav -filter_complex concat=n=3:v=0:a=1 -ac 2 -ar 48000 0.wav` at the end says "time=00:00:23.83", but `ffprobe 0.wav` shows: "Duration: 00:00:25.88". So the output file is correct, but the output TEXT is not.
[00:18] <Timothy_Gu> This is because ffmpeg calculates the output duration for file header in advance with the incorrect ogg duration
[00:18] <cone-826> ffmpeg.git 03Diego Biurrun 07master:49470b940458: build: Add missing replaygain object file for CAF demuxer
[00:18] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:2ca4802601c2: Merge commit '49470b9404583ab3aff6835f78a759349d331696'
[00:20] <Timothy_Gu> michaelni: any idea why they fail on ARM?
[00:21] <Timothy_Gu> and FreeBSD?
[00:22] <michaelni> hqx seems using floats init, this might cause some failure
[00:23] <Timothy_Gu> floats are evil
[00:24] <michaelni> for bitexact regression tests yes 
[00:25] <michaelni> loosing the floats should be easy, ill look at it in an hour or 2 if noone else did it by then
[00:25] Action: J_Darnley recognises those constants
[00:28] <debianuser> Timothy_Gu: ffmpeg says "time=00:00:23.83" at the end, after it's finished encoding. And if that was calculated from incorrect ogg length, that would be 27.24 (10.92+3.43+12.89), not 23.83. Anyway, incorrect output text is not too bad bug for me, I just wasn't aware of it.
[00:29] <Timothy_Gu> Oh, well
[00:30] <Timothy_Gu> At least you now know concat is working :)
[00:31] <cone-826> ffmpeg.git 03Diego Biurrun 07master:03391b5432a2: build: Add missing object files for Matroska demuxer
[00:31] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:738272826f4b: Merge commit '03391b5432a2b8b8dbc3d2db722a54dfb68737b9'
[00:37] <cone-826> ffmpeg.git 03Diego Biurrun 07master:85d805d5cccb: build: Add missing object files for webm muxer
[00:37] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:9e8c92226fe1: Merge commit '85d805d5cccb2f57e7f2e2254d7a870f05926cfd'
[00:38] <ubitux> michaelni: it seems they help finding bugs everywhere ;)
[00:38] <ubitux> for the float i could have a look, but not now
[00:39] <ubitux> probably this week
[00:39] <ubitux> do you have mismatches because of the floats?
[00:40] <michaelni> ubitux, i dont know, ive not reproduced any yet
[00:41] <michaelni> i was thinking its easier to just get rid of the floats than to test if they do cause any of the remaining problems
[00:44] <debianuser> Timothy_Gu: Yes, thank you for your help. And for the great program you all are creating here.
[00:44] <debianuser> While I'm here, I also have a lot of wav files (from rift game resources), which are basically "adpcm_ima_wav", but stereo files have blocks interleaved, not nibbles, like in "adpcm_ima_qt". Are you interested in supporting such files? If yes, how can I help?
[00:45] <cone-826> ffmpeg.git 03Diego Biurrun 07master:c54e118722cb: build: Have the eatqi decoder depend on the MPEG-1 decoder
[00:45] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:9d91fae8a109: Merge commit 'c54e118722cbbdc04945538d1796d4472a1ff406'
[00:51] <cone-826> ffmpeg.git 03Diego Biurrun 07master:fbf4d2197f35: build: Record explicit dependency of intrax8 on error resilience
[00:51] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:7695a538ed24: Merge commit 'fbf4d2197f35bddccec00ad53185ee4ab81778a0'
[00:53] <Timothy_Gu> debianuser: you should probably open a ticket on trac.ffmpeg.org . No guarantee on when we will implement it, but at least we will have a record and samples for that.
[00:57] <cone-826> ffmpeg.git 03Diego Biurrun 07master:a6a675a79287: configure: Drop stray error resilience dependencies
[00:58] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:ffc2c095ef7c: Merge commit 'a6a675a79287565c3c23455439b114676b63a96c'
[01:01] <ubitux> michaelni: i'm removing the float, don't worry about it
[01:02] <michaelni> ok
[01:02] <michaelni> thx
[01:03] <cone-826> ffmpeg.git 03Diego Biurrun 07master:8d695dc411c1: mpeg: Mark ff_mpeg_set_erpic() static
[01:03] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:412015694c08: Merge commit '8d695dc411c1ae28477f107d6f7bf2db59af95df'
[01:12] <cone-826> ffmpeg.git 03Diego Biurrun 07master:f83896abda1e: ape: Replace memset(0) by zero initialization
[01:12] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:efec73043184: Merge commit 'f83896abda1e78ebbbda0f184b682b4fabadc682'
[01:19] <cone-826> ffmpeg.git 03Diego Biurrun 07master:73953df71a28: Replace av_malloc() and memset(0) by av_mallocz()
[01:19] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:32365667cceb: Merge commit '73953df71a28e94d8959facba7906cf5975ca4ef'
[01:25] <kierank> For some reason ff_hevc_decode_nal_pps is slow on this file
[01:26] <nevcairiel> Decoding the pps is slow?
[01:27] <nevcairiel> Isn't that like a few bytes of flags?
[01:27] <wm4> maybe the file consists entirely of repeated PPSs
[01:28] <cone-826> ffmpeg.git 03Luca Barbato 07master:0ef256d51518: libx264: Correctly manage constant rate factor params
[01:28] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:49f1b869b60f: Merge commit '0ef256d51518f0e483c38b9ee3b993cf0709d00e'
[01:36] <cone-826> ffmpeg.git 03Diego Biurrun 07master:463a7cde563f: Differentiate assembler and assembly in documentation and comments
[01:36] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:a125f87ac49b: Merge commit '463a7cde563fd805864c48a76dd1b03fc24671ed'
[01:44] <kierank> https://www.dropbox.com/s/9bvezkhvrv8ne4k/BBC4k.002.ts
[01:44] <kierank> nevcairiel: yeah it's weird
[01:49] <ubitux> michaelni: http://pastie.org/pastes/9315169/text
[01:49] <ubitux> a bit ugly but this seems to do the trick
[01:50] <michaelni> ubitux, LGTM
[01:50] <michaelni> ubitux,  or you could divide by 1000
[01:50] <michaelni> and use prettier numbers
[01:51] <michaelni> hmm no
[01:51] <michaelni> i realize te array is a bit big so divide would possibly suck even for init
[01:52] <ubitux> i'm not sure about the bithack because of signdness
[01:52] <ubitux> and my brain is off :(
[01:52] <ubitux> it's late, gtg
[01:54] <michaelni> ubitux, sleep well in that case, dont worry about it
[02:06] <kierank> smarter: any idea how to optimise a slow ff_hevc_decode_nal_pps 
[02:21] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:42e6fc147b5a: avcodec/apedec: optimize sign combination
[02:21] <cone-826> ffmpeg.git 03James Almer 07master:c172683bf46a: x86/dsputil: remove redundant global motion compensation code
[03:25] <cone-826> ffmpeg.git 03Justin Ruggles 07master:44127546b0a8: Check if an mp3 header is using a reserved sample rate.
[03:25] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:2bcf34a0ef60: Merge commit '44127546b0a81dc9dd6190739a62d48f0044c6f3'
[03:43] <cone-826> ffmpeg.git 03Justin Ruggles 07master:f2f2e7627f0c: Check mp3 header before calling avpriv_mpegaudio_decode_header().
[03:43] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:ca350378de06: Merge commit 'f2f2e7627f0c878d13275af5d166ec5932665e28'
[04:02] <debianuser> Found another problem with concat filter. Is it expected that `ffmpeg -i mono8kHz.wav -i stereo48kHz.wav -ac 2 -ar 48000 -filter_complex concat=n=2:v=0:a=1 concat.wav` produces a file that sounds like 8kHz mono? Isn't concat filter supposed to select "common" sampling rate and channel layout?
[04:06] <cone-826> ffmpeg.git 03Diego Biurrun 07master:7b9ef8d701c3: mpeg: Split error resilience bits off into a separate file
[04:06] <cone-826> ffmpeg.git 03Michael Niedermayer 07master:56afbe8dbc6e: Merge commit '7b9ef8d701c319c26f7d0664fe977e176764c74e'
[04:06] <nevcairiel> debianuser: it probably just sticks to whatever came first
[04:07] <nevcairiel> anyway whats "common"? one needs upsampling, one downsampling. For us its obvious that upsampling may be the better idea, but the filter may not know that :p
[04:13] <michaelni> id say the filter should avoid loosing information by default and downsample would loose info, a patch that improves this would be welcome
[04:20] <debianuser> Hm. So it ignores those "-ac 2 -ar 48000" options... Then I'll use a workaround: ffmpeg -i mono8kHz.wav -i stereo48kHz.wav -filter_complex aresample=48000:och=2,concat=n=2:v=0:a=1 concat.wav
[04:22] <nevcairiel> those options are probably applied after the filter_complex before the final output
[04:22] <nevcairiel> its not smart enough to combine filter_complex with any other options really
[04:52] <Timothy_Gu> Yeah. I personally consider -ac -ar -s and other similar options half deprecated, especially when using filtering.
[11:25] <plepere> what's the time function to test if a function is quicker ?
[11:26] <plepere> I did an AVX version of a function, but I think it's slower than the SSE2 version
[11:26] <J_Darnley> The {START,STOP}_TIMER macros?
[11:26] <plepere> ah great, thanks
[11:27] <J_Darnley> Just make sure the block/fucntion you time processes the same amount of data each time it is called.
[11:27] <plepere> well I can just run the same videos
[11:28] <J_Darnley> No
[11:29] <J_Darnley> I mean the function needs to process the same number of samples everytime it is run.
[11:29] <plepere> it's a function on a specific block size
[11:30] <J_Darnley> If it processes 30 then 10 then 20 then 10 it will give a nonsense result
[11:30] <J_Darnley> goog
[11:30] <J_Darnley> good
[11:38] <plepere> yeah, AVX is slower. :(
[12:10] <kierank> plepere: 12:44 am " kierank https://www.dropbox.com/s/9bvezkhvrv8ne4k/BBC4k.002.ts
[12:10] <kierank> Some bizzare benchmarking on this filw
[12:11] <kierank> File
[12:56] <plepere> kierank, you want me to run some benches on that ?
[12:56] <kierank> plepere: you can but the weirdest bit is that a lot of time is spent on PPS parsing
[12:56] <plepere> it's hevc ?
[12:57] <kierank> yes
[12:57] <kierank> plepere: see pm
[12:58] <kierank> plepere: I've written a small patch to the slow part of the pps parsing but it doesn't make much difference yet
[12:59] <kierank> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/hevc_ps.c;h=30e58a20b1e03443906909a9c65f98a7f832eeb6;hb=HEAD#l1307
[12:59] <kierank> these two loops are sloow
[13:03] <kierank> I tried changing the multiplies into shifts but nothing happened
[13:04] <plepere> ok
[13:33] <kierank> plepere: http://obe.tv/about-us/obe-blog/item/12-first-look-at-bbc
[13:33] <kierank> wrote a small blog post about it
[13:35] <kierank> also I thinks some of the functions need inlining
[13:35] <plepere> the most significant missing optimisations in ffmpeg are the whole idct transforms
[13:35] <kierank> yeah I mention that in my post
[13:36] <plepere> if only we could use intrinsics in ffmpeg.
[13:36] <iive> hum?
[13:37] <plepere> we're even faster with intrinsics than with assembly. :/
[13:39] <iive> we have bad past experience with intrinsics.
[13:40] <plepere> :(
[13:40] <iive> and i don't understand the remark about missing whole idct tranforms.
[13:40] <kierank> iive: we have no hevc idct asm
[13:40] <plepere> we have the idct transforms in intrinsics and in C, but not in ASM
[13:41] <plepere> in openhevc
[13:42] <plepere> and for ffmpeg to be able to do real-time 4K, someone needs to translate the intrinsics to ASM
[13:42] <iive> aha h265
[13:43] <iive> why intrinsics are not committed in ffmpeg in the meantime?
[13:43] <plepere> we were told that intrinsics were more or less banned from ffmpeg
[13:43] <kierank> they are
[13:43] <J_Darnley> Why duplicate the effort?  Why not use openhevc as a library in ffmpeg?
[13:44] <kierank> J_Darnley: openhevc is a copy of libavcodec
[13:44] <kierank> with intrinsics
[13:44] <kierank> and other scalable features
[13:44] <J_Darnley> oh
[13:45] <plepere> the scalable is so good. We've got multi-view coming along too. :D
[13:46] <plepere> but yeah, being able to add intrinsics as a temporary solution would really boost the performance
[13:47] <plepere> at least enough to bring things to realtime
[13:48] <iive> aren't BBB and ubitux working on the asm idct at the moment?
[13:48] <plepere> vp9
[13:48] <plepere> and it's with a different way of doing it
[13:50] <iive> well, imho, you should send RFC to the maillist asking if it is ok to commit the intrinsics, until somebody comes forward and rewrites them in ASM.
[13:51] <plepere> I think things have already been negociated by smarter, michaelni and all.
[13:52] <iive> what have been negotiated?
[13:52] <plepere> well about the "no intrinsics in ffmpeg"
[13:53] <cone-622> ffmpeg.git 03Diego Biurrun 07master:c67b449bebbe: dsputil: Split bswap*_buf() off into a separate context
[13:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:35bb74900b95: Merge commit 'c67b449bebbe0b35c73b203683e77a0a649bc765'
[13:56] <iive> afaik, intrinsics are avoid because they are kind of volatile. compilers might generate horribly inefficient code if some options are tweaked, and sometimes miscompile them.
[13:57] <iive> e.g. ppc assembly is entirely written in intrinsics and it is full of hacks for one version of compiler or another.
[13:57] <plepere> :/
[13:58] <BBB> iive: no
[13:58] <BBB> iive: no intrinsics
[13:59] <iive> that's why we prefer an assembly that is guaranteed to produce consistent results and doesn't need compiler hacks.
[13:59] <iive> BBB: ok :)
[14:00] <BBB> and Im happy to commit intermediate work for all hevc asm
[14:00] <BBB> but no intrinsics
[14:00] <iive> that's why I asked if you are working on it.
[14:00] <kierank> plepere: it's probably the one thing both sides agree on
[14:00] <kierank> out of all the hatred
[14:01] <ubitux> don't yell it too loud, it could be a reason to disagree
[14:01] <BBB> so true
[14:01] Action: iive imagines all libav assembly written in intrinsics.
[14:01] <BBB> iive: Im not working on hevc asm. honestly, if someone wants me to work on it, they probably have a commercial interest in it, and they should pay me marketrate; else, they are free to use vp9
[14:02] <plepere> why so much hate ?
[14:02] <kierank> plepere: blog.pkh.me/p/13-the-ffmpeg-libav-situation.html
[14:03] <ubitux>  create mode 100644 vp9/encoder/vp9_speed_features.c
[14:03] <ubitux> mmh
[14:04] <cone-622> ffmpeg.git 03Dirk Ausserhaus 07master:e121ac634ba3: indeo45: use is_indeo4 context flag instead of checking codec ID
[14:04] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:b222e0773c78: Merge commit 'e121ac634ba324a318f4a97f978dcfb48da6b735'
[14:13] <plepere> kierank, thanks for the read. :)
[14:15] <kierank> plepere: any idea why nal_pps is so slow on ffmpeg but not in openhevc
[14:15] <kierank> even though it's the same code
[14:22] <plepere> the pps in openhevc was just optimized recently
[14:22] <kierank> how do I do that
[14:22] <kierank> whoops
[14:32] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:d6af26c55c1e: avutil/lzo: Fix integer overflow
[14:34] <ubitux> libav, first on security
[14:34] <ubitux> ;)
[15:03] <cone-622> ffmpeg.git 03Luca Barbato 07master:ccda51b14c0f: lzo: Handle integer overflow
[15:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:cf2b7c01f81c: avutil/lzo: add asserts to be double sure against overflows
[15:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:957457a044f6: Merge commit 'ccda51b14c0fcae2fad73a24872dce75a7964996'
[15:49] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:7d9c059a3525: avutil/lzo: Fix integer overflow
[15:49] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:61796a8999f5: avutil/lzo: add asserts to be double sure against overflows
[16:02] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:63e3a9781579: avcodec/aic: fix quantization table permutation
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:e173834af8b5: avformat/flvenc: Do not allow creating h263/mpeg4 in flv without unofficial format extensions being enabled.
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:dad0c9d686a3: avcodec/alsdec: Clear MPEG4AudioConfig so that no use of uninitialized memory is possible
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:252a0ccb80b4: cavsdec: fix qp fixed slice handling
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:e4cdde96b381: avformat/cavsvideodec: Fix probing when the file extension is avs
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:d07be523f5d8: avformat/mpc: attempt to allocate a packet that is not smaller than the data inside it
[16:03] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.2:e0a03d1f9cb1: update for FFmpeg 2.2.4
[16:44] <cone-622> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n2.2.4': unknown revision or path not in the working tree.
[16:44] <cone-622> Use '--' to separate paths from revisions
[16:44] <cone-622> refs/tags/n2.2.4:HEAD: Merge commit 'ccda51b14c0fcae2fad73a24872dce75a7964996'
[17:21] <kierank> mraulet: can you send that pps patch please =p
[17:23] <J_Darnley> Damn!  When explorer crashes, it crashes HARD!
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:946810d1da50: ffmpeg: dont call exit_program() from a signal hander
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:4201c7454551: avformat/mp3enc: use av_copy_packet()
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:2b5b21dbaca0: avcodec: Add padding after the remaining AVFrames
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:b8f6069bf398: swscale/swscale: fix integer overflow
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:998b5ac3a043: dox/scaler:fix bicubiclin typo
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:380587cc34c2: avcodec/x86/mpegvideoenc_template: fix integer overflow
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:61e5001b265b: avcodec/h264_mp4toannexb_bsf: prepend global headers before any in stream parameter sets
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:e4f9a5efa15a: avcodec/libx264: move where x264opts is applied down so it isnt overridden by avctx & defaults
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:4c9c76af8521: avfilter/vf_pullup: fix gray8
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:c5085a030143: avcodec/h264: clear cur_pic structure instead of duplicating it in ff_h264_update_thread_context()
[17:56] <cone-622> ffmpeg.git 03Lukasz Marek 07release/2.1:46b64d981884: lavu/opt: validate range before dereference
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:861732246ff2: avcodec/wma: use av_freep(), do not leave stale pointers in memory
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:9c83914a1fd6: avcodec/msrle: check return code for success before use
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:37f854f5b958: swresample/resample: Limit filter length
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:15871a90aa42: swresample/dither: use av_malloc_array()
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:9d8e2314592a: swresample/resample: use av_malloc_array() where appropriate
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:672919388651: swscale/x86/swscale_template: loose hardcoded dstw_offset
[17:56] <cone-622> ffmpeg.git 03Anthoine Bourgeois 07release/2.1:26d72b3df0ce: avcodec/dirac_arith: Fix build with PIC and stack-check options
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:9c358c6e3b34: avutil/lzo: Fix integer overflow
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:c33770e74bfc: avutil/lzo: add asserts to be double sure against overflows
[17:56] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:d2e3afbbd167: avcodec/x86/idct_sse2_xvid: fix non C99 inline function
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:522865543819: swscale/swscale: fix srcStride/srcSlice typo
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:861e453996f4: avcodec/mjpegen: Fix declared argument size
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:1d9125137039: avformat/mpegts: Remove redundant check
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:ec2ce83913ca: avcodec/diracdec: fix undefined behavior with shifts
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:af109ff12581: avcodec/g723_1: add assert to help static code analyzers
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:90a384cde163: avfilter/f_select: fix loss of precission in SAD calculation
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:4eb0b6c59051: avfilter/vf_deshake: fix loss of precission with odd resolutions
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:df208d6ab939: iavcodec/vc1dec: Fix missing {}
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:9cde200ba70f: avfilter/filtfmts: Support dynamically allocated in/outputs
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:ac182d6d8b50: avformat/mux: Check for and remove invalid packet durations
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:d89e5b20f2b9: avformat/h263dec: Fix h263 probe
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:2b45eab4c913: ffmpeg_filter: fix pointer to local outside scope
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:aaf6ddfa4d9d: avcodec/mjpegdec: Fix undefined shift
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:2a543e58ec52: avfilter/graphdump: Fix pointer to local outside scope
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:2572372f909a: avcodec/diracdec: move mc buffer allocation to per frame
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:6a51bd7360cb: avutil/cpu: force mmx on selection of higher x86 SIMD features
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:e3b6e23a0f64: avcodec/libvorbisenc: dont add the duration to AV_NOPTS_VALUE
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:76838fb171ab: avcodec/aac: fix () in IS_CODEBOOK_UNSIGNED macro
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:60ed43933402: avcodec/golomb-test: fix () in EXTEND() macro
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:a71b4209c26a: avcodec/h264: fix () in macros
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:ccbcb17960a3: avcodec/ivi_dsp: add some missing () to macros
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:f1feb7e388c4: avcodec/mlpdec: fix () in MSB_MASK() macro
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:3065919b6566: avcodec/mss34dsp: fix () in SOP* macros
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:66a9c50d42e0: avcodec/mss4: Fix () in MKVAL() macro
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:4d1346d46acc: avcodec/aic: fix quantization table permutation
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:d192ff0ece61: avformat/flvenc: Do not allow creating h263/mpeg4 in flv without unofficial format extensions being enabled.
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:c2c922460aa4: avcodec/alsdec: Clear MPEG4AudioConfig so that no use of uninitialized memory is possible
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:a45a66599143: avformat/cavsvideodec: Fix probing when the file extension is avs
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:0e3a1d60d393: cavsdec: fix qp fixed slice handling
[17:57] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:9790ed63fcd2: avformat/mpc: attempt to allocate a packet that is not smaller than the data inside it
[17:58] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/2.1:e7873dfccad5: update for FFmpeg 2.1.5
[18:12] <cone-622> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n2.1.5': unknown revision or path not in the working tree.
[18:12] <cone-622> Use '--' to separate paths from revisions
[18:12] <cone-622> refs/tags/n2.1.5:HEAD: Merge commit 'ccda51b14c0fcae2fad73a24872dce75a7964996'
[18:52] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:1be06c853182: swscale/swscale: fix integer overflow
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:92ca3578bf40: avcodec/x86/mpegvideoenc_template: fix integer overflow
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:2af3a0416453: avcodec/h264_mp4toannexb_bsf: prepend global headers before any in stream parameter sets
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:e897eb108b51: avcodec/libx264: move where x264opts is applied down so it isnt overridden by avctx & defaults
[18:53] <cone-622> ffmpeg.git 03Lukasz Marek 07release/1.2:8f7ebb88b816: lavu/opt: validate range before dereference
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:e5546d48717f: avcodec/wma: use av_freep(), do not leave stale pointers in memory
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:a61998e4f4e1: avcodec/msrle: check return code for success before use
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:ce959040135d: swresample/resample: Limit filter length
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:fe97363c38a9: swresample/dither: use av_malloc_array()
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:4065e557593b: swresample/resample: use av_malloc_array() where appropriate
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:4af56488c86e: swscale/x86/swscale_template: loose hardcoded dstw_offset
[18:53] <cone-622> ffmpeg.git 03Anthoine Bourgeois 07release/1.2:90cf672d7a49: avcodec/dirac_arith: Fix build with PIC and stack-check options
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:98fe429ba555: avcodec/x86/idct_sse2_xvid: fix non C99 inline function
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:71104f7cb0fb: swscale/swscale: fix srcStride/srcSlice typo
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:7b38ea4003ed: avcodec/mjpegen: Fix declared argument size
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:424b00a30021: avformat/mpegts: Remove redundant check
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:90c7dfb907ee: avcodec/diracdec: fix undefined behavior with shifts
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:53c02d04e495: avcodec/g723_1: add assert to help static code analyzers
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:8d60d8bef228: avfilter/vf_deshake: fix loss of precission with odd resolutions
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:85b00d798fd1: iavcodec/vc1dec: Fix missing {}
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:753f0738cae9: avfilter/filtfmts: Support dynamically allocated in/outputs
[18:53] <cone-622> ffmpeg.git 03Xi Wang 07release/1.2:9ade5b804ea1: lzo: fix overflow checking in copy_backptr()
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:eada99a854f8: avutil/lzo: Fix integer overflow
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:ecd30280f5ef: avutil/lzo: add asserts to be double sure against overflows
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:c75a915f66f2: avformat/mux: Check for and remove invalid packet durations
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:26d725344155: avformat/h263dec: Fix h263 probe
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:c7e9626b7b59: swresample: fix AV_CH_LAYOUT_STEREO_DOWNMIX input
[18:53] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:3c385ea24cb6: ffmpeg_filter: fix pointer to local outside scope
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:0607c8af2bd1: avcodec/mjpegdec: Fix undefined shift
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:4ecd7182851f: avfilter/graphdump: Fix pointer to local outside scope
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:4b638ea42d31: avcodec/diracdec: move mc buffer allocation to per frame
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:2b8490b762d9: avutil/cpu: force mmx on selection of higher x86 SIMD features
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:5691d9a6c129: avcodec/libvorbisenc: dont add the duration to AV_NOPTS_VALUE
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:b2b4df8d51b3: avcodec/aac: fix () in IS_CODEBOOK_UNSIGNED macro
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:ea7fbd0b1699: avcodec/golomb-test: fix () in EXTEND() macro
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:8d1f38c4b1eb: avcodec/mlpdec: fix () in MSB_MASK() macro
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:5626e11bd553: avcodec/mss34dsp: fix () in SOP* macros
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:d13cf32099fa: avcodec/mss4: Fix () in MKVAL() macro
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:f2aaee49b241: avformat/flvenc: Do not allow creating h263/mpeg4 in flv without unofficial format extensions being enabled.
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:9d77a6293627: avcodec/alsdec: Clear MPEG4AudioConfig so that no use of uninitialized memory is possible
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:c2b72eac6cfe: avformat/mpc: attempt to allocate a packet that is not smaller than the data inside it
[18:54] <cone-622> ffmpeg.git 03Michael Niedermayer 07release/1.2:cc8d20e345e2: Update for FFmpeg 1.2.7
[19:15] <cone-622> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n1.2.7': unknown revision or path not in the working tree.
[19:15] <cone-622> Use '--' to separate paths from revisions
[19:15] <cone-622> refs/tags/n1.2.7:HEAD: Merge commit 'ccda51b14c0fcae2fad73a24872dce75a7964996'
[20:48] <cone-622> ffmpeg.git 03Janne Grunau 07master:f23d26a68641: h264: avoid using uninitialized memory in NEON chroma mc
[20:48] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:32cf26cc6a0d: Merge commit 'f23d26a6864128001b03876b0b92fffe131f2060'
[21:39] <cone-622> ffmpeg.git 03Diego Biurrun 07master:fab9df63a315: dsputil: Split off global motion compensation bits into a separate context
[21:39] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:d7463c681363: Merge commit 'fab9df63a3156ffe1f9490aafaea41e03ef60ddf'
[22:20] <J_Darnley> Does C++ automatically initialize variables to 0?
[22:20] <nevcairiel> no
[22:21] <nevcairiel> unless static
[22:21] <nevcairiel> like C
[22:21] <J_Darnley> Then I wonder what is supposed to be in this float matrix[16]
[22:21] <nevcairiel> maybe its initialized at some point later?
[22:25] <J_Darnley> Oh yes, my bad
[22:25] <J_Darnley> a function call sets values in it.
[22:25] Action: J_Darnley just wrote that!
[22:28] <cone-622> ffmpeg.git 03James Almer 07master:476bd3c7e41f: x86/dsputil: move put_signed_pixels_clamped out of bswapdsp.asm
[23:05] <jamrial> i don't understand that bug. is it a crappy compiler problem or what?
[23:05] <nevcairiel> no idea
[23:05] <nevcairiel> its too much CE in there to read it all
[23:05] <jamrial> the 3dnow functions are not going to be used unless the cpu supports them. it sholdn't matter if the binary has them
[23:06] <nevcairiel> thats what I said in the very first comments in that bug
[23:06] <nevcairiel> but it keeps going and going and going
[23:06] <jamrial> carl's latest patch in that bug is ugly and unnecesary
[23:13] <iive> i'm more curious what sets these flags?
[23:13] <nevcairiel> configure
[23:14] <nevcairiel> apparently he just wants to make --disable-3dnow "work", as in exclude all 3dnow code
[23:14] <nevcairiel> why, is beyond us
[23:18] <iive> if I understand correctly, these flags are in the elf header
[23:19] <iive> the compiler or the linker put them there. my question is... how do they know.
[23:19] <nevcairiel> oh, those flags
[23:19] <nevcairiel> probably the linker, and it scans the objects for instructions
[23:19] <nevcairiel> which seems braindead
[23:19] <nevcairiel> never heard of runtime selection
[23:20] <iive> so, the linker does disassembly?
[23:20] <nevcairiel> well either it sets them just willy-nilly, or it actually finds these instructions in the binary somewhere
[23:21] <iive> also, he complains about .a file, that is just archive with .o files ...
[23:22] <nevcairiel> so the assembler would then set it on the .o file it generates?
[23:22] <nevcairiel> i guess its in a position to know, but still
[23:22] <iive> this looks more plausable.
[23:44] <iive> https://web.archive.org/web/20070210141458/http://blogs.sun.com/alblog/entry/ridding_or_modifying_hardware_capabilities
[23:45] <iive> the above might be useful.
[23:45] <nevcairiel> solaris is all kinds of crap, apparently
[23:47] <iive> well, if `fbe` is the assembler, then adding "-nH" option should clear the hardware flags.
[23:48] <nevcairiel> i figured we would be using some form of yasm or nasm as well on solaris, but what do i know
[23:48] <iive> there is a lot of inline assembly especially for old code.
[23:49] <iive> would you kindly put the above url in the track, so maybe more people could see it.
[23:49] <nevcairiel> hm, if it uses this fbe for some task, might as well try to figure out how to pass the flags
[23:50] <iive> the other solution is using mapfile
[23:51] <cone-622> ffmpeg.git 03Michael Niedermayer 07master:49bf712a8929: libavcodec/wmalosslessdec: add missing emms
[23:54] <iive> https://web.archive.org/web/20081204050526/http://developers.sun.com/solaris/articles/hwcap_modification.html
[23:54] <iive> that's almost the same info.
[00:00] --- Tue Jun 24 2014