[Ffmpeg-devel-irc] ffmpeg-devel.log.20170320

[burek]

[00:24:15 CET] <michaelni> ubitux, iam surprised they are unused but yes if they are you can drop them
[01:34:30 CET] <cone-434> ffmpeg 03Michael Niedermayer 07master:5d996b56499f: avcodec/tiff: Check stripsize strippos for overflow
[01:34:30 CET] <cone-434> ffmpeg 03Michael Niedermayer 07master:a84d610b372c: avcodec/h264_direct: Fix runtime error: signed integer overflow: -9 - 2147483647 cannot be represented in type 'int'
[01:34:30 CET] <cone-434> ffmpeg 03Michael Niedermayer 07master:98da63b3f5f5: avcodec/vp56: Check avctx->error_concealment before enabling EC
[02:39:52 CET] <cone-434> ffmpeg 03Michael Niedermayer 07master:656a17e126c0: avcodec/mjpegdec: Check quant_matrixes values for being non zero
[02:39:53 CET] <cone-434> ffmpeg 03Michael Niedermayer 07master:23f3f92361a3: avcodec/mjpegdec: quant_matrixes can be up to 65535, use uint16_t
[05:35:34 CET] <wm4> is it just me or does git://source.ffmpeg.org/ffmpeg.git not work
[05:35:40 CET] <wm4> fatal: read error: Connection reset by peer
[08:01:47 CET] <ubitux> wm4: works fine here
[08:02:36 CET] <wm4> still doesn't work here
[08:02:55 CET] <wm4> oh on the second and third tries it worked
[08:03:25 CET] <wm4> actually works sporadically
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:08e1376d81f3: fate: add fate-sws-pixdesc-query
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:f052b1b40f94: swscale: use a function for isGray
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:9c2436e1e78f: lavu: add AV_PIX_FMT_FLAG_BAYER
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:c30875e8b2bd: swscale: use a function for isBayer
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:2b9a52bcca7d: swscale: use a function for isAnyRGB
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:ff6bc16c5ad9: swscale: use a (more correct) function for isPacked
[08:06:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:d6635daded80: swscale: remove unused is{RGB,BGR}inBytes
[08:06:34 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e811f84a2ed0: swscale: cosmetics in is{RGB,BGR}inInt
[08:08:43 CET] <cone-785> ffmpeg 03Derek Buitenhuis 07master:eb96505b761e: mov: Remove ancient heuristic hack
[08:08:44 CET] <cone-785> ffmpeg 03Clément BSsch 07master:64722057b41d: Merge commit 'eb96505b761eb02b6a3efc76d854afa6a41941ff'
[08:09:50 CET] <cone-785> ffmpeg 03Derek Buitenhuis 07master:8db804e8f549: mov: Remove old b-frame/video delay heuristic
[08:09:51 CET] <cone-785> ffmpeg 03Clément BSsch 07master:6557d784d21d: Merge commit '8db804e8f549d5b86a1edf62736e0ef80f160da9'
[08:11:34 CET] <cone-785> ffmpeg 03Clément BSsch 07master:5e5e7935523d: doc/APIchanges: fill date & hash for AV_PIX_FMT_FLAG_BAYER
[08:13:37 CET] <cone-785> ffmpeg 03Vittorio Giovara 07master:95f80293456d: avprobe: Fix memory leak
[08:13:38 CET] <cone-785> ffmpeg 03Clément BSsch 07master:b1a80bdb62ae: Merge commit '95f80293456d9d4b1b096621260c38bc90325ec0'
[08:17:33 CET] <cone-785> ffmpeg 03Burt P 07master:728e80cd2e1d: High Definition Compatible Digital (HDCD) decoder filter, using libhdcd
[08:17:34 CET] <cone-785> ffmpeg 03Clément BSsch 07master:45982bdcd0d2: Merge commit '728e80cd2e1d4b7c3e26489efcd77bd7a9e84a99'
[08:19:25 CET] <cone-785> ffmpeg 03Vittorio Giovara 07master:80fc75d51e33: Changelog: Mention mov with multiple stsd
[08:19:26 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e514a1d4047f: Merge commit '80fc75d51e3312e1890591048eb6a3d499b6e49d'
[08:21:38 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:72eba6558ee4: wmavoice: Simplify GetBitContext initialization
[08:21:39 CET] <cone-785> ffmpeg 03Clément BSsch 07master:464fcc979c50: Merge commit '72eba6558ee4f10239ba3f472c0b033ec70082a7'
[08:25:44 CET] <cone-785> ffmpeg 03Mark Thompson 07master:123ccd07c55c: lavc: Rewrite VAAPI decode infrastructure
[08:25:45 CET] <cone-785> ffmpeg 03Mark Thompson 07master:2fe93244ab94: vaapi_h264: Convert to use the new VAAPI hwaccel code
[08:25:46 CET] <cone-785> ffmpeg 03Mark Thompson 07master:102e13c353de: vaapi_mpeg2: Convert to use the new VAAPI hwaccel code
[08:25:47 CET] <cone-785> ffmpeg 03Mark Thompson 07master:520fb77285ff: vaapi_vc1: Convert to use the new VAAPI hwaccel code
[08:25:48 CET] <cone-785> ffmpeg 03Mark Thompson 07master:ccd0316f7cab: vaapi_mpeg4: Convert to use the new VAAPI hwaccel code
[08:25:49 CET] <cone-785> ffmpeg 03Mark Thompson 07master:3e8651a7ccd8: avconv_vaapi: Convert to use hw_frames_ctx only
[08:25:50 CET] <cone-785> ffmpeg 03Mark Thompson 07master:851960f6f8cf: lavc: Remove old vaapi decode infrastructure
[08:25:51 CET] <cone-785> ffmpeg 03Clément BSsch 07master:518961bc99b9: Merge commit '851960f6f8cf1f946fe42fa36cf6598fac68072c'
[08:26:54 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:24da43032473: Changelog: mark the release 12 branch
[08:26:55 CET] <cone-785> ffmpeg 03Clément BSsch 07master:5d2354327725: Merge commit '24da430324735f95880c4a4a54298dc8023125bb'
[08:31:22 CET] <ubitux> nevcairiel: ping
[08:37:12 CET] <ubitux> actually, maybe wm4 knows too: ping :)
[08:37:18 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:d7bc52bf456d: imgutils: add a function for copying image data from GPU mapped memory
[08:37:19 CET] <cone-785> ffmpeg 03Clément BSsch 07master:8200b16a9c77: Merge commit 'd7bc52bf456deba0f32d9fe5c288ec441f1ebef5'
[08:38:01 CET] <ubitux> wm4: the next commit applies cleanly, but we have some differences in dxva
[08:38:14 CET] <ubitux> would you mind merging it and test it?
[08:39:05 CET] <ubitux> michaelni: opinion on the drop of memalign hack?
[08:39:11 CET] <ubitux> (4fb311c804098d78e5ce5f527f9a9c37536d3a08)
[08:39:49 CET] <ubitux> i think it's the followup of 46e3936fb04d06550151e667357065e3f646da1a
[08:44:28 CET] <wm4> ubitux: which hash?
[08:44:51 CET] <ubitux> f01f7a7846529b7c3ef343f117eaa2c0a1457af0
[08:46:52 CET] <wm4> ubitux: I think if it applies cleanly it should be fine
[08:46:58 CET] <ubitux> ok
[08:48:04 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:f01f7a784652: hwcontext_dxva2: use the special UC copy for downloading frames
[08:48:05 CET] <ubitux> thanks
[08:48:05 CET] <cone-785> ffmpeg 03Clément BSsch 07master:a5cf6628d68e: Merge commit 'f01f7a7846529b7c3ef343f117eaa2c0a1457af0'
[08:48:16 CET] <ubitux> alright, now the memalign hack...
[08:48:47 CET] <ubitux> anyone has an opinion on this one?
[08:49:32 CET] <wm4> fuck those hacks with a rusty rake
[08:50:05 CET] <wm4> they're surely not needed anymore, but of course SOMEONE in ffmpeg will argue for keeping them "just in case" or because $FUCKED_UP_OLD_ENVIRONMENT
[08:50:20 CET] <wm4> (if this is what I think it's about)
[08:53:11 CET] <ubitux> ok :)
[08:53:31 CET] <rcombs> does anything actually have neither memalign nor anything like it?
[08:53:54 CET] <ubitux> the mingw case seems to have been avoided by 46e3936fb04d06550151e667357065e3f646da1a
[08:57:22 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:4fb311c80409: Drop memalign hack
[08:57:23 CET] <cone-785> ffmpeg 03Clément BSsch 07master:3835283293bf: Merge commit '4fb311c804098d78e5ce5f527f9a9c37536d3a08'
[09:18:31 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:746c56b7730c: indeo: Change type of array pitch parameters to ptrdiff_t
[09:18:32 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:21e500ba647a: svq1dec: Change type of array pitch parameters to ptrdiff_t
[09:18:33 CET] <cone-785> ffmpeg 03Clément BSsch 07master:bb3ad401fc65: Merge commit '746c56b7730ce09397d3a8354acc131285e9d829'
[09:18:34 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e59d8d030fd4: Merge commit '21e500ba647aec233d5930d3d1081489d0d53ceb'
[09:25:42 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:73f5e17a2037: copy_block: Change type of array stride parameters to ptrdiff_t
[09:25:43 CET] <cone-785> ffmpeg 03Clément BSsch 07master:21c18b087822: Merge commit '73f5e17a203713c4ac4e5a821809823b383b195f'
[09:25:44 CET] <cone-785> ffmpeg 03Clément BSsch 07master:64926292a68a: lavc/copy_block: style fix
[09:26:14 CET] <nevcairiel> ubitux: the dxva changes are indeed fine,  just in case you look for confirmation after the fact =p
[09:26:34 CET] <ubitux> wm4 took responsibility so i'm safe ;)
[09:26:36 CET] <ubitux> thanks :)
[09:26:53 CET] <wm4> sure I'll take any blame
[09:30:32 CET] <ubitux> ah, didn't take long before the memalign thing made someone angry
[09:31:04 CET] <ubitux> michaelni: zhouxiaoyong at loongson.cn doesn't seem to exist anymore; are you still in contact with them?
[09:31:35 CET] <ubitux> i want to poke them about updating the copy_block*_mmi prototypes in libavcodec/mips/h264qpel_mmi.c after the merge of 73f5e17a20
[09:37:34 CET] <wm4> who got angry?
[09:37:52 CET] <ubitux> it was on cvslog, i forwarded
[09:38:27 CET] <wm4> oh.
[09:38:49 CET] <wm4> what next crash is he talking about
[09:38:59 CET] <wm4> and why would freebsd not have a posix function
[09:42:35 CET] <nevcairiel> its carl, he typically doesnt make much sense
[09:44:52 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:5b5ed92d9225: sanm: Change type of array pitch parameters to ptrdiff_t
[09:44:53 CET] <cone-785> ffmpeg 03Clément BSsch 07master:7317b69630de: Merge commit '5b5ed92d92252a685e891a5d636870e223b63228'
[09:46:56 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:2610c9528f86: configure: Move initial VAAPI check to a more sensible place
[09:46:57 CET] <cone-785> ffmpeg 03Clément BSsch 07master:6d6f79c737d4: Merge commit '2610c9528f86286e4c6e174411a26ff5b4815cde'
[09:49:05 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:b8c2d407efa4: configure: Simplify libopenjpeg check
[09:49:07 CET] <cone-785> ffmpeg 03Clément BSsch 07master:715f781834bb: Merge commit 'b8c2d407efa41c3db6813ad67fadd51b814765bd'
[09:49:07 CET] <michaelni> ubitux, why do you ask and then push 4fb311c80409 ? 
[09:49:55 CET] <ubitux> i assumed it wasn't a problem anymore
[09:50:08 CET] <ubitux> it can still be reverted, there is nothing stacked upon
[09:50:20 CET] <ubitux> any reason to keep it?
[09:51:06 CET] <michaelni> whateveraliged() + realloc()/free()/... is supposedly not safe on various places
[09:51:21 CET] <michaelni> some memory debugger at least had an issue
[09:51:25 CET] <durandal_1707> what?
[09:52:53 CET] <michaelni> libav seperates realloc() and malloc() we dont we use the memalign hack to handle these very obscure corner cases IIRC
[09:52:56 CET] <durandal_1707> so it should then be alwaya enabled as we like hacks
[09:54:15 CET] <michaelni> ill leave this to others, iam not interrested in trolling
[09:54:52 CET] <ubitux> michaelni: sorry i don't understand what you're saying wrt realloc
[09:54:59 CET] <durandal_1707> its named hack after all
[09:55:17 CET] <ubitux> michaelni: i'm looking for a concrete example where it breaks
[09:55:26 CET] <nevcairiel> ubitux: apparently on some systems its unsupported to call realloc on a pointer created by memalign
[09:55:35 CET] <nevcairiel> not sure if we actually have any where this applies
[09:55:45 CET] <nevcairiel> since i doubt most people even build with the hack e nabled
[09:56:29 CET] <durandal_1707> i encountered some
[09:56:46 CET] <nevcairiel> the only people you commonly see using the hack are people on windows that cargo-culted the command line from several decades ago
[09:57:09 CET] <ubitux> are we actually supposed to support realloc on memaligned allocated buffer?
[09:57:41 CET] <wm4> <michaelni> ill leave this to others, iam not interrested in trolling <- so you admit arguing against this merge is trolling? wow
[09:57:46 CET] <ubitux>  * @warning Unlike av_malloc(), the returned pointer is not guaranteed to be
[09:57:48 CET] <ubitux>  *          correctly aligned.
[09:57:56 CET] <durandal_1707> when we call memalign?
[09:57:59 CET] <michaelni> ubitux, as it works almost everywhere its very hard to maitain seperation
[09:58:47 CET] <nevcairiel> clearly everyone should be using windows where there is an aligned realloc function </troll
[09:59:16 CET] <michaelni> also specs dont allow memalign() + free(), memalign hack would allow support for such systems too
[09:59:33 CET] <nevcairiel> how are you supposed to free that memory then
[09:59:35 CET] <michaelni> wm4, do not slander me
[10:00:03 CET] <durandal_1707> but hack is not enabled automagically?
[10:00:04 CET] <michaelni> nevcairiel, memalign hack uses malloc()
[10:00:10 CET] <nevcairiel> posix specs do say free should work on memalign'ed memory
[10:01:14 CET] <ubitux> can we add a FATE test doing a bunch of alloc-realloc and assert if misalign to detect such platforms on FATE?
[10:01:32 CET] <michaelni> we could i guess
[10:01:59 CET] <ubitux> i'll revert (or someone else will) if such a case is present
[10:02:30 CET] <durandal_1707> is there FATE instance with this option?
[10:02:41 CET] <michaelni> ubitux,  i doubt we have a fate client affeced
[10:03:08 CET] <wm4> <michaelni> wm4, do not slander me <- uh what
[10:03:28 CET] <michaelni> anyway, asking me, then pushing before i reply and then attacking me when i reply is pretty lame
[10:03:37 CET] <wm4> michaelni: so you didn't mean _you_ are trolling, but others are (including me), so wouldn't this mean that you've slandered me?
[10:04:05 CET] <wm4> but as usual you're wasting our time arguing about unimportant obscurities just because you don't like cleanups
[10:04:27 CET] <durandal_1707> i doubt having options for unreproducible problems is way to go
[10:04:31 CET] <michaelni> wm4, you made a false statement about what i meant and you repeat now other false statements
[10:04:36 CET] <michaelni> wm4 stop this
[10:05:01 CET] <wm4> michaelni: how does "iam not interrested in trolling" not mean "I am not interested in trolling (you)"
[10:05:22 CET] <wm4> michaelni: and the other interpretation, how does it not mean you accuse us of trolling?
[10:05:32 CET] <durandal_1707> i just ignore personal vendetas
[10:05:40 CET] <michaelni> i do accuse some here of trolling  yes
[10:07:39 CET] <wm4> michaelni: ok stop slandering me too, then
[10:08:14 CET] <wm4> always the same shit
[10:11:47 CET] <ubitux> note that i'm ok to revert if it's actually breaking something
[10:12:03 CET] <ubitux> (if that wasn't already obvious)
[10:12:42 CET] <durandal_1707> its already said that it breaks some obscure configuration
[10:13:25 CET] <durandal_1707> "breaks" because option is not auto enabled afaik
[10:15:12 CET] <wm4> still looking for proof that anything was actually broken
[10:46:57 CET] <cone-785> ffmpeg 03Clément BSsch 07master:d0db00c80886: configure: remove pod2man from the config list
[10:47:24 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:0e5dde739943: configure: Fix --disable-pod2man / --disable-texi2html
[10:47:25 CET] <cone-785> ffmpeg 03Clément BSsch 07master:4ae80c375398: Merge commit '0e5dde739943168d6f61d3fb40b3f622e7abfeff'
[10:48:36 CET] <ubitux> what's the pipe for in a make dependency list?
[10:48:45 CET] <ubitux> (refering to 3aa9d37d03da3c9b482d19b3988659287815280e)
[10:49:59 CET] <wbs> ubitux: https://www.gnu.org/software/make/manual/html_node/Prerequisite-Types.html
[10:50:13 CET] <wm4> wbs was faster
[10:50:36 CET] <ubitux> oooh i was wondering how to do that the other day
[10:50:38 CET] <ubitux> thanks!
[10:57:32 CET] <ubitux> i'm not sure we actually need that commit sinces we have a bunch of mkdir in our configure
[10:58:30 CET] <nevcairiel> the commit doesnt exactly elaborate what kind of failure condition it fixes
[10:59:00 CET] <ubitux> it might be useful it cases such as make tests/pixfmts.mak when nothing is generated yet
[10:59:44 CET] <ubitux> i guess it doesn't hurt
[10:59:51 CET] <ubitux> i'll merge it
[11:02:20 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:3aa9d37d03da: build: Fix directory dependencies of tests/pixfmts.mak target
[11:02:21 CET] <cone-785> ffmpeg 03Clément BSsch 07master:38343651a82f: Merge commit '3aa9d37d03da3c9b482d19b3988659287815280e'
[11:06:35 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:ec903058447a: configure: Simplify clock_gettime() test
[11:06:36 CET] <cone-785> ffmpeg 03Clément BSsch 07master:100026bed651: Merge commit 'ec903058447ad5be34d89533962e9ae1aa1c78f7'
[11:16:12 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:6b52762951fa: error_resilience: Change type of array stride parameters to ptrdiff_t
[11:16:13 CET] <cone-785> ffmpeg 03Clément BSsch 07master:d36a42344525: Merge commit '6b52762951fa138eef59e2628dabb389e0500e40'
[11:20:52 CET] <ubitux> that next commit sounds pretty wrong
[11:21:30 CET] <ubitux> he didn't seem to have updated the filter_flt prototype
[11:21:51 CET] <ubitux> (in the context)
[11:22:04 CET] <ubitux> and he a log.h dubious include
[11:23:01 CET] <wbs> ubitux: there's no such context in the code where it came from
[11:23:40 CET] <ubitux> oh indeed, that's only used by the mips code, my bad
[11:23:43 CET] <ubitux> thanks
[11:28:58 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:52730e0f867f: iir_filter: Change type of array stride parameters to ptrdiff_t
[11:28:59 CET] <cone-785> ffmpeg 03Clément BSsch 07master:8316a0e08b89: Merge commit '52730e0f867fe77b7d2353d8b44e92edb7079ca5'
[11:34:48 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:131a85a1fed9: utvideo: Change type of array stride parameters to ptrdiff_t
[11:34:49 CET] <cone-785> ffmpeg 03Clément BSsch 07master:eed8ccde3e56: Merge commit '131a85a1fed9966bbd38517f76abfac0237e39dc'
[11:35:55 CET] <ubitux> jkqxz: do you mind merging your vp8 hwaccel commits?
[11:43:29 CET] <wm4> didn't he post them on the ML?
[11:47:24 CET] <ubitux> ah?
[11:48:13 CET] <ubitux> oh indeed
[11:49:09 CET] <ubitux> i'll skip them then
[11:54:39 CET] <cone-785> ffmpeg 03Mark Thompson 07master:4e528206bc4d: vp8: Add hwaccel hooks
[11:54:40 CET] <cone-785> ffmpeg 03Mark Thompson 07master:a9fb134730da: lavc/vaapi: Add VP8 decode hwaccel
[11:54:41 CET] <cone-785> ffmpeg 03Mark Thompson 07master:11c191b52ce0: vaapi_decode: Ignore the profile when not useful
[11:54:42 CET] <cone-785> ffmpeg 03Mark Thompson 07master:75d642a944d5: vaapi_vp8: Explicitly include libva vp8 decode header
[11:54:43 CET] <cone-785> ffmpeg 03Clément BSsch 07master:9785b1e21be1: Merge commit '75d642a944d5579e4ef20ff3701422a64692afcf'
[11:56:43 CET] <cone-785> ffmpeg 03Luca Barbato 07master:e89cef40506d: checkasm: Read the unsigned value as it should
[11:56:44 CET] <cone-785> ffmpeg 03Clément BSsch 07master:3c8f7a8f6b52: Merge commit 'e89cef40506d990a982aefedfde7d3ca4f88c524'
[11:58:36 CET] <cone-785> ffmpeg 03Luca Barbato 07master:caccb3a0cdc7: audiodsp: ppc: Add VSX variant
[11:58:37 CET] <cone-785> ffmpeg 03Clément BSsch 07master:9e8fd5c423da: Merge commit 'caccb3a0cdc7ee32cbed7eab156d35025133eadc'
[12:02:49 CET] <ubitux> mmh
[12:03:31 CET] <ubitux> so our ppc actually support BE
[12:03:53 CET] <nevcairiel> ppc is primarily be
[12:06:41 CET] <ubitux> sorry, i meant LE
[12:07:27 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:6ce93757ee6b: ppc: Update #endif comments
[12:07:28 CET] <cone-785> ffmpeg 03Clément BSsch 07master:7c54e5870f2c: Merge commit '6ce93757ee6b81fe727bfdc9f546fd0ddf9139c3'
[12:08:39 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:468bfe38c66d: ppc: mpegvideo: Add proper runtime AltiVec detection
[12:08:40 CET] <cone-785> ffmpeg 03Clément BSsch 07master:8e9dfe0d298b: Merge commit '468bfe38c66d4d020984158e53b09a6a5749f394'
[12:23:47 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:ab3554e1a7c0: configure: Drop check_lib()/require() in favor of check_lib2()/require2()
[12:23:48 CET] <cone-785> ffmpeg 03Clément BSsch 07master:4563a86f011b: Merge commit 'ab3554e1a7c04a5ea30f9c905de92348478ef7c8'
[12:43:49 CET] <jkqxz> I did put them on the ML twice, but I didn't really care very much and they got bikeshedded so I gave up.
[13:40:25 CET] <ubitux> jkqxz: so you don't want to handle this with the suggested change from wm4 & michaelni?
[13:47:01 CET] <jkqxz> I don't much like the idea of adding hacks to a clean decoder which does get some use to "fix" a terrible one that is barely used at all and which should never have been written like that in the first place.
[13:48:22 CET] <ubitux> okay
[13:48:37 CET] <ubitux> i won't fix it either, it's in the libav merge doc anyway
[13:54:00 CET] <wm4> is this still about the vp8 hwaccel?
[13:56:22 CET] <ubitux> yes
[13:56:24 CET] <ubitux> vp8/ebp
[13:56:27 CET] <ubitux> webp*
[14:29:47 CET] <feliwir> hey, what is the exact purpose of using those macros: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/get_bits.h#L554
[14:30:10 CET] <feliwir> (why using macros instead of inline functions?)
[14:31:01 CET] <wm4> feliwir: well, some declare variables
[14:31:13 CET] <wm4> how would that work with inline functions?
[14:31:26 CET] <wm4> also, bad compilers
[14:31:39 CET] <feliwir> well not at all, but this is hardly readable/understandable
[14:31:47 CET] <wm4> BBB: any chance you could look at those threading patches I've merged from Libav and posted on the ML?
[14:31:59 CET] <BBB> any specifically?
[14:32:43 CET] <wm4> BBB: hm, all of them I guess
[14:32:50 CET] <atomnuker> feliwir: its a bitstream reader, just write one yourself, it takes like an hour and its fun
[14:33:17 CET] <BBB> wm4: 1/9 is ok
[14:33:21 CET] <feliwir> well the combination with get_vlc2 makes it difficult :D
[14:33:26 CET] <wm4> feliwir: the way it's implemented sure is unreadable, although the functions should be rather easy to use
[14:33:43 CET] <BBB> wm4: 2/9 is ok
[14:34:02 CET] <wm4> BBB: that's a suspiciously quick review
[14:34:17 CET] <BBB> wm4: Ive seen these two before, they were sent by chrome devs
[14:34:27 CET] <wm4> yes, but the ffmpeg code is different
[14:34:38 CET] <BBB> they were initially sent here
[14:34:40 CET] <BBB> theyre fine
[14:34:53 CET] <wm4> eh then I should compare them to the initial patches
[14:34:56 CET] <BBB> I told them to send them to libav because anton was planning to do the same thing and I wanted to prevent the chaos from getting worse
[14:35:07 CET] <wm4> and that was good
[14:35:12 CET] <BBB> (50% chance that he chooses another solution and crap goes to shit even worse)
[14:35:37 CET] <BBB> 3/9 is probably ok but maybe ask btbn or so?
[14:35:59 CET] <BBB> 4/9 same
[14:36:15 CET] <BBB> 5/9 is ok
[14:37:10 CET] <BBB> 6/9 same as 3-4
[14:37:11 CET] <BtbN> I don't think the series affects cuvid/nvenc/cuda at all. As they don't use threads themselves, and are not hooked into external decoders.
[14:38:38 CET] <BBB> 7/9 is probably fine, I am assuming you tested it in the relevant settings with dxva2 to make sure it works?
[14:38:49 CET] <BBB> 8/9 is ok
[14:39:00 CET] <BBB> 9/9 is ok
[14:40:18 CET] <wm4> I didn't test with dxva yet, but with others
[14:40:35 CET] <wm4> and yes, it only affects "classic" hwaccels
[14:41:56 CET] <BBB> I would probably suggest testing it with dxva2 if that is practical
[14:42:16 CET] <wm4> I will
[14:42:31 CET] <BBB> ty
[14:42:33 CET] <BBB> no comments then
[14:42:33 CET] <wm4> I mean, I'm convinced that it works as well or as badly as say vaapi
[14:42:35 CET] <BBB> :)
[14:42:53 CET] <BBB> I guess I should commit the first few gsoc qualification task patches
[14:46:48 CET] <durandal_1707> BBB: which ones?
[14:46:55 CET] <BBB> vp8 tm avx2
[14:47:04 CET] <BBB> and vp9 ddl avx2 16x16
[14:47:06 CET] <RiCON> ubitux: xavs and probably x264 checks are probably broken with 4563a86f011
[14:47:09 CET] <BBB> vp8 tm avx2 is also 16x16
[14:47:55 CET] <ubitux> RiCON: broken how?
[14:48:09 CET] <ubitux> did i derped the merge or the original commit is problematic?
[14:48:14 CET] <RiCON> they need stdint.h before {xavs,x264}.h
[14:48:28 CET] <jamrial> don't those use pkgconfig?
[14:48:33 CET] <RiCON> x264 does
[14:48:35 CET] <RiCON> xavs doesn't
[14:48:35 CET] <jamrial> and not check_lib or require
[14:48:38 CET] <jamrial> ah
[14:48:52 CET] <RiCON> and if there's no pkgconfig x264 will probably fail
[14:49:01 CET] <RiCON> since it's also missing stdint.h
[14:51:31 CET] <ubitux> i don't see how that's a problem
[14:51:50 CET] <ubitux> i didn't drop the stdint.h
[14:52:27 CET] <jamrial> just tried wihtout pkgconfig. stdint.h is included but after x264.h so it breaks
[14:53:19 CET] <jamrial> god, configure is painfully slow on windows
[14:53:26 CET] <cone-785> ffmpeg 03Ronald S. Bultje 07master:f3cd2302a9c9: wmavoice: remove unused or write-only variables.
[14:53:27 CET] <cone-785> ffmpeg 03Mirage Abeysekara 07master:5eb4f95bef2f: h264pred: added AVX2 implementation for tm_vp8 16x16.
[14:53:28 CET] <cone-785> ffmpeg 03Ilia 07master:2f3d10a01ac5: avcodec/vp9: avx2 implementation of ipred_dl_16x16_16
[14:54:20 CET] <ubitux> jamrial: but what's different with previously?
[14:54:42 CET] <DHE> can confirm x264 now fails without pkg-config whereas it did work before
[14:54:45 CET] <jamrial> no idea if it's different. always used pkgconfig
[14:54:48 CET] <DHE> (with warning)
[14:55:24 CET] <jamrial> i'm trying a fix
[14:55:35 CET] <ubitux> adding stdint.h to the require?
[14:55:35 CET] <RiCON> couldn't the stdint.h always be before the added headers?
[14:56:05 CET] <RiCON> or would that break if the headers then add stdint.h again?
[14:56:15 CET] <ubitux> i understand it's broken now, but i don't understand how it's a regression
[14:56:16 CET] <jamrial> http://pastebin.com/raw/JZK3ni1r seems to fix it
[14:56:37 CET] <ubitux> jamrial: yes, exactly my thought; but can you confirm it worked before the merge?
[14:56:49 CET] <jamrial> DHE did
[14:57:45 CET] <ubitux> i'm actually confused how it could work
[14:57:53 CET] <DHE> quick fix: http://pastebin.com/FPYTqhjQ
[14:58:02 CET] <DHE> not sure HOW it broke, but it did
[15:00:01 CET] <RiCON> 'check_header $header && check_func $func "$@"' does not do the same as 'check_func_headers "$headers" "$funcs" "$@"'
[15:04:22 CET] <ubitux> olol check_func_headers has a stdint.h
[15:04:37 CET] <RiCON> i can confirm http://sprunge.us/FTJV fixes it
[15:05:39 CET] <ubitux> can you make a proper patch, including the regression hash in the description and an explanation about check_func_headers having stdint.h?
[15:05:52 CET] <ubitux> and apply or share the patch
[15:07:27 CET] <jamrial> http://pastebin.com/raw/3XUG8Pqq completely different checks
[15:07:53 CET] <RiCON> oh, libav's xavs check does have stdint.h
[15:08:05 CET] <jamrial> before this merge the check just made sure x264.h existed, but didn't actually try to use it
[15:08:59 CET] <RiCON> ubitux: 20abcaa273a6 will fix it too
[15:09:23 CET] <ubitux> ah sorry i missed that
[15:09:30 CET] <ubitux> care to format your patch with both anyway?
[15:13:23 CET] <rounaq> Hi. Sorry to interrupt an ongoing conversation. I am new to audio/video formats and other related areas. Is there a good reference to study all these terms so that I can understand the terms like "square decoder" etc.
[15:15:36 CET] <durandal_1707> rounaq: if this is about gsoc,  too late, another student completed that qualification task
[15:16:00 CET] <RiCON> ubitux: http://sprunge.us/RYPQ how's that?
[15:16:43 CET] <rounaq> Okay. No problem. But I would like to learn anyway. Thanks
[15:17:03 CET] <ubitux> RiCON: LGTM; do you have push access?
[15:17:07 CET] <RiCON> no
[15:17:12 CET] <ubitux> ok will apply in a moment
[15:19:03 CET] <cone-785> ffmpeg 03Ricardo Constantino 07master:20c4fb2e010f: configure: add stdint.h to x264 and xavs checks
[15:19:06 CET] <ubitux> RiCON: thanks
[15:37:25 CET] <cone-785> ffmpeg 03Paul B Mahol 07master:ce818d90bdb2: avcodec/wmaprodec: reset offsets when error happens
[15:56:01 CET] <ubitux> please don't push anything for the next minutes, i'm rebasing a merge
[15:59:55 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:de452e503734: pixblockdsp: Change type of stride parameters to ptrdiff_t
[15:59:56 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e07fa3008bca: Merge commit 'de452e503734ebb0fdbce86e9d16693b3530fad3'
[16:00:03 CET] <ubitux> i hope i didn't break anything in that one
[16:05:30 CET] <ubitux> jkqxz: any reason 09a145b3c837273b1379321e44386a3233156e75 wasn't cherry-picked? it applies cleanly, i just deleted the #undef we had
[16:41:23 CET] <feliwir> atomnuker, another thing i don't understand: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/huffman.c#L165 This is the only line where i see the symbols getting set. I thought the according symbols would depend on the bitstream, but there it is just set to i
[16:41:25 CET] <jkqxz> ubitux:  Don't think so.  Probably missed because it just fixes a warning, hence meh?
[16:41:47 CET] <ubitux> ok :)
[16:42:00 CET] <cone-785> ffmpeg 03Mark Thompson 07master:09a145b3c837: hwcontext_vdpau: Remove duplicate definition of GET_CALLBACK
[16:42:01 CET] <cone-785> ffmpeg 03Mark Thompson 07master:7081620aca36: hwcontext_vdpau: Fix missing subscripts
[16:42:02 CET] <cone-785> ffmpeg 03Clément BSsch 07master:2feef75cb5e6: Merge commit '09a145b3c837273b1379321e44386a3233156e75'
[16:42:03 CET] <cone-785> ffmpeg 03Clément BSsch 07master:6c3d2dad9e29: Merge commit '7081620aca36e616ea96f71fd71d2703e3abae09'
[16:43:08 CET] <cone-785> ffmpeg 03Mark Thompson 07master:3a9662af6c74: vaapi_h264: Fix HRD bit_rate/cpb_size scaling
[16:43:09 CET] <cone-785> ffmpeg 03Clément BSsch 07master:9a0f91314adc: Merge commit '3a9662af6c741f8354b1ca97642f78f5c02e2e8f'
[16:44:16 CET] <ubitux> jkqxz: how about bdf7610eb266fd3de650040c97328791868abd82?
[16:45:00 CET] <ubitux> (it applies cleanly)
[16:45:23 CET] <jkqxz> That's not in ffmpeg already?  Oops.  Yeah, apply that.
[16:45:45 CET] <ubitux> ok :)
[16:46:11 CET] <cone-785> ffmpeg 03Mark Thompson 07master:bdf7610eb266: vf_scale_vaapi: Crop input surface to active region
[16:46:12 CET] <cone-785> ffmpeg 03Clément BSsch 07master:bc1023eb36c6: Merge commit 'bdf7610eb266fd3de650040c97328791868abd82'
[16:47:52 CET] <cone-785> ffmpeg 03Martin Storsjö 07master:df3795025337: rtsp: Fix a crash with the RTSP muxer
[16:47:53 CET] <cone-785> ffmpeg 03Clément BSsch 07master:48cc083a301f: Merge commit 'df3795025337479a639cb3cd26c93a4e82ccd4db'
[16:53:07 CET] <RiCON> wbs: any opinion on something like https://i.fsbn.eu/mipK.txt ?
[16:54:11 CET] <cone-785> ffmpeg 03Josh de Kock 07master:bc7399934def: libdc1394: Distinguish between enumeration errors and no cameras found
[16:54:12 CET] <cone-785> ffmpeg 03Clément BSsch 07master:f91bf71d69d9: Merge commit 'bc7399934def210c2a84ea51375d50f79c676c96'
[17:19:17 CET] <durandal_1707> how much those recent asm optimizations helps overall performance?
[17:22:31 CET] <kierank> durandal_1707: for h264?
[17:23:17 CET] <durandal_1707> and others
[17:24:34 CET] <kierank> i dunno what asm you talk about then
[17:25:59 CET] <durandal_1707> those that were gsoc task
[17:27:17 CET] <nevcairiel> Probably minimal improvements from single functions being optimized
[17:27:29 CET] <nevcairiel> But a lot of small things...
[17:32:56 CET] <jamrial> h264 and vp8 barely make a modern cpu sweat, so no real noticeable difference
[17:37:27 CET] <durandal_1707> and vp9 and hevc?
[17:37:52 CET] <jamrial> avx2 on those may help with 4k
[17:38:06 CET] <jamrial> in any case, last time i checked perf for vp9 showed that half the time spent decoding was in decode_coeff()
[17:39:10 CET] <tdjones> The preprocessed input from the aac psychoacoustic model is saved in the context's 'planar_samples' member, and the values are copied and used in the psy.model->window() for the current frame and look ahead samples. Where would be a good place to find these values within the vorbis context? The vorbis encoder does not copy samples frame to frame in the same way that aac enc does.
[17:40:26 CET] <tdjones> That is for the qualification task with the vorbis encoder, detecting transients using the aac psychoacoustic model
[17:40:50 CET] <durandal_1707> atomnuker: ^
[17:48:48 CET] <thebombzen> Does anyone know why nutenc defaults to mpeg4 for video codecs?
[17:49:05 CET] <thebombzen> I can't see how anyone muxing a nut file would want that
[17:49:20 CET] <cone-785> ffmpeg 03Matthieu Bouron 07master:d839c4716cdc: configure: error out if jni is enabled and cannot be found
[17:52:35 CET] <atomnuker> tdjones: there isn't one, just introduce one in the context which points to the AVFrame's extended_data planes and feed that in
[17:52:46 CET] <atomnuker> (into s->psy.model->window that is)
[17:53:09 CET] <atomnuker> if you want to you can just make it like the AAC encoder and make it an array instead of a pointer
[17:53:38 CET] <atomnuker> AAC did it this way since there was some time domain stuff going on
[17:54:37 CET] <tdjones> atomnuker: thanks, I'll try that
[18:29:33 CET] <cone-785> ffmpeg 03Diego Biurrun 07master:8c201dde0ab6: build: doc: more fine-grained dependencies for generated texi files
[18:29:34 CET] <cone-785> ffmpeg 03Clément BSsch 07master:465a7a1b9f03: Merge commit '8c201dde0ab62e5cd581d958e78d7609e0ba710d'
[18:30:31 CET] <cone-785> ffmpeg 03Janne Grunau 07master:15fcf6292ed7: build: remove hardcoded name of version header
[18:30:32 CET] <cone-785> ffmpeg 03Clément BSsch 07master:6d43533286fa: Merge commit '15fcf6292ed79be274c824fedb099c2665f4cc15'
[18:34:24 CET] <cone-785> ffmpeg 03Michael Niedermayer 07master:136f55207521: mpegvideo_motion: Handle edge emulation even without unrestricted_mv
[18:34:25 CET] <cone-785> ffmpeg 03Clément BSsch 07master:71d3d96c9f69: Merge commit '136f55207521f0b03194ef5b55ba70f1635d6aee'
[18:34:44 CET] <gh0st__> durandal_1707: I made the ipred_dl_16x16_16 avx2 implmentation. IIRC Amdahl's law can be applied here, e.g if the C version took 1% of the cpu time and the avx2 is 10 times faster then theoretically the overall performance of vp9 increases by 0.9%.
[18:36:45 CET] <durandal_1707> my cpu doesnt  have avx2 :(
[18:36:49 CET] <gh0st__> durandal_1707: But real world tests are needed to get the practical performance impact.
[18:37:14 CET] <cone-785> ffmpeg 03Yogender Gupta 07master:de64dd13cbd4: avcodec: Add the extended pixel format profile for HEVC
[18:37:15 CET] <cone-785> ffmpeg 03Clément BSsch 07master:37cf0d0bbfa0: Merge commit 'de64dd13cbd47fd54334b6aa2a2cd3c7c36daae2'
[18:37:51 CET] <gh0st__> durandal_1707: I can give you access to a broadwell machine.
[18:38:37 CET] <gh0st__> If you want to run some tests.
[18:38:41 CET] <cone-785> ffmpeg 03Alexandra Hájková 07master:07e1f99a1bb4: x86util: Document SBUTTERFLY macro
[18:38:42 CET] <cone-785> ffmpeg 03Clément BSsch 07master:3898e346b335: Merge commit '07e1f99a1bb41d1a615676140eefc85cf69fa793'
[18:39:39 CET] <durandal_1707> gh0st__: not needed
[18:43:17 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:1d6c76e11feb: audiodsp/x86: fix ff_vector_clip_int32_sse2
[18:43:18 CET] <cone-785> ffmpeg 03Clément BSsch 07master:072fad7cf536: Merge commit '1d6c76e11febb58738c9647c47079d02b5e10094'
[18:46:08 CET] <nevcairiel> if functions achieve the expected optimizations, then its fine either way (ie. 2x the speed as 128-bit SSE2/3/.. would be ideal for AVX2, so getting somewhere close to that is fine), of course optimizing single functions will not make the entire encoder speed up substantially at that point, but it all counts
[18:46:24 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:75d98e30afab: audiodsp/x86: clear the high bits of the order parameter on 64bit
[18:46:25 CET] <cone-785> ffmpeg 03Clément BSsch 07master:43a4c729d4f2: Merge commit '75d98e30afab61542faab3c0f11880834653bd6b'
[18:47:23 CET] <nevcairiel> (of course not all functions can linearly scale with register size increase)
[18:48:03 CET] <ubitux> philipl: what should we do about 340f12f71207513672b5165d810cb6c8622c6b21?
[18:48:25 CET] <atomnuker> it doesn't help how avx2 is implemented as well
[18:48:39 CET] <ubitux> it looks like we should noop it since they use AV_PIX_FMT_YUV444P16 because they lack P016
[18:48:43 CET] <nevcairiel> the vp8 function from above is even faster then 2x though, while the vp8 function didnt achieve 2x
[18:48:53 CET] <nevcairiel> second vp8 should be vp9
[18:48:54 CET] <nevcairiel> :D
[18:49:08 CET] <ubitux> philipl: but maybe we do need AV_PIX_FMT_YUV444P16 as well?
[18:52:16 CET] <BtbN> Adding AV_PIX_FMT_YUV444P16 won't hurt, but doesn't gain anything.
[18:53:08 CET] <BtbN> it's not a straight forward merge though, as the logic in ffmpeg is slightly different, because of alignment
[18:53:34 CET] <ubitux> yes, and i can't test it
[18:53:47 CET] <ubitux> it's blocking my merge so anyone can take over for now
[18:53:48 CET] <BtbN> I'd say just noop it. If we'll ever need it, it will be added in turn.
[18:53:59 CET] <ubitux> ok
[18:55:25 CET] <cone-785> ffmpeg 03Yogender Kumar Gupta 07master:340f12f71207: hwcontext_cuda: Add P010 and YUV444P16 pixel format
[18:55:26 CET] <cone-785> ffmpeg 03Clément BSsch 07master:87007ebc1688: Merge commit '340f12f71207513672b5165d810cb6c8622c6b21'
[19:01:51 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:eea9857bfd69: blockdsp: drop the high_bit_depth parameter
[19:01:52 CET] <cone-785> ffmpeg 03Clément BSsch 07master:9010676ea324: Merge commit 'eea9857bfd6925d0c34382c00b971ee6df12ad44'
[19:01:53 CET] <cone-785> ffmpeg 03Clément BSsch 07master:b78243c50401: lavc/arm: fix indent in blockdsp_init_neon
[19:07:53 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:2eb97af66af9: checkasm: add a test for blockdsp
[19:07:54 CET] <cone-785> ffmpeg 03Clément BSsch 07master:c50b2164a6cd: Merge commit '2eb97af66af90ca3978229da151f0b8b3a5d9370'
[19:11:28 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:e9ef6171396d: checkasm: add tests for audiodsp
[19:11:29 CET] <cone-785> ffmpeg 03Clément BSsch 07master:84147554865b: Merge commit 'e9ef6171396dc4106526aaa86b620c61ca3d1017'
[19:11:44 CET] <BtbN> Wasn't there some new avx2 asm recently? I'd like to bench it.
[19:12:08 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:bf58545aace7: audiodsp: fix vector_clipf documentation
[19:12:09 CET] <cone-785> ffmpeg 03Clément BSsch 07master:90f6433dcf37: Merge commit 'bf58545aace7d14522ce4fa680c7b3ff62109a3a'
[19:16:48 CET] <BtbN> ubitux, I just got an E-Mail from Github, because there is @BtbN in the commit message. :D
[19:17:03 CET] <ubitux> :D
[19:17:11 CET] <ubitux> sorry ;)
[19:17:18 CET] <JEEB> :D
[19:22:59 CET] <ubitux> jamrial: opinion on 12004a9a7f20e44f4da2ee6c372d5e1794c8d6c5?
[19:23:11 CET] <ubitux> it was apparently yasmified differently
[19:23:42 CET] <ubitux> (it depends on the previous commit arg shuffle)
[19:24:25 CET] <ubitux> i don't see the magic movsxdifnidn in your port btw
[19:24:55 CET] <ubitux> it's a port from 1d36defe94c7d7ebf995d4dbb4f878d06272f9c6
[19:25:07 CET] <jamrial> seems better thanks to said arg shuffle
[19:25:18 CET] <jamrial> the loop is the same, but init is simpler
[19:25:43 CET] <jamrial> probably worth merging
[19:25:52 CET] <ubitux> OK
[19:25:56 CET] <ubitux> thanks
[19:26:05 CET] <jamrial> curious they didn't make len ptrdiff_t after the previous bunch of commits, heh
[19:26:34 CET] <ubitux> yeah indeed
[19:27:04 CET] <ubitux> maybe it was a pain due to the other arch
[19:27:09 CET] <ubitux> +s
[19:46:52 CET] <cone-785> ffmpeg 03Clément BSsch 07master:bbc3bde14f14: configure: fix crystalhd detection
[19:49:34 CET] <JEEB> crystalhd... now even the people who used to maintain it have their HW dying
[19:53:05 CET] <wbs> RiCON: probably ok I guess, I don't know the swfverify stuff very well
[19:54:07 CET] <kierank> will avfilter_graph_free free any internally buffered frames?
[20:04:38 CET] <durandal_1707> michaelni: can you explain dnxhr changes to parser?
[20:28:53 CET] <kierank> durandal_1707: i have memory leaks in drawtext, who do I ask if my fix is correct?
[20:31:34 CET] <michaelni> durandal_1707, i assume you mean ed0dc14ebba1d2e65fe39b7e8c10359adbfba36f ? if so the author probably would be best in explaining the code
[20:32:05 CET] <durandal_1707> kierank: your code?
[20:32:16 CET] <kierank> in the filter itself
[20:32:26 CET] <durandal_1707> hmm
[20:33:09 CET] <kierank> i also have general big memory leaks in libavfilter which need fixing but I will fix simple stuff first
[20:33:10 CET] <durandal_1707> kierank: valgrind report?
[20:33:41 CET] <kierank> i have a patch coming
[20:33:46 CET] <kierank> but I dunno who to ask if correct
[20:34:21 CET] <kierank> i have lots of valgrind reports with lots of avfilter leaks :(
[20:34:23 CET] <kierank> but I use old ffmpeg
[20:34:25 CET] <kierank> from 2015
[20:34:28 CET] <durandal_1707> ask Carl or Nicolas
[20:35:05 CET] <nevcairiel> lavfi was changed quite a lot s ince 2015
[20:35:15 CET] <kierank> the bug is still present
[20:35:25 CET] <kierank> but it's a minor bug compared to the megabytes of frames leaks i have
[20:35:38 CET] <durandal_1707> have way to reproduce?
[20:35:55 CET] <kierank> not easily, will spend tonight trying to fix
[20:36:37 CET] <durandal_1707> if it leaks mbs than you are not freeing refcounted frames
[20:37:05 CET] <kierank> I am freeing every frame as soon as it leaves lavfi actually
[20:37:12 CET] <kierank> in my debug code
[20:38:46 CET] <durandal_1707> hard to guess without looking at code
[20:39:39 CET] <kierank> I will start sending patches now, don't worry
[20:39:48 CET] <kierank> but i regret using lavfi for anything serious now
[20:49:32 CET] <michaelni> durandal_1707, i assume you plan to work on the dnxhd parser consulting job ? 
[20:52:04 CET] <durandal_1707> michaelni: no if you want to do it
[20:53:37 CET] <michaelni> iam happy to leave it to you, i dont remember the code it just feels like not a impossible bug to fix
[20:55:18 CET] <michaelni> also i have lots of other things to do
[21:01:48 CET] <ubitux> kierank: a better $subj would be "lavfi/drawtext: fix a_pexpr memory leak" though (or s/lavfi/avfilter/ depending on your preference)
[21:02:14 CET] <durandal_1707> michaelni: looks like best way is to pick compressed frame size from cid table
[21:06:28 CET] <durandal_1707> but that doesnt fix issues with variable stuff
[21:11:04 CET] <michaelni> Is that code actually skiping over the compressed frame size ? if not that should improve detection of the next header
[21:15:33 CET] <atana> michaelni, repo updated. 
[21:18:01 CET] <michaelni> atana,  memmove(p->data, p->data, SIZECHECK/2); moves data from p->data to p->data thats exactly where it was, you meant  memmove(p->data, p->data + p->windowSize/2 ...
[21:27:24 CET] <durandal_1707> michaelni: yes, current dnxhd parser code is very fragile
[21:27:50 CET] <michaelni> atana, also the size in memmove should consider the element/sample size
[21:33:37 CET] <hemalpatil> Hey everyone! New user here
[21:33:47 CET] <hemalpatil> Any GSoC 2017 applicants?
[21:36:28 CET] <Rathann> hemalpatil: welcome
[21:38:10 CET] <hemalpatil> Hi
[21:38:36 CET] <hemalpatil> has anyone taken up the Ambisonic decoder project?
[21:38:37 CET] <feliwir> can someone explain me why the symbols for the huffman tree are just set to i here: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/huffman.c#L165 ?
[21:38:40 CET] <feliwir> it makes no sense to me
[21:41:33 CET] <atana> michaelni, code pushed
[21:41:50 CET] <durandal_1707> hemalpatil: yes its taken
[21:42:22 CET] <hemalpatil> any open mentored projects that I can take up for GSoC?
[21:42:52 CET] <atomnuker1> feliwir: pengvado wrote that code but I doubt he'll respond
[21:47:10 CET] <michaelni> atana, does it produce a better score now ?
[21:47:53 CET] <atana> michaelni, no, now there are 114 failures
[21:49:14 CET] <feliwir> atomnuker1, is he not active anymore?
[21:57:07 CET] <michaelni> atana, sizeof(p->data) in the memmove should be  sizeof(*p->data), its the size of a sample not the pointer
[21:58:47 CET] <ubitux> dammit that DASH patch is huuuuge
[22:01:26 CET] <hemalpatil> @durandal_1707: can the "Ambisonic decoder" project still be taken up?
[22:01:43 CET] <durandal_1707> hemalpatil: nope
[22:01:51 CET] <atana> michaelni, corrected and code pushed. still 114 failures
[22:10:14 CET] <michaelni> atana, thats odd i have fewer failures here
[22:12:43 CET] <atana> michaelni, what's the failure count at your side?
[22:12:50 CET] <michaelni> 4
[22:13:28 CET] <atana> let me check
[22:18:17 CET] <atana> I think I found it. For debugging I made some changes in test scripting and was printing file names that might got included in the count. checking now..
[22:20:16 CET] <atana> michaelni, yes it's 4 
[22:23:06 CET] <michaelni> atana, very good !
[22:23:44 CET] <michaelni> atana, you can try if any other window works better
[22:25:14 CET] <atana> Was thinking about it. I will try with diff window sizes also I should replace SIZECHECK with p->windowSize in the filter_frame code
[22:25:36 CET] <michaelni> yes
[22:27:34 CET] <hemalpatil> @BBB: mentor for "VMAF video filter". I want to take up this project
[22:35:34 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:683da86aabb4: audiodsp: reorder arguments for vector_clipf
[22:35:35 CET] <cone-785> ffmpeg 03Anton Khirnov 07master:12004a9a7f20: audiodsp/x86: yasmify vector_clipf_sse
[22:35:36 CET] <cone-785> ffmpeg 03Clément BSsch 07master:83cd80d10aeb: Merge commit '12004a9a7f20e44f4da2ee6c372d5e1794c8d6c5'
[22:38:39 CET] <cone-785> ffmpeg 03Luca Barbato 07master:352741b5ead1: nvenc: Make sure that enum and array index match
[22:38:40 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e3ee67c85e69: Merge commit '352741b5ead1543d775ccf6040f33023e4491186'
[22:39:51 CET] <ubitux> BtbN: any reason e02e2515b24bfc37ede6ca1744696230be55e50b wasn't merged completely?
[22:40:27 CET] <ubitux> ah that's actually present in the other nvenc files, my bad
[22:40:42 CET] <BBB> hemalpatil: ok& we need to find you a suitable qualification task
[22:40:53 CET] <BBB> where is thilo
[22:44:12 CET] <cone-785> ffmpeg 03Yogender Gupta 07master:e02e2515b24b: nvenc: Add some easier to understand presets that match x264 terminology
[22:44:13 CET] <cone-785> ffmpeg 03Clément BSsch 07master:a5dea2b77305: Merge commit 'e02e2515b24bfc37ede6ca1744696230be55e50b'
[22:46:55 CET] <ubitux> i see checks in 358c887a9fa0fb2e7ce089eaea71ab924a3e47a7 that are not present in our version
[22:47:07 CET] <ubitux> anyone with a nvidia up to the merges please?
[22:47:13 CET] <ubitux> i can't test that
[22:49:24 CET] <jamrial> ubitux: you should probably just noop it. we already seem to have high bit support in nvenc
[22:49:44 CET] <jamrial> also, libav supports old sdks while we only support the latest i think
[22:50:01 CET] <ubitux> we don't want to support the old ones?
[22:50:17 CET] <jamrial> don't think so. we're shipping the header in the compat folder after all
[22:50:28 CET] <nevcairiel> indeed, no reason to bother
[22:50:31 CET] <jamrial> and doing dynlink
[22:52:13 CET] <ubitux> OK
[22:53:08 CET] <ubitux> there are a few other differences though
[22:53:33 CET] <ubitux> i'll reduce the cosmetics
[22:54:20 CET] <ubitux> or maybe not..
[22:57:25 CET] <jamrial> IMO, just noop them and ask philipl_ or timo if these are useful/needed. they can be adapted for our tree later like with other commits before
[22:57:43 CET] <cone-785> ffmpeg 03Yogender Gupta 07master:358c887a9fa0: nvenc: Add support for high bitdepth
[22:57:44 CET] <cone-785> ffmpeg 03Yogender Gupta 07master:70de2ea4261f: nvenc: Extended rate-control support as provided by SDK 7
[22:57:45 CET] <cone-785> ffmpeg 03Clément BSsch 07master:99d081e6380e: Merge commit '358c887a9fa0fb2e7ce089eaea71ab924a3e47a7'
[22:57:46 CET] <cone-785> ffmpeg 03Clément BSsch 07master:e849296d0a68: Merge commit '70de2ea4261f860457a04e3d0c58c5543f403325'
[22:59:44 CET] <durandal_1707> BBB: i think its over for quallis, we just pick students now
[23:00:03 CET] <BBB> what if new students come in?
[23:02:21 CET] <nevcairiel> there is still a bit time until april 3, but if you have promising candidates already, might as well tell students that, can only have one per task afterall, and probably only a small subset overall
[23:05:09 CET] <cone-785> ffmpeg 03Clément BSsch 07master:b7cc4eb3030b: lavc/nvenc: misc cosmetics to reduce diff with Libav
[23:09:50 CET] <ubitux> tomorrow, swscale merges, yeeey ~ T_T
[23:14:42 CET] <ubitux> ah it's actually stuff we have in ffmpeg since 2012
[23:15:15 CET] <ubitux> i guess i'll be able to continue my noop party
[23:16:54 CET] <atana> michaelni,  size->failures : 1024->4, 2048->138, 3072->143, 4096->4, 5120->147, 6144->148, 7168->148, 8192->145 
[23:20:57 CET] <michaelni> atana, i have 7 failures with 2048, did you update all 1024 and 512 ?
[23:21:32 CET] <michaelni> atana, also only power of 2 will work probably that is 1024, 2048, 4096, 8192, ...
[23:24:41 CET] <atana> michaelni,  does 'lim = 512' also needs to be update? I have updated others before testing
[23:33:45 CET] <michaelni> atana, did you update SIZECHECK and the default window size in the AVOption table ?
[23:34:18 CET] <michaelni> atana, also p->index = 512;
[23:34:30 CET] <michaelni> lim = 512 isnt used i think currently
[23:35:13 CET] <atana> michaelni,  lim += 512
[23:35:21 CET] <michaelni> yes
[23:35:26 CET] <atana> ok
[23:52:21 CET] <stevenliu> Thanks guys
[23:52:59 CET] <stevenliu> i will merge all the review comment :)
[00:00:00 CET] --- Tue Mar 21 2017