[Ffmpeg-devel-irc] ffmpeg-devel.log.20150531

[burek]

[00:12:59 CEST] <cone-766> ffmpeg 03Luca Barbato 07release/2.4:70642090960c: msrle: Use FFABS to determine the frame size in msrle_decode_pal4
[00:13:00 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07release/2.4:2c180cfc5dd5: Merge commit '70642090960c35dcd6da941c869bdf55d4f3bb00' into release/2.4
[00:21:16 CEST] <cone-766> ffmpeg 03Luca Barbato 07release/2.4:3b69f245dbe6: h264: Make sure reinit failures mark the context as not initialized
[00:21:17 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07release/2.4:b50fa26885dc: Merge commit '3b69f245dbe6e2016659a45c4bfe284f6c5ac57e' into release/2.4
[00:45:45 CEST] <Timothy_Gu> rcombs: your securetransport patch seems to break iOS: http://fatebeta.ffmpeg.org/report/arm64-darwin-clang-apple-5.1/20150530030336
[00:56:58 CEST] <Timothy_Gu> rcombs: also two bogus warnings: diff -u0 <(curl -s http://fatebeta.ffmpeg.org/log/i386-darwin-gcc-4.8/20150529
[00:57:01 CEST] <Timothy_Gu> 141247/compile | grep -i warning | sort) <(curl -s http://fatebeta.ffmpeg.org/log/i386-darwin-gcc-4.8/20150530134417/compile | grep -i warning | sort )
[00:57:40 CEST] <Timothy_Gu> let me try again: diff -u0 <(curl -s http://fatebeta.ffmpeg.org/log/i386-darwin-gcc-4.8/20150529141247/compile | grep -i warning | sort) <(curl -s http://fatebeta.ffmpeg.org/log/i386-darwin-gcc-4.8/20150530134417/compile | grep -i warning | sort)
[01:21:43 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07master:ebe919cce2f4: postproc/postprocess_template: Compute packedYScale and QPCorrecture without floats
[01:42:28 CEST] <rcombs> Timothy_Gu: hmmmmmm
[01:42:52 CEST] <rcombs> Timothy_Gu: wouldn't hurt to zero-initialize those `size_t`s, I suppose
[01:43:53 CEST] <rcombs> Timothy_Gu: but as for the SecItemImport and related type issues, things are a bit trickier; it looks like neither that API nor any equivalent exist on iOS
[01:45:30 CEST] <rcombs> so, for now, I can add SecItemImport to the configure check, or add a separate check for SecItemImport and just return an error if the relevant options are used when Secure Transport is enabled and it's missing
[01:47:03 CEST] <rcombs> for the real fix, I could add back the AVIO PEM parser I had there originally before I found out about SecItemImport
[01:47:53 CEST] <rcombs> or I could add an AV_OPT_TYPE_BINARY for the raw key/cert data, which might be more useful on iOS anyway
[02:11:37 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07master:6f51674c91cf: postproc: Avoid floats in maxClipped computation
[03:17:56 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07master:11aa050a2516: avcodec/parser: Print an error in case of reallocation fails in ff_combine_frame()
[04:47:01 CEST] <cone-766> ffmpeg 03Michael Niedermayer 07master:03637762d94c: avfilter/vf_fade: Force alpha pixel format if alpha is set to 1
[04:57:21 CEST] <Timothy_Gu> rcombs: in any case, the master should always be green, so I'd just add a configure check for now
[05:10:35 CEST] <rehunted> does git clone --depth 1 git://source.ffmpeg.org/ffmpeg is the stable branch ? 
[05:15:29 CEST] <rehunted> I guess I will be fine with git checkout af5917698bd44f136fd0ff00a9e5f8b5f92f2d58 (2.6)
[05:16:01 CEST] Action: rehunted no reference.. changing dockerfile to use https://github.com/FFmpeg/FFmpeg/archive/n2.6.3.tar.gz 
[05:43:30 CEST] <rehunted> so hard to get --enable-pic working
[05:44:43 CEST] <rehunted> I guess is easy to setup a socket proxy to use with trickle instead of using shared libs on ffmpeg
[05:52:40 CEST] Action: rehunted Do'h I need compile all libs with shared too
[06:13:22 CEST] <Timothy_Gu> rehunted: don't use --enable-pic if you need shared libs, just use --enable-shared --disable-static
[06:15:22 CEST] <rehunted> Timothy_Gu, okay, trying again, because, this time, I get ffmpeg running, but trickle does not (I was trying to find what lib ffmpeg use for opening the MJPEG sockets)
[06:15:49 CEST] <rehunted> I'm trying to limit the bandwidth of MJPEG 
[06:17:18 CEST] <rehunted> trickle hack recv/send/ioctl and printf of applications 
[06:17:27 CEST] <rehunted> http://static.usenix.org/event/usenix05/tech/freenix/full_papers/eriksen/eriksen.pdf
[06:19:13 CEST] <rehunted> Timothy_Gu, build finished fine, let me test it. thank you
[06:32:57 CEST] <rehunted> Timothy_Gu, no success with trickle. trying a hack with proxy
[12:00:18 CEST] <cone-113> ffmpeg 03wm4 07master:881b80b32944: network: prevent SIGPIPE on OSX
[12:00:19 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:168195f25add: Merge commit '881b80b3294483696a21b21a69fa5eee30f33037'
[12:09:52 CEST] <cone-113> ffmpeg 03Rodger Combs 07master:c729021f450f: configure: Check for SecItemImport for securetransport as well
[12:21:06 CEST] <cone-113> ffmpeg 03Clément BSsch 07master:b60c445965b5: tests: add some ASS/SSA/MKS remux and transcode tests
[13:01:54 CEST] <cone-113> ffmpeg 03Peter Cordes 07master:fb1be6303d9b: avfilter/vf_mpdecimate: pass the same AVClass context as use elsewhere to av_pixelutils_get_sad_fn()
[13:10:45 CEST] <ubitux> do we still need the libzvbi text output now that we have a native cc decoder?
[13:27:14 CEST] <cone-113> ffmpeg 03Peter Cordes 07master:50f112c91535: avfilter/vf_mpdecimate: debug logging of threshold stats
[13:40:52 CEST] <Compnn> ubitux : zvbi is for teletext
[13:40:58 CEST] <Compnn> cc decoder is for closed captions
[13:41:06 CEST] <Compnn> i think
[13:42:12 CEST] <ubitux> my bad.
[14:18:01 CEST] <cone-113> ffmpeg 03Peter Cordes 07master:6590acf0fd0b: avfilter/vf_mpdecimate: support more pixel formats, including GBRP
[14:18:02 CEST] <cone-113> ffmpeg 03Peter Cordes 07master:3ad3529b567e: avfilter/vf_mpdecimate: add comment about 8x8 SAD
[18:57:52 CEST] <Daemon404> "ffmpeg players are noticeably more magenta and when comparing, Safari playback does feel more natural."
[18:58:07 CEST] <Daemon404> from chromium dev re: wrong colormatrix 
[18:58:14 CEST] Action: Daemon404 cringes at "more natural"
[18:58:26 CEST] <wm4> I almost thought I was reading something from an audiophile
[18:58:32 CEST] <Mavrik> Apple way the natural way!
[18:58:52 CEST] <JEEBsv> lol
[18:58:53 CEST] <Daemon404> wm4, me too.
[18:59:00 CEST] <rehunted> kk
[19:00:14 CEST] <Daemon404> oh well, best if it actually gets some attention
[19:01:11 CEST] Action: Daemon404 spies rehunted 
[19:18:20 CEST] <cone-113> ffmpeg 03Gilles Chanteperdrix 07master:d61386a69002: avformat/libquvi: fix error handling
[20:14:03 CEST] <jamrial> so libav added their own nvenc encoders...
[20:18:15 CEST] <wm4> of course
[20:29:06 CEST] <BBB> did they backport the vp9 decoder?
[20:29:10 CEST] <BBB> or are they writing their own?
[20:29:46 CEST] <jamrial> they backported it long ago, then reindented it and made it hard to backport the asm or any other change you wrote since then
[20:29:57 CEST] <BBB> awkward
[20:30:08 CEST] <BBB> ohwell not my problem i suppose
[20:32:38 CEST] <jamrial> rcombs: the tls securetransport stuff seems to have broken mac builds on arm
[20:50:24 CEST] <Compnn> people run mac on arm.... ok then
[20:56:21 CEST] <wm4> iOS?
[21:05:09 CEST] <Compnn> oh yeah, nevermind
[21:31:47 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:254eb5b6faeb: ppc: avutil: Drop a potentially dangerous workaround
[21:31:48 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:876a663c6681: Merge commit '254eb5b6faebb7bcfc3cefc1edced6652fe9d5be'
[21:46:42 CEST] <rcombs> jamrial: huh?
[21:47:16 CEST] <philipl> jamrial: yay. Of course they did...
[21:47:57 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:72cebae0d981: ppc: avutil: Use the abriged vector types
[21:47:58 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:e8676c758a62: Merge commit '72cebae0d981dde144340cf51f3c323f01e215e5'
[21:51:45 CEST] <ubitux> :/
[21:53:54 CEST] <jamrial> rcombs: http://fate.ffmpeg.org/report.cgi?time=20150531030339&slot=arm64-darwin-clang-apple-5.1 http://fate.ffmpeg.org/report.cgi?time=20150531030718&slot=armv7-darwin-clang-apple-5.1
[21:54:45 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:7014b65995b1: ppc: pixblockdsp: Use the abriged vector types
[21:54:46 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:08f229d81e41: Merge commit '7014b65995b1fe6188fb53447bf61b08e3963355'
[21:54:49 CEST] <philipl> This vdpau hevc stuff is pretty broken right now.
[21:55:36 CEST] <philipl> The decoder spits out frames that are laid out differently in memory than the rest of vdpau suspects, so you end up with a frame that has the bottom half interlaced with the top half.
[21:55:46 CEST] <philipl> s/suspects/expects
[21:56:37 CEST] <philipl> I also suspect there are internal alignment problems which mean that the bottom half gets partially written into the top halfs alignment scratch space so you lose content too.
[21:56:41 CEST] <rcombs> jamrial: I already sent a patch to fix that
[21:56:41 CEST] <philipl> *sigh*
[21:57:12 CEST] <jamrial> rcombs: ah, nevermind then
[21:57:22 CEST] <rcombs> it just hasn't run again since it was applied
[21:57:52 CEST] <rcombs> (it's a dumb fix: it disables Secure Transport in cases where that function is missing [which is all iOS])
[21:58:30 CEST] <rcombs> (optimally, we should implement it ourselves or work some other method out for handling those cases [PEM imports])
[21:58:43 CEST] <rcombs> (but the build is green so at least we're not regressing in the short-term)
[22:19:06 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:baa94563fede: ppc: linux: Check altivec using the auxv
[22:19:07 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:3058872c293e: ppc: Clarify and extend the cpuid check
[22:19:08 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:ff1995437fa2: Merge commit 'baa94563fede8959a638b0fa132dd2124acd93e8'
[22:19:09 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:47cc1a2bf1b7: avutil/ppc/cpu: Check if unistd.h is available before including it
[22:19:10 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:8dcbfb73d58c: Merge commit '3058872c293e239e3b51e86fe18cfbe720aadff1'
[22:42:01 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:f22ebd2555d1: ppc: configure: Support ISA 2.06 and later
[22:42:02 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:3f21cdd85bfb: Merge commit 'f22ebd2555d15b59e109a5b630dd71374b076d0b'
[22:44:59 CEST] <nevcairiel> philipl: sounds like drivers are broken
[22:50:09 CEST] <wm4> nvidia didn't post patches this time
[22:50:16 CEST] <wm4> (they did when vdpau was introduced)
[22:50:22 CEST] <wm4> did they even try it?
[22:50:27 CEST] <nevcairiel> maybe they know its not done yet
[22:50:37 CEST] <nevcairiel> did a driver release advertise this as a feature?
[22:58:31 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:da60b99a8857: ppc: Restrict some Altivec implementations to Big Endian
[22:58:32 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:7d07ee5a9bd1: ppc: cpu: Add support for VSX and POWER8 extensions
[22:58:33 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:902ce2a6c436: Merge commit 'da60b99a8857d5ca236f32c1799a066e0135a866'
[22:58:34 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:58a420487337: Merge commit '7d07ee5a9bd170a06d26fd967cf8de5d3b1ce331'
[23:08:04 CEST] <wm4> nevcairiel: somewhat, http://www.nvidia.com/Download/driverResults.aspx/84043/en-us
[23:08:43 CEST] <wm4> but if the API advertises support, I'd say it implies to be confirmed as working by nvidia at least
[23:08:54 CEST] <nevcairiel> i guess so
[23:09:17 CEST] <nevcairiel> i was happy to see that the 10-bit decoding on the GTX960 worked right away out of the box on release
[23:09:21 CEST] <nevcairiel> on windows, that is
[23:16:03 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:1aff90504aa6: configure: ldbrx seems not working on qemu power8, disable for now
[23:43:29 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:eecd29b3fd7f: ppc: vsx: Implement float_dsp
[23:43:30 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:da9a6f4a2b9a: Merge commit 'eecd29b3fd7fee221580798346d6582b75c7ade4'
[23:43:31 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:1046b6b09326: avutil/ppc/float_dsp_init: Disable duplicate functions
[23:49:13 CEST] <Compnn> philipl : maybe someone should email nvidia ask whats up
[23:56:36 CEST] <cone-113> ffmpeg 03Luca Barbato 07master:d0bf20a4f25a: ppc: vsx: Implement diff_pixels and get_pixels
[23:56:37 CEST] <cone-113> ffmpeg 03Michael Niedermayer 07master:1ff47770ca50: Merge commit 'd0bf20a4f25ac5de021c860a0c8ad05638ee2078'
[00:00:00 CEST] --- Mon Jun  1 2015