[Ffmpeg-devel-irc] ffmpeg-devel.log.20111223

[burek]

[02:03] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r9eb0d8bab1 10ffmpeg/libavutil/fifo.c: 
[02:03] <CIA-49> ffmpeg: fifo: Make writes atomic.
[02:03] <CIA-49> ffmpeg: Prior to this a X bytes write could be seen as less than X bytes being
[02:03] <CIA-49> ffmpeg: available if the check was done at an unfortunate moment.
[02:03] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:03] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r3dcbafc777 10ffmpeg/libavformat/udp.c: 
[02:03] <CIA-49> ffmpeg: udp: return circular buffer error if such error happened.
[02:03] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:03] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rbc900501e0 10ffmpeg/libavformat/udp.c: (log message trimmed)
[02:03] <CIA-49> ffmpeg: udp: Replace double select() by select+mutex+cond.
[02:03] <CIA-49> ffmpeg: When no data was available both the buffer thread as well as
[02:03] <CIA-49> ffmpeg: the main thread would block in select(), when data becomes
[02:03] <CIA-49> ffmpeg: available both should move forward and as data is read in the
[02:03] <CIA-49> ffmpeg: buffer thread the main thread would block in select() later
[02:03] <Daemon404> hmmm
[02:03] Action: Daemon404 sees an error with fate.texi or doc/Makefile but cant decide which
[02:14] <Daemon404> is alexander strasser ever on irc?
[02:21] <iive> i think his nick is beastd
[02:21] <Daemon404> oic
[02:28] <Daemon404> iive, i just use the mailing list anyway.
[03:38] <CIA-49> ffmpeg: 03Anton Khirnov 07master * r131609dc2a 10ffmpeg/libswscale/swscale_internal.h: 
[03:38] <CIA-49> ffmpeg: sws: readd PAL8 to isPacked()
[03:38] <CIA-49> ffmpeg: Fixes PAL8 to YUV conversion.
[03:38] <CIA-49> ffmpeg: 03Anton Khirnov 07master * r196c9e5c8c 10ffmpeg/ (configure libavfilter/Makefile libavfilter/allfilters.c): 
[03:38] <CIA-49> ffmpeg: lavfi: always build vsrc_buffer.
[03:38] <CIA-49> ffmpeg: It's a part of public API.
[03:38] <CIA-49> ffmpeg: 03Diego Biurrun 07master * raaf47bcde7 10ffmpeg/libavcodec/ (27 files): 
[03:38] <CIA-49> ffmpeg: Drop ALT_ prefix from BITSTREAM_READER_LE name.
[03:38] <CIA-49> ffmpeg: The prefix is a historic remnant that probably meant "alternative".
[03:38] <CIA-49> ffmpeg: Now that the A32 bitstream reader has been dropped it makes no sense anymore.
[03:38] <CIA-49> ffmpeg: 03Justin Ruggles 07master * rf264d336fe 10ffmpeg/libavcodec/truespeech.c: 
[03:38] <CIA-49> ffmpeg: truespeech: fix invalid reads in truespeech_apply_twopoint_filter()
[03:38] <CIA-49> ffmpeg: fixes Bug 171
[03:38] <CIA-49> ffmpeg: 03Alex Converse 07master * r4bf3c8f226 10ffmpeg/ (5 files in 3 dirs): 
[03:38] <CIA-49> ffmpeg: mov: Don't stick the QuickTime field ordering atom in extradata.
[03:38] <CIA-49> ffmpeg: The 'fiel' atoms can be found in H.264 tracks clobbering the extradata.
[03:38] <CIA-49> ffmpeg: MJPEG supports non field based extradata, and this data should be
[03:38] <CIA-49> ffmpeg: preserved when copying.
[03:38] <CIA-49> ffmpeg: 03Janne Grunau 07master * rc32e4029d9 10ffmpeg/libavcodec/options.c: 
[03:38] <CIA-49> ffmpeg: options: set minimum for "threads" to zero
[03:38] <CIA-49> ffmpeg: A negative number of threads does not make sense and 0 is used for
[03:38] <CIA-49> ffmpeg: autodetection. Adds a symbolic name for autodetection.
[03:38] <CIA-49> ffmpeg: 03Janne Grunau 07master * rea4d5f4837 10ffmpeg/ (configure libavcodec/pthread.c): 
[03:38] <CIA-49> ffmpeg: linux: use number of CPUs as automatic thread count
[03:38] <CIA-49> ffmpeg: Use sched_getaffinity to determine the number of logical CPUs.
[03:38] <CIA-49> ffmpeg: Limits the number of threads to 16 since slice threading of H.264
[03:38] <CIA-49> ffmpeg: seems to be buggy with more than 16 threads.
[03:39] <CIA-49> ffmpeg: 03Shitiz Garg 07master * r355d917c0b 10ffmpeg/libavcodec/4xm.c: 
[03:39] <CIA-49> ffmpeg: 4xm: Add a check in decode_i_frame to prevent buffer overreads
[03:39] <CIA-49> ffmpeg: Fixes bugzilla #135
[03:39] <CIA-49> ffmpeg: Signed-off-by: Janne Grunau <janne-libav at jannau.net>
[03:39] <CIA-49> ffmpeg: 03Gaurav Narula 07master * rad3161ec1d 10ffmpeg/libavcodec/kmvc.c: 
[03:39] <CIA-49> ffmpeg: kvmc: fix invalid reads
[03:39] <CIA-49> ffmpeg: Signed-off-by: Janne Grunau <janne-libav at jannau.net>
[03:39] <CIA-49> ffmpeg: 03Paul B Mahol 07master * r01a01bf8bd 10ffmpeg/libavcodec/adpcm.c: 
[03:39] <CIA-49> ffmpeg: adpcm: fix IMA SMJPEG decoding
[03:39] <CIA-49> ffmpeg: Signed-off-by: Janne Grunau <janne-libav at jannau.net>
[03:39] <CIA-49> (30 lines omitted)
[03:46] <pasteeater> michaelni: the "clone git://source..." is fine with me. my download page suggestion wasn't that strong of a preference of mine.
[03:48] <pasteeater> also, i'm out of town for a week, so the FFmpeg living spam filter might be a little slower than normal
[03:49] <michaelni> np
[03:49] <michaelni> ill think about how to best word it ...
[03:50] <michaelni> iam not sure myself ATM what is best
[03:50] <pasteeater> the original was good enough
[06:48] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rae7a6441df 10ffmpeg/libavcodec/h264_ps.c: 
[06:48] <CIA-49> ffmpeg: h264: Decode short VUI without error
[06:48] <CIA-49> ffmpeg: Fixes 47j9R7PXBep.mov
[06:48] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:48] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rf810ab45e3 10ffmpeg/libavfilter/ (avfilter.c avfilter.h): 
[06:48] <CIA-49> ffmpeg: libavfilter: increase LIBAVFILTER_VERSION_MICRO to 100 like the other libs.
[06:48] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[11:57] <CIA-49> ffmpeg: 03Carl Eugen Hoyos 07master * re5de928923 10ffmpeg/libavcodec/aacdec.c: 
[11:57] <CIA-49> ffmpeg: Fix a possible endless loop when decoding aac.
[11:57] <CIA-49> ffmpeg: Fixes ticket #789.
[11:58] <CIA-49> ffmpeg: 03Carl Eugen Hoyos 07master * r8e88145d0b 10ffmpeg/libavcodec/svq1enc.c: 
[11:58] <CIA-49> ffmpeg: Release buffers after encoding svq1.
[11:58] <CIA-49> ffmpeg: Fixes ticket #820.
[11:58] <CIA-49> ffmpeg: 03Carl Eugen Hoyos 07master * rc25c7ba87c 10ffmpeg/libavutil/imgutils.c: Make systematic palette opaque.
[16:10] <CIA-49> ffmpeg: 03Janne Grunau 07master * r8d4668cc5f 10ffmpeg/libavcodec/pthread.c: 
[16:10] <CIA-49> ffmpeg: pthread: include sys/types.h before sys/sysctl.h
[16:10] <CIA-49> ffmpeg: Fixes compilation on FreeBSD with clang 3.
[16:10] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:10] <CIA-49> ffmpeg: 03Janne Grunau 07master * r5c33a9a5c1 10ffmpeg/libavcodec/pthread.c: 
[16:10] <CIA-49> ffmpeg: threads: check defines before using them in automatic thread detection
[16:10] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:37] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r7de733b752 10ffmpeg/libavcodec/pthread.c: 
[16:37] <CIA-49> ffmpeg: pthread: sysctl.h needs param.h on openbsd
[16:37] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:18] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * rc2c284b3eb 10ffmpeg/doc/Makefile: 
[18:18] <CIA-49> ffmpeg: Revert "doc: build: TEXIDEP: Recognize verbatim includes"
[18:18] <CIA-49> ffmpeg: This reverts commit e5cbf24e45ed21178674fff8f485d919128b793c.
[18:18] <CIA-49> ffmpeg: This commits breaks out of tree builds
[22:24] <GordonFreeman> hi
[22:24] <GordonFreeman> round(y+v*8) or y+round(v*8) for performance? (y is int)
[22:27] <iive> integer addition is considered faster, but don't forget that round still returns double
[22:28] <cbsrobot> I guess you could use some shift operations 
[22:28] <iive> if v is integer, then i don't see point in using round.
[22:29] <cbsrobot> v<<3 = v*8
[22:29] <caytchen> the compiler will already do that.
[22:29] <cbsrobot> but I think the compiler will optimize it
[22:30] <iive> cbsrobot: you are missing the point, you can't shift floats
[22:30] <cbsrobot> is v a float ?
[22:30] <iive> "if v is integer, then i don't see point in using round."
[22:31] <cbsrobot> true
[22:31] <Daemon404> [16:30] <@iive> cbsrobot: you are missing the point, you can't shift floats <-- on a bit-level, you could if you really wanted to
[22:31] <Daemon404> it'd just break shit
[22:34] <iive> Daemon404: the compiler won't accept it. you need to cast or alias it first.
[22:34] <Daemon404> iive, merely meant in theory you can
[22:34] <Daemon404> as a float is still a bunch of bits
[22:35] <iive> if it is in memory, there are special registers for floats. aka the same used by mmx.
[22:35] <Daemon404> not talking about in practice
[22:38] <iive> ~~
[23:08] <CIA-49> ffmpeg: 03Michael Niedermayer 07master * r8ff4fff774 10ffmpeg/libavformat/flvdec.c: 
[23:08] <CIA-49> ffmpeg: flvdec: disable new midstream param change code not only for h264 but all cases.
[23:08] <CIA-49> ffmpeg: It breaks some samples and iam not aware of one that it fixes.
[23:08] <CIA-49> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:37] <pengvado> GordonFreeman: y+lrintf(v*8)
[00:00] --- Sat Dec 24 2011