[Ffmpeg-devel-irc] ffmpeg-devel.log.20140625

[burek]

[01:01] <cone-96> ffmpeg.git 03Lukasz Marek 07master:a30f1b15c7ee: doc/build_system: document examples and config targets
[01:17] <BBB> jamrial: ok so I fixed the asm to be same speed as gcc-4.8.3 inline for 32bit
[01:17] <nevcairiel> \o/
[01:17] <jamrial> swr? cool
[01:17] <BBB> jamrial: your 64bit asm dump wasnt 100% useful because it works only on win64, not unix64& Id be very interested in a unix64 asm dump from gcc-4.8.3 so I can compare closer
[01:18] <BBB> I pushed to github, please test if you get a chance
[01:18] <jamrial> i think the win64 asm dump was from J_Darnley
[01:18] <BBB> oh, right
[01:18] <jamrial> i can give you a gcc 4.7.x or a 4.9.0 unix64 dump if it's useful
[01:18] <BBB> J_Darnley: can I get unix64 asm dump from gcc-4.8.3 also?
[01:18] <BBB> is it faster?
[01:19] <jamrial> well, the 4.7.x unix64 gave me those awful results i mentioned the other day. the 4.9.x unix64 was only slightly slower
[01:20] <jamrial> awful results = bogus, probably
[01:20] <BBB> hm& so apparently gcc-4.8.3 has the great touch
[01:20] <BBB> ok lets wait for that one then
[01:20] <BBB> Im just hoping theres something I can run locally to confirm speed against
[01:20] <BBB> the 32bit asm dump was quite useful for that
[01:20] <J_Darnley> BBB: not from me
[01:20] <BBB> hm...
[01:21] <J_Darnley> perhaps I should clarigy that...
[01:21] <BBB> J_Darnley: the asm dumps were from you and were gcc-4.8.3 right?
[01:21] <J_Darnley> yes
[01:21] <jamrial> ah wait, you mean a dump for the current inline version? if so then give me a moment, because for both 4.7 and 4.9 inline was faster than the yasm port
[01:22] <J_Darnley> I cannot give you a unix64 disassebly because my compiler cannot produce "unix" code
[01:22] <BBB> jamrial: ah yes
[01:22] <jamrial> ok, sorry for the confusion
[01:23] <BBB> also if you can confirm theyre still faster, thatd be great
[01:23] <BBB> but poking me the asm is fine also
[01:23] <BBB> then Ill compare myself
[01:23] <J_Darnley> Sorry, was that request direct at me?
[01:23] <J_Darnley> *directed
[01:24] <BBB> no, at jamrial
[01:25] <BBB> if you want to help testing speed, thatd be useful (I prefer to confirm across a bunch of systems that the yasm version is not slower)
[01:25] <jamrial> http://pastebin.com/Ez2fSdiV unix64 4.9.0 (with golomb bug fix backported from gcc-4.9 branch, but that's not relevant for this i guess)
[01:26] <BBB> thanks!
[01:26] <jamrial> i'll test what you uploaded to github and compare again later
[01:30] <BBB> crap your syntax is the inverse of what j_darnley gave me
[01:30] <BBB> J_Darnley: what did you use to dump the asm?
[01:31] <nevcairiel> with inverse oyu mean the operand order? thats weird, msvc tools should use the same as gcc
[01:32] <nevcairiel> this dump looks the wrong side up =p
[01:35] <jamrial> BBB: this better? http://pastebin.com/hS1HFPjT
[01:35] <jamrial> Intel instead of AT&T
[01:35] <BBB> yes!
[01:35] <jamrial> cool :P
[01:37] <nevcairiel> ah that looks better
[01:37] <J_Darnley> BBB: objdump -M intel
[01:37] <J_Darnley> ... too slow
[01:49] <BBB> nope that still beats me by a lot
[01:49] <BBB> 4 cycles/sample :(
[01:49] <BBB> bah
[01:49] <BBB> 3.5 really
[01:49] <BBB> still
[01:50] <BBB> must be register pre-loading& so I guess I have to do that next then
[01:50] <BBB> bah
[01:50] <BBB> ohwell
[02:00] <jamrial> BBB: i'm going to test using the measure macros you have on github, see if they help (i was getting lots of skips with the normal timer.h ones sometimes)
[02:00] <jamrial> just noticed them
[05:10] <debianuser> Hello. Got one more concat bug. Command `ffmpeg $(ls -1 ???.wav | sed 's/^/-i /') -filter_complex concat=n=399:v=0:a=1 0.wav` freezes with zero CPU/disk usage, it just does nothing. Output: pastebin.com/VZYew6f7 Files: ge.tt/1Eg60nl1/v/2 (150KB)
[05:17] <debianuser> Is that my build is buggy?
[05:26] <debianuser> Or have I hit some buffer overflow with my files?
[05:36] <BtbN> i'd use the concat input
[05:49] <debianuser> I can't. Files have different frame rates mono/stereo. I was told that "-f concat" can't handle those.
[07:55] <ubitux> is there an expected processing between ff_ref_[fi]dct()?
[07:56] <ubitux> it seems i can't get block==idct(fdct(block))
[08:59] <kurosu> depending on the dct, sometimes some rescaling is left to the quantization
[09:00] <kurosu> don't know about the _ref_ version, but I would try various rescaling after fdct or idct
[09:00] <kurosu> (by a constant)
[09:05] <ubitux> i tried to scale by n bits after fdct but it didn't seem to be help
[09:05] <ubitux> i'm still getting weird negative numbers in the output
[09:08] <ubitux> mmh maybe that's because i badly adjusted to 16-bit that code actually
[09:11] <ubitux> 16-width*
[10:16] <ubitux> ok, found my mistake
[10:17] <ubitux> anyway, the ref dct has a weird bias
[10:17] <ubitux> http://pastie.org/pastes/9321987/text is this expected?
[10:18] <ubitux> and adjusted for 16px: http://pastie.org/pastes/9321988/text
[10:24] <ubitux> seems i get better result if i scale before rounding in the fdct
[10:25] <ubitux> but still not perfect
[10:30] <ubitux> or just rounded scaling
[10:34] <ubitux> so basically, in dctref, instead of doing block[i + j] = floor(tmp + 0.499999999999) in fdct (and later a -rounded or not- scaling by N), i'm doing block[i + j] = lrint(tmp / BSIZE) (BSIZE being 1<<N, so 8 or 16)
[10:35] <ubitux> and i'm getting this: http://pastie.org/pastes/9322044/text
[10:35] <ubitux> i have no idea where the other inaccuracy comes from
[10:35] <ubitux> any ideas?
[10:37] <ubitux> (code: http://pastie.org/9322056 - based on dctref)
[10:47] <debianuser> Hello. I have something that looks like concat bug. Command `ffmpeg $(ls -1 ???.wav | sed 's/^/-i /') -filter_complex concat=n=399:v=0:a=1 0.wav` freezes with zero CPU/disk usage, it just does nothing. Output: pastebin.com/VZYew6f7 Files: ge.tt/1Eg60nl1/v/2 (150KB)
[10:47] <debianuser> Is that my build is buggy? Or have I hit some buffer overflow with my files?
[11:19] <plepere> wow, avx2 pack is retarded : if you want to do a 256b pack, you need to shuffle afterwards. ;.
[11:19] <plepere> :/
[11:37] <J_Darnley> There's always something retarded about x86 simd instructions
[11:39] <J_Darnley> I find the lack of some operations at some data sizes to be particularly annoying
[12:57] <michaelni> ubitux, fdct+idct isnt really guranteed to give the same integers you started with due to finite precission and rounding
[13:20] <ubitux> michaelni: right, but it seems i'm able to get exact in/out with what doing some naive double av_rdft 
[13:20] <ubitux> and so i was wondering why the reference wasn't exact at all
[13:21] <ubitux> a better scale/rounding within the idct really helps
[13:21] <ubitux> but not so much
[13:22] <michaelni> the idct is specified to be as it is, if you intend to use it in mpeg/h26*
[13:22] <michaelni> the fdct could be changed in any way we see fit
[13:23] <ubitux> i was talking about the fdct
[13:23] <ubitux> block[i + j] = floor(tmp + 0.499999999999);
[13:23] <ubitux> this weird rounding
[13:23] <ubitux> i changed that to block[i + j] = lrint(tmp / 8)
[13:24] <ubitux> (so including the scaling within)
[13:24] <ubitux> i get way better results, but still not perfect
[13:24] <ubitux> i wonder why i have not a single diff with a the double av_rdft_calc
[13:25] <michaelni> because its double ?
[13:25] <ubitux> dctref is double
[13:25] <michaelni> but its output is int16
[13:25] <ubitux> (i meant av_dct_calc sorry)
[13:25] <ubitux> ah, mmh yeah right
[13:26] <ubitux> ok
[13:26] <michaelni> also tmp/8 may have effects on trellis performance
[13:26] <ubitux> can you elaborate?
[13:27] <ubitux> a post scaling with something like (lrint(tmp)+4)>>3 doesn't give a better accuracy than lrint(tmp/8) here
[13:27] <ubitux> i have more mismatches
[13:28] <ubitux> ah, i missed the "performance" ok
[13:28] <ubitux> note that dctref is only used in dct-test
[13:28] <ubitux> so it shouldn't matter much
[13:28] <ubitux> and speaking of this
[13:29] <ubitux> in dct-test, a see no rounding in the scaling:
[13:29] <ubitux>             for (i = 0; i < 64; i++)
[13:29] <ubitux>                 block[i] >>= 3;
[13:29] <michaelni> we quantize the values after the fdct, and higher precission should allow higher precission distortion calculation
[13:29] <michaelni> hmmm
[13:29] <ubitux> i'm gonna write a patch
[13:30] <michaelni> also block[i + j] = floor(tmp + 0); in fdct and (...+round)>>3 later should give the same results
[13:30] <ubitux> http://pastie.org/9322657
[13:30] <ubitux> basically i mean this ^
[13:31] <ubitux> and no, having block[i]>>=3 changed so it has rounding doesn't give as good result as rounding within ff_ref_fdct()
[13:31] <ubitux> but yeah it's better
[13:36] <ubitux> michaelni: so basically currently i have this: http://pastie.org/9322666
[13:36] <ubitux> that is, with the floor(x + 0.4999...) and scaling >>= 3
[13:36] <ubitux> if i do a good rounding, that is x+4>>3
[13:36] <michaelni> http://pastie.org/9322657 <-- breaks dct-test
[13:36] <ubitux> i get way better result: http://pastie.org/9322668
[13:37] <ubitux> and if i removing the scaling completely and integrate within fdct (that is lrint(tmp/8)), i get the best results: http://pastie.org/9322667
[13:38] <ubitux> michaelni: yeah, haven't tested properly, that's just so you get the idea
[13:40] <michaelni> floor(x/8+0.5) and floor(x) with later +4>>3 should give the same results
[13:41] <ubitux> doesn't seem to, but maybe i'm doing something wrong
[13:42] <michaelni> so for what value of x does it differ ?
[13:43] <michaelni> anyway like i said we can change the fdct in any way if it improves things
[13:44] <ubitux> any idea why we had floor(x+0.499999...) instead of floor(x+0.5) btw?
[13:44] <ubitux> i'd use lrint(x) anyway for this but well
[13:50] <michaelni> the 0.49 comes from some referenc from iso/itu i think but would have to check
[13:50] <ubitux> ok
[13:51] <michaelni> or at least it was there as long as i can remember
[13:51] <michaelni> mateo`, is the new iteration of the  "set/force channelcount in MXF D-1" patch ok to push ?
[14:03] <Gizmoomzig> Hello, I am trying to get certain information out of the H264 decoder. One of them is the ei_flag but i cant seem to find it. Can someone please point me in the right direction where i can find the error indicator if error concealment has to be used on a macroblock?
[14:39] <plepere> michaelni, should I re-do a patch removing the mmxext function prototype changes ?
[14:41] <michaelni> plepere, if you have time yes, otherwise i can do it
[14:42] <plepere> I'm busy on avx2 mc functions, but I'll do it probably tomorrow. :)
[14:50] <Luigi12_work> hello.  I'm preparing updated ffmpeg packages for Mageia.  http://ffmpeg.org/security.html for 1.1.x is up to date as of 1.1.9.  Do 1.1.10 and 1.1.11 fix any further CVEs?
[14:53] <michaelni> Luigi12_work, i suggest you wait a day or 2 before updating from packages from branch theres some backports pending
[14:54] <michaelni> packages from 1.1 branch
[14:54] <Luigi12_work> michaelni: ok thanks.  We have 1.1.x in Mageia 3 and 2.0.x in Mageia 4.
[14:56] <michaelni> ok, didnt know, maybe someone should start a wiki page on who uses what 
[14:56] <Luigi12_work> not a bad idea :o)
[14:58] <Luigi12_work> michaelni: are these backports pending for 2.0.x as well?
[14:59] <michaelni> yes, ill try to get them out ASAP
[14:59] <Luigi12_work> michaelni: thanks again
[15:06] <cone-910> ffmpeg.git 03plepere 07master:942e22c65116: avcodec/x86/hevc: add avx2 dc idct
[15:07] <cone-910> ffmpeg.git 03plepere 07master:9ba6b17add28: avcodec/x86/hevc_idct: fix number of sse registers
[15:07] <plepere> oh thanks, michaelni. :D
[15:07] <michaelni> np
[15:26] <Daemon404> btw not sure if i ever posted this here: https://gist.github.com/dwbuiten/7101755
[15:26] <Daemon404> useful for fuzzing ffmpeg
[16:10] <michaelni> no, i think ive not seen that before, thanks
[17:01] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:91d024d75dce: avutil/timestamp: Warn about missing __STDC_FORMAT_MACROS for C++ use
[17:01] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:9933e06595c0: swscale/x86/swscale: fix missing xmm clobbers in yuv2yuvX_sse3()
[17:01] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:d581567e0948: avcodec/utils: fix sizeof(AVFrame) dependence in  avcodec_encode_audio2()
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:6387aa94d664: avcodec/h263dec: Fix use of uninitialized memory from the bitstream buffer
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:f65501f80cbc: ffmpeg: dont call exit_program() from a signal hander
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:673f679c8ae2: avformat/mp3enc: use av_copy_packet()
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:3ae71dd33ed3: avcodec: Add padding after the remaining AVFrames
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:36fc9bff08f9: swscale/swscale: fix integer overflow
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:102df43f49cb: dox/scaler:fix bicubiclin typo
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:b556432f5a75: avcodec/x86/mpegvideoenc_template: fix integer overflow
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:c7f419efd1df: avcodec/h264_mp4toannexb_bsf: prepend global headers before any in stream parameter sets
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:02055da69ace: avcodec/libx264: move where x264opts is applied down so it isnt overridden by avctx & defaults
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:5ac53d07a276: avcodec/h264: clear cur_pic structure instead of duplicating it in ff_h264_update_thread_context()
[17:02] <cone-910> ffmpeg.git 03Lukasz Marek 07release/2.0:c61c66cfb17e: lavu/opt: validate range before dereference
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:5805c0c2b6cf: avcodec/wma: use av_freep(), do not leave stale pointers in memory
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:491e1ec95db4: avcodec/msrle: check return code for success before use
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:1c321b7c31aa: swresample/resample: Limit filter length
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:c9ed7a4fc13c: swresample/dither: use av_malloc_array()
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:e315549ffa16: swresample/resample: use av_malloc_array() where appropriate
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:218c853774a9: swscale/x86/swscale_template: loose hardcoded dstw_offset
[17:02] <cone-910> ffmpeg.git 03Anthoine Bourgeois 07release/2.0:853bbe1aed3b: avcodec/dirac_arith: Fix build with PIC and stack-check options
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:2ca68b98ba49: avcodec/x86/idct_sse2_xvid: fix non C99 inline function
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:690a3c4fb7c9: swscale/swscale: fix srcStride/srcSlice typo
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:6f6f5f9c8b3d: avcodec/mjpegen: Fix declared argument size
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:ed38ed3b629f: avformat/mpegts: Remove redundant check
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:8162b69091b6: avcodec/diracdec: fix undefined behavior with shifts
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:66ca450fcb07: avcodec/g723_1: add assert to help static code analyzers
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:98805cbac5a4: avfilter/f_select: fix loss of precission in SAD calculation
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:6882a04706b9: avfilter/vf_deshake: fix loss of precission with odd resolutions
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:4d9aad95b517: iavcodec/vc1dec: Fix missing {}
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:eaa6d7910545: avfilter/filtfmts: Support dynamically allocated in/outputs
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:d3bcc9cfa9fc: avformat/mux: Check for and remove invalid packet durations
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:ee282be99b9d: avformat/h263dec: Fix h263 probe
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:0be9a2e9d9bb: swresample: fix AV_CH_LAYOUT_STEREO_DOWNMIX input
[17:02] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:db33fb1c4866: ffmpeg_filter: fix pointer to local outside scope
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:d9d66d2d16fa: avcodec/mjpegdec: Fix undefined shift
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:99d4d3532fac: avfilter/graphdump: Fix pointer to local outside scope
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:0e18480fc3b3: sws: dont use the optimized 410->420 unscaled conversion when height%4
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:62b9c99aec0f: avcodec/diracdec: move mc buffer allocation to per frame
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:332777970c8b: avutil/cpu: force mmx on selection of higher x86 SIMD features
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:bdd2f7c7a761: avcodec/libvorbisenc: dont add the duration to AV_NOPTS_VALUE
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:2f1f4070446e: avcodec/aac: fix () in IS_CODEBOOK_UNSIGNED macro
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:db6f3be9418c: avcodec/golomb-test: fix () in EXTEND() macro
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:a83304814c28: avcodec/h264: fix () in macros
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:4965da92eff9: avcodec/ivi_dsp: add some missing () to macros
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:a144d305dd65: avcodec/mlpdec: fix () in MSB_MASK() macro
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:81a2740919ae: avcodec/mss34dsp: fix () in SOP* macros
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:3c169251eec2: avcodec/mss4: Fix () in MKVAL() macro
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:9fe7e450305c: avcodec/aic: fix quantization table permutation
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:a2b4e38729ca: avformat/flvenc: Do not allow creating h263/mpeg4 in flv without unofficial format extensions being enabled.
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:b94621ac6ba6: avcodec/alsdec: Clear MPEG4AudioConfig so that no use of uninitialized memory is possible
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:b93659db5bbe: avformat/cavsvideodec: Fix probing when the file extension is avs
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:01e2dd53a067: cavsdec: fix qp fixed slice handling
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:1e9ae5dbbd67: avformat/mpc: attempt to allocate a packet that is not smaller than the data inside it
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:083d0b7b2d92: avutil/lzo: Fix integer overflow
[17:03] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:4a237657047b: avutil/lzo: add asserts to be double sure against overflows
[17:15] <cone-910> ffmpeg.git 03Michael Niedermayer 07release/2.0:205e2264c3d5: update for FFmpeg 2.0.5
[18:24] <cone-910> ffmpeg.git 03Michael Niedermayer 07master:1ace0ca60ff8: avcodec/x86/hevc_idct: fix function name in comment
[18:24] <cone-910> ffmpeg.git 03Michael Niedermayer 07master:6dffc8f5aaab: avfilter/vf_pullup: use ptrdiff_t as stride argument for dsp functions
[18:24] <cone-910> ffmpeg.git 03Michael Niedermayer 07master:b8255a4c7096: avfilter/x86/vf_pullup: fix old typo
[22:23] <cone-533> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n2.0.5': unknown revision or path not in the working tree.
[22:23] <cone-533> Use '--' to separate paths from revisions
[22:23] <cone-533> refs/tags/n2.0.5:HEAD: avfilter/x86/vf_pullup: fix old typo
[23:15] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:1f2f6b247aaf: avutil/lzo: Fix integer overflow
[23:15] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:574bf2ce4ccd: avutil/lzo: add asserts to be double sure against overflows
[23:15] <cone-533> ffmpeg.git 03Luca Barbato 07release/1.1:503322f97c5a: lzo: Handle integer overflow
[23:15] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:a8dd7fe5b994: Merge commit '503322f97c5a25a020933ed4ab510697d5f5b4af' into release/1.1
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:4533e8b30a38: avcodec/diracdec: move mc buffer allocation to per frame
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:e68114868ffd: avutil/cpu: force mmx on selection of higher x86 SIMD features
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:bd2266975cbd: avcodec/libvorbisenc: dont add the duration to AV_NOPTS_VALUE
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:80588930fffb: avformat/flvenc: Do not allow creating h263/mpeg4 in flv without unofficial format extensions being enabled.
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:71a4a9183f3e: avcodec/alsdec: Clear MPEG4AudioConfig so that no use of uninitialized memory is possible
[23:35] <cone-533> ffmpeg.git 03Michael Niedermayer 07release/1.1:f0d56d4f5542: avformat/mpc: attempt to allocate a packet that is not smaller than the data inside it
[00:00] --- Thu Jun 26 2014