[Ffmpeg-devel-irc] ffmpeg-devel.log.20150204

Thu Feb 5 02:05:03 CET 2015

[00:05] <cone-276> ffmpeg.git 03Luca Barbato 07master:898276c16b16: r3d: Fix pos type
[00:05] <cone-276> ffmpeg.git 03Michael Niedermayer 07master:d304cb49eb09: Merge commit '898276c16b1683ac77723e97574a3bfdb29507fd'
[00:29] <cone-276> ffmpeg.git 03Christophe Gisquet 07master:e93d3a22cb53: x86: lavu/x264asm: fix ymm register instantiation
[00:48] <cone-276> ffmpeg.git 03wm4 07master:56cc02422088: avformat/mpc8: fix hang with fuzzed file
[01:03] <compn> .ram is rm playlist , so maybe is good ?
[01:03] <cone-276> ffmpeg.git 03wm4 07master:b737a2c52857: avformat/mpc8: fix broken pointer math
[02:07] <cone-276> ffmpeg.git 03Fabrice Bellard 07master:da81cc38e8a9: avcodec/hevc: reduce memory for SAO
[02:07] <cone-276> ffmpeg.git 03Christophe Gisquet 07master:00fe77855b2a: avcodec/hevc: Do not zero is_pcm on allocation
[02:07] <cone-276> ffmpeg.git 03Christophe Gisquet 07master:52f2adc015ba: avcodec/hevc: Update the USE_SAO_SMALL_BUFFER case for the alignment requirements in FFmpeg
[05:12] <cone-652> ffmpeg.git 03Michael Niedermayer 07master:77f326dc36b7: ffmpeg: Make CFR frame drop threshold user adjustable
[10:26] <ubitux> http://b.pkh.me/kdtree-try6.png finally.
[10:30] <nevcairiel> neat balance
[10:32] <ubitux> yes, but still slow :)
[10:32] <ubitux>    87.81%  ffmpeg_g  ffmpeg_g            [.] colormap_nearest_node
[10:32] <ubitux>     4.35%  ffmpeg_g  ffmpeg_g            [.] set_frame_sierra2_4a
[10:32] <ubitux>     2.26%  ffmpeg_g  ffmpeg_g            [.] ff_lzw_encode
[10:32] <ubitux> :(
[10:33] <ubitux> maybe i should try removing the recursion in the search somehow
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:6e7ac5ee554b: avcodec/webp: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:e9eb94ac5a66: avcodec/qcelpdec: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:c93a7b264f76: avcodec/ac3dec: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:49f825713536: avcodec/tak_parser: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:1634d2bcd141: avcodec/on2avc: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:63cb546f0a55: avcodec/wnv1: use init_get_bits8()
[10:59] <cone-616> ffmpeg.git 03Paul B Mahol 07master:cecc4ea1fc2f: avcodec/evrcdec: use init_get_bits8()
[12:56] <cone-616> ffmpeg.git 03Carl Eugen Hoyos 07master:adfe89bc4813: Fix standalone compilation of the tblend filter.
[13:10] <cone-616> ffmpeg.git 03Alex Converse 07master:82bd8a57692d: aacdec: Refactor decode_ics_info.
[13:10] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:0afc26d330e6: Merge commit '82bd8a57692d8f3643ad0903c97e4a0f84393f53'
[13:18] <cone-616> ffmpeg.git 03Liviu Oniciuc 07master:d2330e584411: doc/muxers/image2: remove the non-negative number requirement for start_number
[13:18] <cone-616> ffmpeg.git 03Stefano Sabatini 07master:fd242b468d66: lavf/utils: fix if_( style
[13:30] <cone-616> ffmpeg.git 03Alex Converse 07master:edd9ea3ffba1: aacdec: Refactor aac_er_decode_frame.
[13:30] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:5348d7b605c0: Merge commit 'edd9ea3ffba1d5873dead1c57c74b1e3cc912167'
[13:39] <cone-616> ffmpeg.git 03Niel van der Westhuizen 07master:0ee2573347ec: aacdec: Support for ER AAC in LATM
[13:39] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:f53aa3f4f331: Merge commit '0ee2573347ecdb9cb5656001f7201d819eec16d8'
[13:48] <cone-616> ffmpeg.git 03Niel van der Westhuizen 07master:7640c4a371a9: aacdec: Tables for length 480 AAC ELD.
[13:48] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:dd72e26d0cea: Merge commit '7640c4a371a97899dfb443d980fd0bf8ac587754'
[14:03] <cone-616> ffmpeg.git 03Alex Converse 07master:d615187f74dd: aacdec: Support for ER AAC ELD 480.
[14:04] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:ba4fba8f4873: Merge commit 'd615187f74ddf3413778a8b5b7ae17255b0df88e'
[14:11] <cone-616> ffmpeg.git 03Paul B Mahol 07master:469a17013554: avformat/siff: fix if_( style
[14:27] <cone-616> ffmpeg.git 03Reimar Döffinger 07master:8c1ce7787ad7: aacsbr: silence message for SBR extension "padding".
[14:27] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:85a8069d23ed: Merge commit '8c1ce7787ad7af8fc321b4a145e9dc965f6d0e86'
[14:40] <cone-616> ffmpeg.git 03Paul B Mahol 07master:9f1a5dbc26bf: avcodec/cdxl: use init_get_bits8()
[15:07] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:05e161952954: avformat/mpc8: Use uint64_t in *_get_v() to avoid undefined behavior
[16:44] <cone-616> ffmpeg.git 03Hendrik Leppkes 07master:b11994593ae4: dxva2_hevc: properly fill the scaling list structure
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:8a21620c26b8: avformat/mov: Fix memleaks for duplicate STCO/CO64/STSC atoms
[17:18] <cone-616> ffmpeg.git 03Stefano Sabatini 07release/1.2:fa8d407c1f8f: lavf/segment: remove duplicated and inconsistent cleanup code in seg_write_packet()
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:2dd1d0c60b06: avformat/cdxl: Fix integer overflow of image_size
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:fd4c80177b00: avformat/flvdec: Increase string array size
[17:18] <cone-616> ffmpeg.git 03wm4 07release/1.2:931f4313b2ec: avcodec/dvdsubdec: fix out of bounds accesses
[17:18] <cone-616> ffmpeg.git 03Dale Curtis 07release/1.2:07f634f94876: mov: Avoid overflow with mov_metadata_raw()
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:b6351f9978a4: avformat/mov: fix integer overflow in mov_read_udta_string()
[17:18] <cone-616> ffmpeg.git 03Dale Curtis 07release/1.2:54a9f64d4296: mov: Fix negative size calculation in mov_read_default().
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:e81236d1a5ac: avformat/matroskadec: Use av_freep() to avoid leaving stale pointers in memory
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:529488afa315: avformat/mov: Fix mixed declaration and statement warning
[17:18] <cone-616> ffmpeg.git 03Johan Andersson 07release/1.2:5410d6d26fc6: cmdutils: update copyright year to 2015.
[17:18] <cone-616> ffmpeg.git 03wm4 07release/1.2:8a3a3e3154b7: avcodec/dvdsubdec: error on bitmaps with size 0
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:08fe58ef3280: ffmpeg: Clear error message array at init.
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:b00a3bf33206: avformat/movenc: workaround bug in "PathScale EKOPath(tm) Compiler Suite Version 4.0.12.1"
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:a86cc0cc1bd2: avcodec/flac_parser: fix handling EOF if no headers are found
[17:18] <cone-616> ffmpeg.git 03Vittorio Giovara 07release/1.2:71bdcd66e15a: opt: check memory allocation
[17:18] <cone-616> ffmpeg.git 03Vittorio Giovara 07release/1.2:051b47dc4e08: swscale: check memory allocations
[17:18] <cone-616> ffmpeg.git 03Martin Storsjö 07release/1.2:237cf31a5736: rtpdec_h263_rfc2190: Clear the stored bits if discarding buffered data
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:abfa41618124: avformat/rmdec: Check for overflow in ff_rm_read_mdpr_codecdata()
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:9939d98b2c25: avformat/utils: Fix number suffixes in tb_unreliable()
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:0e79fd1ec925: swresample/dither: Cleanup number suffixes
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:ba5e66cf65f3: avformat/matroskadec: Fix number suffixes
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:ef3687998fb4: avformat/smacker: Fix number suffix
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:5f2902bdc76c: avcodec/h264_cabac: use int instead of long for mbb_xy
[17:18] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:7881d3b2a5c3: avcodec/mpegvideo_enc: Fix number suffixes in rc_buffer_size calculation
[17:18] <cone-616> ffmpeg.git 03wm4 07release/1.2:5d7320be52c3: avformat/tta: fix crash with corrupted files
[17:18] <cone-616> ffmpeg.git 03wm4 07release/1.2:bdc4abd694a3: avformat/mpc8: fix hang with fuzzed file
[17:19] <cone-616> ffmpeg.git 03wm4 07release/1.2:49dd89f9027f: avformat/mpc8: fix broken pointer math
[17:19] <cone-616> ffmpeg.git 03Michael Niedermayer 07release/1.2:70ff376e8357: avformat/mpc8: Use uint64_t in *_get_v() to avoid undefined behavior
[17:32] <kierank> great, a crash in libavfilter
[17:41] <JEEB> my condolences
[17:50] <wm4> as long as it doesn't crash in the format negotiation code...
[17:55] <aetasx> finally got a profiled + LTOed build done with all the libs that don't break when enabling LTO
[17:57] <aetasx> shaved about 27% of the encoding time off, pretty decent imo
[17:58] <iive> :O
[17:58] <iive> 27% total time off is enormous!
[17:59] <Daemon404> that's a very vague statement
[17:59] <Daemon404> encoding time *with what*
[17:59] <Daemon404> and compared to what
[17:59] <aetasx> libx264 placebo, non-profiled/lto vs profiled/lto
[18:00] <Daemon404> that seems VERY unlikely
[18:00] <Daemon404> most of the time is in its asm
[18:00] <wm4> yeah
[18:00] <Daemon404> to which LTO makes 0 difference
[18:00] <Daemon404> unless you did something silly like disable asm.
[18:00] <aetasx> lol no
[18:00] <wm4> how did you test?
[18:01] <Daemon404> if you perf x264, you see almost all its time spent in e.g. avx functions
[18:01] <aetasx> they're just encoding times based on what I use it for and I don't have avx sooo
[18:01] <aetasx> ;)
[18:01] <Daemon404> s/avx/sse4.2/ then
[18:01] <Daemon404> or whatever
[18:02] <aetasx> wish libspeex and SDL would LTO properly, those two I had problems with
[18:02] <Daemon404> it will also depend on what input, and if the input decoder has asm or not
[18:02] <Daemon404> in which case youre jus tbenchmarking the decoder.
[18:02] <aetasx> hey I didnt declare it scientific, its the speedup I get on my own
[18:02] <wm4> aetasx: they're also useless, so you can just skip them
[18:03] <aetasx> SDL yeah, yeah cant encode in speex without libspeex though
[18:04] <aetasx> its still used in flash crap
[18:06] <aetasx> that and I think it was gnutls or gcrypt that crashes the optimizer out when attempting it but I don't really have a need for them
[18:08] <aetasx> wm4: I did take your suggestion and just start stripping everything off Ill never need too, cut quite a few megs out of the static build
[18:31] <ubitux> so how do we reproduce the speedup?
[18:41] <aetasx> the configure script using --enable-lto and tossing in fprofile-generate/use simply works right on its own so that much is basically a handout.  configuration in ffmpeg is so much more resilient than the libs I had to deal with with the exception of maybe x264 which you can feed a sample file and it'll automate the entire profiling build
[18:45] <aetasx> it might even justify just chucking an option for it in configure considering how easy it was to do with a little alternate environments testing.  either way passing the flags is just as easy
[18:59] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:7f02fcd91774: avfilter/vf_fps: Do not drop a random subset of frames
[20:59] <ubitux> we should write down the politic of auto selection of libs
[21:00] <ubitux> because OSX stuff or x11 junk autodetected is evil
[21:00] <Daemon404> any autodetect of x11 is evil
[21:00] <Daemon404> packagers will hate you
[21:00] <Daemon404> SURE LETS DRAG IN X11 DEPS
[21:00] <ubitux> yes, i don't understand why we enabled it by default
[21:02] <ubitux> i mean, sdl is already a cancer
[21:02] <nevcairiel> osx stuff is fine, since every osx system is going to have it anyway, but x11 is stupid
[21:02] <ubitux> but we justify its probing for ffplay
[21:02] <ubitux> nevcairiel: what about cross compiling from OSX?
[21:02] <kepstin-laptop> the sdl stuff is only needed for ffplay, right?
[21:03] <ubitux> kepstin-laptop: we have a sdl device output too, which creates a dep to it too
[21:04] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:afa92907f3c6: avcodec/mjpegdec: Check escape sequence validity
[21:04] <cone-616> ffmpeg.git 03Michael Niedermayer 07master:fabbfaa09566: avcodec/mjpegdec: Check number of components for JPEG-LS
[21:06] <nevcairiel> ubitux: why would your cross-compiling env have the headers or libs for configure to succeed on these things
[21:07] <ubitux> dunno, just wondering
[21:14] <wm4> can we get a ssl lib autodetected?
[21:14] <Daemon404> no
[21:15] <Daemon404> since they are all lgplv2.1 incompatible
[21:15] <wm4> who cares
[21:15] <Daemon404> well anyone who distributes anything
[21:17] <Stefan_G> Good evening! Anybody listening at the moment? I am experiencing segfaults with a static build of ffmpeg on Ubuntu 14.04 while attempting to transcode Magic Lantern MLV files.
[21:17] <Stefan_G> This is no support request.
[21:17] <Stefan_G> I suspect a bug and would provide an MLV file for analysis
[21:18] <beastd> Stefan_G:  https://ffmpeg.org/bugreports.html
[21:18] <Stefan_G> Ack
[21:19] <Stefan_G> Is there any option without having to register somewhere? I'm starting to lose count of the sites where I had to register just to report one bug...
[21:24] <beastd> Stefan_G: You could also report the bug on the ffmpeg-user mailing list.
[21:24] <beastd> But I would not really recommend it, though it would be better than not reporting the crash. Also did you try with current git master version of FFmpeg?
[21:24] <Stefan_G> Well, ok, that's it then... There is no ffmpeg_g file in the static package I downloaded, so no gdb...
[21:24] <Stefan_G> Yeah, I downloaded a staic build package with the latest git
[21:25] <Stefan_G> ffmpeg version N-44018-g77f326d-   http://johnvansickle.com/ffmpeg/    Copyright (c) 2000-2015 the FFmpeg developers
[21:25] <Stefan_G>   built with gcc 4.9.2 (Debian 4.9.2-10)
[21:27] <beastd> Stefan_G: I think building FFmpeg yourself should not be too hard. It would also be helpful as you could test patches should it be necesarry. There are guides in the wiki: https://trac.ffmpeg.org/#CompilingFFmpeg
[21:28] <Stefan_G> You're probably right but getting involved in a full debugging of the issue would be way over my time budget at the moment. I have too much other stuff on my plate.
[21:29] <Stefan_G> And I just recently switched from SuSE to Ubuntu, so I'm not yet familiar with the distro that I could easily build stuff. Practically all dev-packages have slightly different names...
[21:29] <Stefan_G> Total nightmare.
[21:30] <beastd> Stefan_G: The packages are named in the wiki guides. and your probably do not need to many.
[21:31] <beastd> Stefan_G: Also if you report without a backtrace there is still a chance a developer could reproduce with the help of your sample file. I can understand being short on time.
[21:32] <Stefan_G> Unfortunately... I'm way over my head in projects... Everything more than git clone <repo>\n cd <dir>\n make config all install\n would be too much for me at the moment
[21:33] <Daemon404> just open a bug with the sample file
[21:33] <Daemon404> and the output
[21:33] <Daemon404> that should be enough
[21:33] <Stefan_G> Where can I put the sample file? It's nearly 2gig. MLV files tend to be huge...
[21:34] <Stefan_G> I could put it to my home server but I only have 5mbit, so download might be slow
[21:34] <Daemon404> we have an ftp but i dont know what its limits are
[21:34] <Daemon404> i you can, say, reproduce with the first MB of the file, thats finee too.
[21:35] <Stefan_G> If I would know the internals of MLV files...
[21:35] <beastd> description about the ftp is on  https://ffmpeg.org/bugreports.html
[21:36] <Stefan_G> that file is only 30 secs of video!
[21:36] <Daemon404> beastd, good point.
[21:38] <beastd> if the crash is not reproducible with the small file, iirc uploading larger files worked fine in the past.
[21:48] <Timothy_Gu> Stefan_G: your version looks odd. Isn't FFmpeg over N-65000 now?
[21:48] <Stefan_G> I have no idea. But the crash also occurs with the stable releases
[21:50] <Timothy_Gu> Stefan_G: See http://ffmpeg.org/bugreports.html#SubmittingSampleMedia
[21:51] <Timothy_Gu> Stefan_G: try to dd the first maybe 10MB of the file and test if it crashes, and if it does just uploading the first 10MB of the file is fine
[21:52] <ubitux> i'm too tired to try to explain to carl about libxcb, does anyone care to do it?
[21:54] <Stefan_G> Yeah, it does...
[21:54] <Stefan_G> Somebody can delete my other two attempts of uploading the files...
[21:55] <cone-616> ffmpeg.git 03Seppo Tomperi 07master:74d7faf4001e: hevcdsp: separated sao edge filter and pixel restore funcs
[21:55] <cone-616> ffmpeg.git 03Seppo Tomperi 07master:4386e1fd94cd: hevcdsp: simplified sao_edge_filter
[21:55] <cone-616> ffmpeg.git 03James Almer 07master:121ccc336551: hevcdsp: further simplify sao_edge_filter
[21:55] <cone-616> ffmpeg.git 03James Almer 07master:7457afc64d5f: hevcdsp: replace the SAOParams struct parameter from sao_edge_filter
[21:55] <Stefan_G> The correct file name is bug_4296_sample_file_segfault_MLV_format_first10MB.mlv
[21:57] <Stefan_G> Timothy_Gu: The readme says:
[21:57] <Stefan_G>                 build: ffmpeg-git-20150204-64bit-static.tar.xz
[21:57] <Stefan_G>               version: N-44018-g77f326d
[23:29] <aetasx> fate verifies you have support for something built-in before running a test against it yes?  (Im assuming so since theres that randomized build configure option)
[23:36] <nevcairiel> yes it should
[23:36] <nevcairiel> if not thats a bug
[23:45] <aetasx> thanks
[23:56] <aetasx> whats your guys thoughts on letting fate run its tests against a fprofile-generate build as like a generic profiling option?  I was going to play around with that idea and see what it turns out
[00:00] --- Thu Feb  5 2015