[Ffmpeg-devel-irc] ffmpeg-devel.log.20130319

[burek]

[00:07] <cone-274> ffmpeg.git 03Thierry Foucu 07master:823efd9286b3: Fix termination of mov_ch_layouts_wav array lookup
[00:07] <cone-274> ffmpeg.git 03Michael Niedermayer 07master:0e02a5193f80: segment: copy metadata
[00:49] <cone-274> ffmpeg.git 03Michael Niedermayer 07master:6128d33946fc: AVFrame.pkt_dts: improve docs
[01:26] <ubitux> saste: any idea why asrc_sine when we have aevalsrc?
[01:26] <ubitux> (except bit-exactness)
[01:26] <saste> ubitux, performance
[01:26] <saste> and bit-exactness
[01:27] <ubitux> performance? does it really matter here?
[01:27] <saste> bit-exactness
[01:27] <ubitux> :D
[01:27] <ubitux> ok ok
[01:27] <saste> i think the point is to use it for tests
[01:27] <ubitux> to replace audiogen?
[01:28] <saste> possible, or for more fancy tests
[01:28] <Daemon404> replacing audiogen is a bad idea...
[01:28] <Daemon404> makes asynth rely on lavfi being built
[01:28] <ubitux> i'm not sure why we didn't add a s16 output to aevalsrc though
[01:29] <saste> yes that can be done, but i think you still have possible non bit-exact double -> int conversions
[01:29] <Daemon404> write a fp emulator like lagarith did =p
[01:30] Action: ubitux will continue to play with sin(2*PI*t*400)*exp(-4*mod(t\,1))
[01:31] <ubitux> 440*
[01:35] <michaelni> if someone wants bitexact floats look at libavutil/softfloat
[01:41] <saste> ubitux, still there is the performance issue, even in case we plug eval to bit-exact softfloat
[01:42] <saste> since the periodic values are computed just once
[01:42] <saste> but i really knows nothing, better to ask on list
[01:42] <cone-274> ffmpeg.git 03Carl Eugen Hoyos 07master:e4f5aa5e8d5c: Inform the user if a pix_fmt != yuv420p was chosen for x264 encoding.
[01:42] <saste> *know
[02:15] <llogan> ubitux: "Something got wrong" in normalize.sh sounds weird
[02:17] <ubitux> yeah well i'm not sure about how to raise the errors properly
[02:17] <ubitux> any suggestion welcome
[02:17] <ubitux> note that's just a demo script
[02:19] <llogan> a comment at top indicating it is an example might help users who may think otherwise.
[02:20] <ubitux> ok
[02:21] Action: llogan gives llogan Worthless Nit Lord Award.
[02:32] <cone-274> ffmpeg.git 03Carl Eugen Hoyos 07master:55bb8f4589a4: Fix Makefile dependency for the dvd_nav parser.
[02:32] <cone-274> ffmpeg.git 03Hendrik Leppkes 07master:97f8c6e14753: pthread: update all get_buffer checks for get_buffer2
[03:20] <cone-274> ffmpeg.git 03Matthieu Bouron 07master:69e93792d607: lavf/mxfdec: read field dominance flag
[03:23] <michaelni> Tjoppen, mateo` Whats the status of the patchset: (i applied the first as it was trivial ...) "Matthieu Bouron (0.8K) [FFmpeg-devel] Support MXF separate fields layout with interlaced j2k"
[03:23] <michaelni> ?
[03:43] <llogan> i thought it was time to screw up a simple ticket invalidation
[08:55] <hnrayer> Hello
[08:56] <hnrayer> These are some of the free projects and programs known to incorporate work from FFmpeg. 
[08:56] <hnrayer> in this page:http://ffmpeg.org/projects.html
[08:57] <hnrayer> anybody knows which one is used c/c++ to develop?
[08:58] <hnrayer> I am a newcomer here myself,I came from China,forgive me my poor English.
[09:01] <hnrayer> can you see my post?
[09:14] <av500> yes
[09:34] <Tjoppen> michaelni: oh yeah, just saw that
[09:42] <mateo`> michaelni: the complete patch list could be found here. It needs more testing and more samples. (https://github.com/mbouron/FFmpeg/tree/mxfdec_j2ki)
[09:43] <mateo`> Tjoppen: if you can have a look at it
[09:43] <mateo`> the samples provided by Dave could be decoded with openjpeg 1.5.1
[10:28] <durandal_1707> what anton changed in lavfi options?
[10:28] <durandal_1707> do we need again another branch for merge?
[10:30] <mateo`> michaelni, Tjoppen: the libopenjpegdec patch is broken after rebase, i should fix it first.
[10:43] <cone-669> ffmpeg.git 03Paul B Mahol 07master:4d16ba2dd878: ffplay: silence warning of mixed enumeration types
[10:49] <durandal_1707> michaelni: can i get informed when new minor release happen?
[11:02] <nizvuka> hi
[11:03] <nizvuka> wmalossless codec asks for sample
[11:04] <michaelni> durandal_1707, which release branch ? but in principle there should be an announcement before each release
[11:05] <michaelni> nizvuka, you have a wmalossless that does not decode correctly ?
[11:05] <nizvuka> yes
[11:05] <durandal_1707> michaelni: monor ones, like 1.0.X
[11:05] <durandal_1707> aka security fixes
[11:06] <nizvuka> Unsupported extradata size
[11:07] <durandal_1707> anyway decoder is not full featured and there are already some bug reports opened for it
[11:08] <nizvuka> just want to help here is sample ftp://upload.ffmpeg.org/MPlayer/incoming/wmall.wma
[11:13] <nizvuka> http://pastie.org/6624427
[11:17] <nizvuka> sorry, not wmall.wma, sample is lossless.wma and description - wmall.txt
[11:19] <durandal_1707> nizvuka: and decoded output sounds wrong?
[11:19] <durandal_1707> does other players that do not use libavcodec/libavformat/FFmpeg exibit same problem?
[11:19] <durandal_1707> it may be known asf demuxer bug(s)
[11:20] <durandal_1707> previous questions still stands
[11:25] <nizvuka> durandal_1707, decoded output sounds good, other players handle it without problem
[11:26] <Tjoppen> mateo`: k
[11:58] <durandal_1707> is there filter which just duplicates one frame n times?
[11:58] <JEEB> the one that the frame rate conversion filter uses?
[11:59] <JEEB> because by setting the -r x2 the original rate you get doubled frames f.ex.
[11:59] <nevcairiel> i dont think that works if you have only one frame =p
[11:59] <durandal_1707> lol, i think i'm looking for split fiter
[11:59] <durandal_1707> *filter
[12:07] <durandal_1707> *filte3
[12:07] <durandal_1707> ^undo
[12:24] <ubitux> durandal_1707: trying to port harddup filter?
[12:24] <Bor0> why do I get h264: no picture after attempting to play a bit larger .mp4 file, but it works for small files? (not even sure that large/small is the issue, maybe something else?)
[12:25] <Bor0> or to be more precise, how can I extend verbosity of "h264: no picture" ?
[12:25] <durandal_1707> ubitux: what is that? note I dislike anime pseudo-filters
[12:26] <ubitux> mp=harddup
[12:26] <ubitux> (why are you talking about animes? :P)
[12:26] <durandal_1707> animes is only thing vapour software is used for
[12:27] <ubitux> i was just trying to answer your dup frame filter thing :p
[12:27] <durandal_1707> ubitux: doesn't split do same?
[12:28] <durandal_1707> and i think mp=harddup does nothing
[12:28] <ubitux> no idea
[12:31] <Bor0> logs say it crashes at queue_picture(), at "wait until we have space to put a new picture"
[12:31] <durandal_1707> Bor0: use #ffmpeg
[12:31] <durandal_1707> this channel is for development
[13:11] <durandal_1707> michaelni: our dca encoder is just more/less ripof/clone of older dcaenc
[13:14] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:8d0da20ca6c7: avidec: fix duration and bitrate of truncated files
[13:32] <cone-669> ffmpeg.git 03chinshou 07master:b619a5485362: alac: Fix order of values in comment
[14:04] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:f1c3d8b3444c: avutil/add_to_pool: remove unused assgnment
[14:04] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:ad9856719864: mpegvideo_enc: remove stray semicolon
[14:10] <michaelni> ubitux, see CID991855  (vf_curves)
[14:11] <michaelni> also CID991854 (ffprobe)
[14:11] Action: michaelni looks around and sees no saste
[14:13] <ubitux> michaelni: will look, thx
[14:19] <cone-669> ffmpeg.git 03Clément BSsch 07master:dec9800c9dc9: lavf: do not make a hard dep on internal subtitles helpers.
[14:20] <Compn> has there ever been a report on a person being compromised via ffmpeg ?
[14:20] <Compn> seem to get a ton of these vulns. i'd like to know if anyone was harmed
[14:20] <av500> vulns are not about being harmed
[14:23] <durandal_1707> also hard dep on riff is funny
[14:24] <nevcairiel> thats mostly the fourcc list used by oh so many formats
[14:25] <durandal_1707> than ditch id3*
[14:25] <ubitux> michaelni: looks like a false positive, i'll mark as such
[14:35] <michaelni> ubitux, ok
[14:35] <ubitux> i'll add some fate coverage of the filter though
[14:35] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:fdda9b440c0a: matroska_parse_frame: fix memleak
[14:35] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:2c993fece28b: audio_get_buffer: add assert to ensure nb_extended_buf has a safe value
[14:41] <cone-669> ffmpeg.git 03Clément BSsch 07master:21aa4fafcd70: fate: add ebur128 metadata test.
[14:44] <durandal_1707> ubitux: did you find my transmedia filter useful?
[14:45] <ubitux> transmedia filter? jf filter?
[14:45] <durandal_1707> yes
[14:46] <ubitux> i don't know if it's useful, but it looked fun :)
[14:46] <ubitux> and i like fun things
[14:50] <durandal_1707> i dunno if i need to pass samples to some additional filter
[14:53] <durandal_1707> there is also rotated variant - but that is trivial
[14:53] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:1e41b8cd072e: hpeldsp: Increase put_no_rnd_pixels_tab[][] size
[14:53] <durandal_1707> and polar one, but can't find how it is actually calculated
[15:04] <ubitux> michaelni: why not 3 instead of 2 here?
[15:05] <ubitux> (or instead of the 4 you put)
[15:06] <durandal_1707> why not 42?
[15:06] <michaelni> ubitux, it was 4 originally, (for all the blocksizes of h264) we could put 3 there and also for the other related tables yes
[15:07] <ubitux> ok
[15:08] <durandal_1707> what functions audio visualizations "plugins" mostly use?
[15:15] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:5f476cfe6e38: avformat/tee: add forgotten ret assignment
[15:34] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:feeb8ca56dc0: vp56: Check av_frame_ref() return code
[16:08] <durandal_1707> doc/RELEASE_NOTES is still containing obsolete info
[16:13] <durandal_1707> ubitux: harddup is yet another filter usefull only with mencoder
[16:20] <durandal_1707> michaelni: fine to remove it?
[16:21] <durandal_1707> posting this to ml is waste of energy
[16:21] <cone-669> ffmpeg.git 03Clément BSsch 07master:bf0e44fa67e9: fate: add curves filter test.
[16:25] <durandal_1707> i give 10 minutes for comments
[16:41] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:d275f6eda81b: doc: update release_notes by taking the ones from release/1.2
[16:41] <cone-669> ffmpeg.git 03Michael Niedermayer 07master:4ce03a95e27d: lmlm4: check packet_size against lower limit too
[16:42] <michaelni> durandal_1707, possibly ok
[16:43] <durandal_1707> why possibly? why not "should be ok"?
[16:48] <michaelni> possibly ok because its not posted to the ML for wider scruntini and review.
[16:49] <durandal_1707> i posted it
[16:58] <cone-669> ffmpeg.git 03Nedeljko Babic 07master:d23aefa4984d: mips: inline assembly clobber list polishing
[17:28] <saste> ubitux: ffprobe  -of compact=p=0 -show_entries frame=pkt_pts,metadata -bitexact -f lavfi testsrc
[17:28] <saste> works fine here
[17:28] <saste> no need to -show_frames
[17:30] <ubitux> feel free to update the fate test; i removed the -show_frames and it broke the test
[17:30] <ubitux> i didn't look closer
[17:31] <ubitux> saste: btw interested in a ffprobe crash?
[17:31] <saste> ubitux, sure
[17:31] <ubitux> saste: compact=p=0:k=0 was crashing the other day
[17:32] <ubitux> (k option not found because it is nk)
[17:32] <ubitux> (but trigger the crash)
[17:35] <cone-669> ffmpeg.git 03Stefano Sabatini 07master:608bccf47054: lavf/avformat.h: fix a few typos in AVFormatContext:avoid_negative_ts doxy
[17:35] <cone-669> ffmpeg.git 03Stefano Sabatini 07master:2bac6b6f5ee8: doc/ffmpeg-formats: reword and extend documentation for avoid_negative_ts option
[17:35] <cone-669> ffmpeg.git 03Stefano Sabatini 07master:18f43de96497: doc/ffmpeg,ffmpeg-formats: avoid to mention ffmpeg options from formats.texi
[17:39] <cone-669> ffmpeg.git 03Clément BSsch 07master:879ca313c02f: fate: add missing filter dep in metadata tests.
[17:43] <ubitux> saste: doc/developer.texi:651: @ref reference to nonexistent node `Regression Tests'
[17:43] <ubitux> do you also have that warning?
[17:43] <ubitux> (i don't get it since there are other similar ref+anchor)
[17:43] <saste> ubitux, yes since the last merge
[17:43] <saste> if @ref mention an internal anchor it is fine
[17:43] <saste> if it is referencing an anchor in another page can't work
[17:43] <ubitux> it is internal afaict
[17:44] <saste> ubitux, [compact @ 0x2af8d20] Key 'k' not found.
[17:45] <saste> no crash
[17:45] <ubitux> oh i get the warn.
[17:45] <ubitux> saste: will retry in a moment
[17:46] <cone-669> ffmpeg.git 03Clément BSsch 07master:1c3b4281148c: doc/developer: fix typo in regression tests ref.
[17:47] <ubitux> saste: try with valgrind
[17:48] <saste> oh you just waster 2 minutes of my life
[17:48] <saste> *wasted
[17:49] <ubitux> the doc thing? :)
[17:49] <saste> Regression *T*est bug
[17:49] <ubitux> :)
[17:50] <saste>  jump or move depends on uninitialised value(s)
[17:50] <saste> willing to chase it?
[17:52] <ubitux> nope
[17:52] <ubitux> :)
[17:52] <ubitux> michael was pointing out a problem spotted by coverity in ffprobe
[17:52] <ubitux> 'might be related
[17:56] <saste> av_bprint_finalize i see
[18:01] <cone-669> ffmpeg.git 03Stefano Sabatini 07master:e292d75113fd: ffprobe: set writer context to 0 after allocation
[18:16] <cone-669> ffmpeg.git 03Compn 07master:b964b0fc5c63: changelog: 10% faster aac encoding on x86 and MIPS
[18:17] <durandal_1707> Compn: huh, really?
[18:19] <cone-669> ffmpeg.git 03Paul B Mahol 07master:17c1881d5823: lavfi: remove harddup libmpcodecs wrapper
[18:20] <Compn> durandal_1707 : yea, mips guy got 10% speedup with that little change to aacpsy :D
[18:20] <Compn> course, benchmarks would be nice to have as well
[18:20] <durandal_1707> but on x86?
[18:20] <Compn> it wasnt mips specific :P
[18:20] <saste> Compn, also aac -> AAC
[18:21] <Compn> damn
[18:21] <Compn> i knew i missed soemthing :P
[18:22] Action: Compn afk
[18:24] <Compn> [libav-devel] [PATCH] tinterlace backported  
[18:24] <Compn> neat
[18:25] <Compn> oh he rewrote it too
[18:25] <Compn> in a later patch
[18:25] Action: Compn afk for reals
[20:07] <ubitux> any idea what's wrong with openbsd & gif?
[20:21] <ubitux> also, it's curious the audio metadata memleak but not the video one
[20:32] <saste> regarding monolithic docs, how should we include them in the web page?
[20:33] <ubitux> what do you mean?
[20:33] <ubitux> the title of the href?
[20:33] <saste> i'm adding an option --enable-monolithic-tools-doc
[20:34] <saste> which make tools page include all the relevant stuff for the tool page
[20:34] <saste> probably we should add a link in the documentation page on the server to the monolithic page as well
[20:34] <ubitux> why don't you just add a ffmpeg-all manpage?
[20:34] <ubitux> that's what zsh does for instance
[20:35] <saste> ubitux, what do you mean?
[20:35] <saste> what should it contain?
[20:35] <ubitux> everything?
[20:35] <saste> including tools?
[20:36] <ubitux> yeah
[20:36] <saste> no
[20:36] <saste> that would be a mess
[20:36] <ubitux> the idea is to be able to /option in everything
[20:36] <ubitux> that's what users want.
[20:37] <ubitux> i'm not personally asking for it
[20:37] <saste> have a look at ffmpeg.texi
[20:37] <ubitux> but at least burek and derek were asking for it
[20:37] <saste> it is organized as a collection of chapters
[20:37] <ubitux> oh it will sucks, no doubt about it
[20:37] <saste> now if you put all chapters together you get a flat list of chapter, with no organization
[20:38] <saste> at some point the ffmpeg example chapter end, and the ffplay synopsis start
[20:38] <saste> that's bad
[20:38] <saste> what i am doing right now (already implemented, basically)
[20:38] <saste> i build a monolithic ff* page including all the other subsystems
[20:38] <ubitux> (you can add a figlet <toolname> on top of each)
[20:39] <ubitux> ok
[20:39] <saste> figlet??
[20:39] <saste> we are talking about texinfo here
[20:39] <ubitux> it was a bad joke :)
[20:40] <ubitux> you can use http://caca.zoy.org/wiki/toilet if you don't like figlet
[20:43] <ubitux> saste: though, why the option?
[20:43] <ubitux> will it be enabled by default?
[20:43] <saste> ubitux, no
[20:43] <saste> i strongly oppose to that
[20:44] <ubitux> so that's just for online documentation?
[20:44] <saste> ubitux, no, insane users can still do --enable-monolithic-tools-doc if they prefer that way
[20:45] <ubitux> right, online documentation only
[20:45] <ubitux> users complaining about not being able to find an option in the doc won't enable that option if they are too lazy to git grep
[20:46] <saste> lazy users never relies on local docs
[20:46] <saste> they try the first command which seems fairly relevant on internet, and complains if that doesn't work
[20:47] <llogan> fear will bring these users in line.
[20:47] <saste> *they try the first command which seems fairly relevant found on internet, and complain if that doesn't work
[20:58] <ubitux> http://ffmpeg.org/pipermail/ffmpeg-devel/2013-February/139421.html  dead patch?
[20:58] <ubitux> Nicolas was asking for you saste ^
[20:59] <saste> yes forgotten after the mergestorm
[21:26] <cone-134> ffmpeg.git 03Clément BSsch 07master:d4982b115dcd: lavd/dshow: fix style and typo recently introduced.
[22:44] <ubitux> ok leak fixed finally
[22:46] <ubitux> seems like a problem from one of the monster commit again
[23:15] <Daemon404> i am honored to have a new git tag mae for me
[23:17] <ubitux> it will soon become a standard if you continue your efforts
[23:18] <ubitux> btw that's not the first time
[23:18] <ubitux> it got used already in 00cae8675
[23:21] <saste> ubitux, what's wrong with s/-show_frames//?
[23:21] <saste> do you still have issues with it?
[23:23] <ubitux> saste: as i said, try to remove it from the test, you'll see it doesn't work
[23:24] <ubitux> maybe because metadata/tags are special
[23:24] <saste> ubitux, what happens?
[23:24] <ubitux> saste: metadata are not displayed
[23:24] <saste> ubitux, sure
[23:25] <saste> -show_entries frame=metadata,pts ?
[23:25] <saste> basically -show_frames should be equivalent to -show_entries frame
[23:26] <ubitux> same, doesn't work with metadata in first position
[23:27] <ubitux> pts are always displayed
[23:27] <ubitux> i mean as long as i set them of course
[23:38] <saste> ubitux, -show_entries frame=pts,tags ?
[23:39] <kangstarr> hi everyone
[23:40] <ubitux> saste: nope, doesn't help
[23:40] <kangstarr> does anyone know if or when ffmpeg will support rec. 2020?
[23:41] <ubitux> never heard of it, and can't find a related ticket
[23:42] <cone-134> ffmpeg.git 03Stefano Sabatini 07master:33b054986767: lavc/utils: add VP6F hack for setting correct video size in stream
[23:49] <llogan> http://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.2020-0-201208-I!!PDF-E.pdf
[23:50] <ubitux> adding a new table in vf colormatrix might be enough
[23:50] <kangstarr> ubitux: ah, ok, i'm new to all this so I'll figure out how to open an issue
[23:50] <kangstarr> basically UltraHD (4K and 8K) supports Rec. 2020
[23:50] <llogan> kangstarr: also provide a sample if possible
[23:50] <ubitux> https://ffmpeg.org/trac/ffmpeg
[23:51] <kangstarr> Rec. 2020 defines a much wider color gamut than Rec. 709
[23:51] <kangstarr> and also defines larger resolutions and higher frame rates
[23:51] <llogan> at least no interlacing
[23:51] <kangstarr> thank god
[23:51] <ubitux> patching libavfilter/vf_colormatrix.c should be do-able
[23:52] <ubitux> coeff are on top of the file
[23:52] <saste> ubitux: -show_entries frame=pkt_pts:frame_tags
[23:52] <ubitux> then you need to add a string and here you go.
[23:52] <ubitux> saste: haha
[23:52] <saste> ubitux: -show_entries frame=pkt_pts:tags -> also works here
[23:52] <ubitux> actually no, doesn't work
[23:53] <kangstarr> ubitux: i only see a link to submit a bug report....?
[23:53] <ubitux> saste: are you able to change the fate test without any breakage?
[23:53] <ubitux> kangstarr: new ticket, type: enhancement
[23:54] <llogan> priority: wish
[23:54] <llogan> also you'll have to register first: https://ffmpeg.org/trac/ffmpeg/register
[23:57] <saste> ubitux, funny
[23:58] <saste> tags selects also stream_tags, and format_tags
[23:58] <saste> that's why it adds some empty lines
[23:58] <saste> it is in line with expected behavior
[23:58] <saste> so at the end the correct "nice" syntax is -show_entries frame=pkt_pts:frame_tags
[00:00] --- Wed Mar 20 2013