[Ffmpeg-devel-irc] ffmpeg-devel.log.20121001

[burek]

[00:42] <ubitux> did anyone understand what Ronald was so angry about?
[00:42] <ubitux> (honestly)
[00:43] <ubitux> cbsrobot_:     I:         -23.4 LUFS
[00:43] <ubitux> you might want to test the latest version in my branch
[00:43] <cbsrobot_> ubitux: ok will do
[00:44] <ubitux> i need to improve the code a little
[00:44] <ubitux> and likely do some similar changes to the LRA
[00:45] <ubitux> so i'll push this week
[00:45] <ubitux> and will work later on the true peak
[01:15] <Compn> ubitux : libav side never could understand michael's english
[01:15] <saste> I guess the "right way" to implement the XYZ -> RGB transform would be in libswscale
[01:16] <Compn> its either a pixfmt or a raw decoder , yep
[01:16] <Compn> conversion does not belong in filter :P
[01:16] <saste> yes a raw decoder could also work
[01:16] <saste> i suppose adding another colorspace in libswcale would be painful
[01:18] <michaelni> iam planing to make it easy to add colorspaces
[01:18] <Daemon404> youve got a lot to learn about social skills^W^Wmailing list trolling guys
[01:19] <ubitux> hey i'm not trolling :(
[01:19] <michaelni> Daemon404, did you understand why ronald is angry ?
[01:19] <Daemon404> the email wasnt very coherent...
[01:19] <ubitux> i think he just wanted an occasion to express his frustration about something, but took a bad thread and misunderstood it
[01:20] <saste> yes ^^
[01:20] <Daemon404> that said
[01:20] <Daemon404> i am not touching this with a 10 foot pole
[01:23] Action: michaelni locks Daemon404 in a small box with "it" and burries the box 10 meters below quick drying cement
[01:23] <ubitux> :D
[01:23] <ubitux> this is not very social michaelni 
[01:24] <Daemon404> it's a fun fact that party guests dont like being murdered
[01:25] Action: michaelni pulls the box quickly back out before the cement dries and lets Daemon404 back out
[01:26] <Daemon404> /nick Houdini
[01:26] <ubitux> :)
[01:27] <ubitux> +2 social skills for michaelni 
[01:27] <michaelni> \o/
[01:27] <cbsrobot_> michaelni: but now Daemon404 will have lifelong trauma !
[01:28] <ubitux> anyway, 'night ppl :)
[01:28] <michaelni> night ubitux 
[01:29] Action: michaelni goes back to coding
[01:39] <ohsix> huge letters too
[02:16] <Compn> michaelni : now is good time to ask ronald on the list if he has any suggestions or wanted features in swscale
[02:16] <Compn> extend the olive branch of discussion :)
[02:16] <Compn> ask him if there are any patches which should be merged
[02:16] <Compn> or any you missed
[02:18] <Compn> and it is customary culture of some english countries to apologize for any misunderstandings of statements made. of course , this feels like 2011 all over again
[02:18] Action: Compn goes to do something else before brain explodes
[02:32] <iive> he is frustrated by the news of the 1.0 release.
[02:35] <iive> I guess he is also frustrated that ffmpeg is releasing more of the new libav code, than libav itself.
[02:38] <Compn> maybe hes stuck between a rock and a google.
[03:02] <iive> I have no idea about that.
[03:17] <iive> He was such a promising nice guy. Then the takeover happened and he turned into vicious angry master. It is heart wrenching.
[06:50] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/K_18JA
[06:50] <KGB> [FFmpeg/master] movenc: fix edit list for the case of negative pts. - Michael Niedermayer
[10:42] <saste> how can i fill gaps in an audio stream?
[10:43] <saste> i could use aresample, but IIRC it only fills the initial gap
[10:43] <durandal_1707> is there a way to skip n frames from demuxer?
[10:46] <ubitux> michaelni: i kind of find a way to debug the race
[10:46] <ubitux> found*
[10:48] <saste> would it make sense to implement a timepad filter for that?
[10:48] <saste> we have many filters which only work with continuous streams (e.g. amerge)
[10:50] <ubitux> saste: you want to write some sounds generation interpolation between two points? 
[10:50] <saste> ubitux: a silence pad would be enough
[10:51] <saste> also i see there is a asyncts, and i wonder if there is a corresponding option in ffmpeg/swr
[10:52] <ubitux> it's done manually in ffmpeg.c
[10:52] <ubitux> asyncts insert is disabled
[10:54] <saste> ubitux, in the ffmpeg manual -async is deprecated, in favor of asyncts
[10:55] <ubitux> according to michaelni something is wrong with it
[10:55] <ubitux> so it's been disabled in ffmpeg
[10:55] <ubitux> iirc
[10:55] <saste> what the current behavior with regards to audio gaps?
[10:56] <ubitux> or maybe it was changed
[10:56] <saste> if i don't specify special options, and i don't insert asyncts/aresample?
[11:00] <ubitux> http://pastie.org/4890153 got you, bitch
[11:20] <ubitux> it seems the race is around the H264Context->ref_count[2]
[11:21] <ubitux> (frame context)
[12:03] <ubitux> and maybe another (?) around ff_h264_lps_range in ff_init_cabac_states (cabac.c:146)
[12:13] <burek> where does cia get its input from?
[12:13] <burek> (the bot)
[12:15] <av500> cia is dead
[12:16] <burek> btw i've seen cia bots all over freenode
[12:16] <burek> is that something provided by freenode or what
[12:16] <av500> no
[12:16] <av500> cia.vc
[12:17] <av500> but it's dead atm
[12:17] <av500> there is FBI and KGB though
[12:18] <durandal_1707> KGB is sloow (not in real time)
[12:20] <av500> durandal_1707: write a new one and call it FSB
[12:20] <av500> or MOSSAD
[12:21] <divVerent> or... Verfassungsschutz
[12:22] <cbsrobot_> call it NDB - they seem to leak fast
[12:41] <burek> or wikileaks :)
[12:41] <burek> g2g bbl
[13:14] <michaelni> saste, aresample can fill gaps in the middle, see the AVOption table in libswresample/swresample.c
[13:14] <michaelni> for parameter, i dunno if its in the docs
[13:19] <saste> michaelni, ok, i'll try to write a document explaining the various options
[13:22] <michaelni> thx
[15:00] <burek> i'm back
[15:00] <burek> michaelni, where does cia get its input from?
[15:13] <michaelni> burek, dunno exactly but there must be a hook on the ffmpeg @ vlc git repo
[15:13] <michaelni> that pushed  something toward CIA
[15:13] <j-b> I think so too
[15:14] <j-b> Btw, how are we supposed to set the cpu flags those days?
[15:16] <michaelni> for ffmpeg via command line -cpuflags, for the libs via av_get_cpu_flags() and av_force_cpu_flags()
[15:17] <michaelni> theres also av_parse_cpu_caps()
[15:17] <michaelni> to parse a string 
[15:17] <michaelni> of things like +mmx-sse2
[15:19] <j-b> get_get_cpu_flags seems to be the opposite
[15:20] <j-b> parse takes a string, not a flag
[15:21] <j-b> av_set_cpu_flags_mask is deprecated
[15:22] <michaelni> ... and av_force_cpu_flags()
[15:22] <j-b> ok
[15:23] <j-b> great, a unique API...
[15:23] <j-b> thx michaelni 
[15:46] <cbsrobot_> ubitux: the summary is nice
[15:52] <ubitux> :)
[15:52] <ubitux> results should be closer to what you expect
[15:52] <cbsrobot_> yeah looks great
[15:52] <ubitux> i still want to improve the precision, and something still looks fishy with the LRA so i need to dig a little more
[15:53] <ubitux> but i'll push soon a similar version
[15:53] <cbsrobot_> I could also crosscheck with another tool - just to make sure it's correct
[15:53] <ubitux> feel free to :)
[15:53] <cbsrobot_> is the threshold ajustable ?
[15:54] <ubitux> no, the specs impose a value
[15:54] <ubitux> you're refering to the I and LRA gate threshold, right? 
[15:54] <cbsrobot_> yes - so maybe it could be omited in the summary - no ?
[15:55] <ubitux> ah this one
[15:55] <cbsrobot_> hmmm, it changes from file to file
[15:55] <ubitux> this threshold is moving over the analysis
[15:55] <cbsrobot_> ah ok
[15:55] <ubitux> it's based on hardcoded thresholds
[15:56] <ubitux> (-10 for I, -20 for LRA iirc)
[15:56] <ubitux> it's the "relative" threshold :p
[15:56] <ubitux> it might not be useful
[15:57] <ubitux> but i wanted to print it anyway otherwise the "sections" would have been empty ;)
[15:57] <cbsrobot_> ha
[15:57] <cbsrobot_> ok
[16:03] <cbsrobot_> nit: maybe the stats could also be printed depending on the loglevel
[16:05] <ubitux> they are in visual mode
[16:05] <ubitux> (except the summary)
[16:34] <KGB> [FFmpeg] michaelni pushed 9 new commits to master: http://git.io/k-VrEg
[16:34] <KGB> [FFmpeg/master] ac3dec: ensure get_buffer() gets a buffer for the correct number of channels - Justin Ruggles
[16:34] <KGB> [FFmpeg/master] smacker: read escape codes in single get_bits() call - Paul B Mahol
[16:34] <KGB> [FFmpeg/master] Remove some silly disabled code. - Diego Biurrun
[17:47] <Daemon404> oh saste's code went in
[17:48] Action: Daemon404 updates his attached_pic logi c
[17:54] <ubitux> mmmh
[17:56] <ubitux> "The FFmpeg/libav developpers were begging for us to drop libpostproc"
[17:56] <ubitux> who from the FFmpeg developpers?
[17:56] <ubitux> !
[17:57] <j-b> see my answer
[17:57] <av500> where?
[17:57] <ubitux> vlc-devel
[17:58] <av500> is that where the drama moved to?
[17:59] <Daemon404> oh boy whats this
[17:59] Action: Daemon404 reads
[17:59] <ubitux> :)
[18:00] <Daemon404> i'd be interested who even uses postproc in vlc
[18:00] <Daemon404> or at all in anything
[18:02] Action: Daemon404 should sync libpostproc.git soon to build with msvc like ffmpeg's
[18:17] <KGB> [FFmpeg] michaelni pushed 2 new commits to master: http://git.io/YRrSwQ
[18:17] <KGB> [FFmpeg/master] cpu: improve av_get_cpu_flags() doxy - Michael Niedermayer
[18:17] <KGB> [FFmpeg/master] jpeglsdec: move pict debug log under correct if() - Michael Niedermayer
[18:18] <av500> j-b: bah, no drama at all in that thread
[18:18] <av500> you can do better
[18:19] <j-b> of course
[19:12] <iive> btw, what does "not required by default" means in the vlc configure context? that configure would not die refusing to continue without it, or that it won't be enabled even if present on the system?
[20:38] <MP4_maniac> michaelni: do you have any plan to define Opus in MOV like this? http://wiki.xiph.org/Oggless
[20:43] <michaelni> MP4_maniac, no but maybe ask nicolas he might be interrested to work on it
[20:47] <MP4_maniac> michaelni: where is nicolas on irc?
[20:50] <michaelni> i dont think he is on IRC
[20:51] <MP4_maniac> he is on ML only?
[20:52] <michaelni> yes i think so
[20:59] <MP4_maniac> michaelni: may i ask him or post topic about opus in mov on ffmpeg-devel ml?
[20:59] <michaelni> sure
[20:59] <MP4_maniac> k
[21:40] <ubitux> where is saste? :(
[21:49] <nyuhu> speaking of saste, in my patch review he told me that it would be useful to write a test, how does it work in ffmpeg ?
[21:50] <ubitux> we mostly use "the old way" to cover quite some cases using tests/lavfi-regression.sh
[21:50] <ubitux> "the new way" is to add a test in tests/fate/filter.mak
[21:50] <ubitux> but it's less effective
[21:50] <nyuhu> I see :o
[21:51] <ubitux> note that adding a test in tests/lavfi-regression.sh will need you to run configure again and stuff like that
[21:51] <nyuhu> ok
[21:52] <ubitux> nyuhu: check 5780f9bbd5668a372eac63dddcb4cda06edfaed5
[21:52] <ubitux> or 39b0d40d9219b895dbd1dd5ba5873f8594360750
[21:53] <ubitux> nyuhu: did you ever run fate?
[21:53] <nyuhu> no :x
[21:53] <ubitux> do you have the samples?
[21:53] <nyuhu> yes
[21:53] <ubitux> make fate-rsync SAMPLES=... ?
[21:53] <nyuhu> yep
[21:54] <ubitux> then well, just make fate -j20 SAMPLES=...
[21:54] <nyuhu> oki
[21:54] <ubitux> (without SAMPLES=... if you have --samples=... in your configure line)
[22:27] <ubitux> KGB :(
[22:27] <ubitux> CIA :(
[22:27] <ubitux> ffbi :(
[22:31] <gnafu> Bots, I am disappoint.
[22:36] <ubitux> ok now, vobsub demuxer!
[22:36] <ubitux> or maybe i should just watch some animes
[22:36] Action: JEEB just checked the next cour's shows
[22:37] <ubitux> dunno what to watch..
[22:37] <ubitux> i think i have some classics still unwatched
[22:38] <JEEB> I have a couple of older shows around to watch atm: RahXephon, ARIA, Kidou senkan nadeshiko
[22:40] <ubitux> the first two seasons of Aria are nice, and the last one is insanely good :)
[22:40] <JEEB> also I need to watch Gunslinger Girl
[22:40] <iive> JEEB: i couldn't watch more than 1-2 episodes of nadeshiko. but there are people who like it.
[22:40] <ubitux> kind of nice, but way below Noir or Phantom IMO
[22:41] <ubitux> saste \o/
[22:41] <JEEB> Phantom I watched until ep10 or so if you mean Phantom of the Inferno
[22:41] <saste> EBU R.128 in, going to try it
[22:41] <iive> aria is slice of life. rahxephon is simply classic.
[22:42] <JEEB> yeah, also aria's manga is just pretty <3
[22:42] <iive> JEEB: Gunslinger Girl is highly recommended.
[22:42] <ubitux> JEEB: requiem for the phantom
[22:42] <JEEB> need to get more of it when I get to a Book-Off the next time
[22:42] <JEEB> ubitux, same
[22:43] <ubitux> speaking of phantom, Kurau is surprisingly nice
[22:44] <JEEB> ah, that sounds like a nice one by Bones
[22:46] <ubitux> and the last best anime i saw was kemonozume
[22:46] <ubitux> anyway, if you have some kind of recommendations, i'm all ear
[22:47] <ubitux> i like a lot of weird/peaceful/violent/mindbreaking stuff so...
[22:47] <iive> ubitux: i guess you've watched madoka :)
[22:48] <JEEB> I have this thing I hate in memory, I usually forget about most things I've watched good or not if someone asks me for recommendations
[22:48] <ubitux> iive: yes
[22:49] <ubitux> JEEB: don't you have a list stored somewhere?
[22:49] <Compn> JEEB : i'd pass on rahxephon. one of those series with little answers and lots of booooooooring
[22:49] <JEEB> Compn, I already watched most of it and deemed it relatively enjoyable while it did have its shortcomings
[22:49] <ubitux> iive: and pretty good surprisingly :p
[22:49] <Compn> :P
[22:49] <JEEB> ubitux, I try to keep a list on anidb of what I've seen, but I often forget to add stuff there
[22:52] <ubitux> so i'm condemned to write code tonight? :(
[22:52] <JEEB> I liked both baccano and durarara, I liked haibane renmei, higashi no eden, bungaku shoujo (OVAs and the movie) looking at my current list
[22:53] <JEEB> and yes, this shows how freaking short my list is :<
[22:54] <JEEB> then I liked code geass that I watched after the summer
[22:54] <ubitux> saw everything except bungaku shoujo
[22:55] <ubitux> code geass? please... :/
[22:55] <JEEB> dunno, I enjoyed it for what it was
[22:55] <JEEB> bungaku shoujo is kind of an odd one
[22:58] <JEEB> anyways, I wish I kept better books of what I've watched
[22:58] <JEEB> most of it is crap, but there are some better ones
[23:00] <iive> code geass was good, they just made it twice, so the repetition was boring.
[23:01] <JEEB> also I'm one of those weirdos that began their mecha watching with Macross F and Gundam 00
[23:03] <iive> ever seen eva ?
[23:04] <JEEB> ok, that came before -- but I really didn't watch it as mecha back then
[23:04] <JEEB> I think one of the renewed DVDs for Eva was the first thing I ever imported from the US
[23:04] <JEEB> that was before I habla'd Japanese
[23:13] <saste> ubitux, yeah, ebur128 is cool
[23:13] <saste> i loove retro-style CGA fonts :)
[23:13] <ubitux> :D
[23:57] <saste> michaelni: ping on duration in AVFrame
[23:58] <Compn> saste : do you have a filter wishlist? can you put an autocropping filter on there ?
[23:58] <Compn> automatic cropdetect  :)
[23:58] <saste> i already asked you some weeks ago, you said it was somehow duplicating repeat_pict and told me to check that the patch was not breaking B-frames in mpeg or something like that
[23:58] <saste> i don't know how to test that
[23:58] <saste> Compn, trac
[23:58] <Compn> k
[23:59] <saste> also I don't know what do you mean by automatic cropdetect
[23:59] <saste> we have already cropdetect
[00:00] --- Tue Oct  2 2012