[Ffmpeg-devel-irc] ffmpeg-devel.log.20130109

[burek]

[00:24] <ubitux> michaelni: do you mind looking at #2045? i don't mind looking at fixing this but please tell me the purpose of your commit
[00:26] <michaelni> that commit fixed the timebase and the duration of something 
[00:26] <michaelni> i dont remember exactly what thet something was
[00:27] <michaelni> looking at ffprobe output of the generated files probably should show what it fixed
[00:28] <michaelni> that is diff with and without the problematic commit
[00:32] <ubitux> the generated files from fate?
[00:33] <michaelni> ubitux, actually fate fails an assert if i revert that commit: it fails here: libavformat/aviobuf.c:152
[00:33] <ubitux> ok
[00:33] <michaelni> mov_write_tmcd_tag (track=0x15bd960, pb=0x15c2a20) at libavformat/movenc.c:1160
[00:34] <michaelni> above calls that failing code
[00:34] <ubitux> btw, aren't the conditions in ea6da80c inverted?
[00:34] <ubitux> michaelni: ok i'm going to dig this after this one ^
[00:35] <ubitux> i'm looking at #2123 now
[00:35] <ubitux> and it's doing a 640%2==0 || 360%2==0
[00:36] <ubitux> should it be 640%2 || 360%2?
[00:36] <ubitux> (’ failure)
[00:39] <michaelni> ubitux, yes they are inverted :/
[00:39] <ubitux> this makes me realize we don't have a dirac fate test.
[00:43] <ubitux> michaelni: i can't really fix that myself (since i don't have access to the j00ru report), but if you fix it and upload the sample from the trac on the fate samples i'd gladly add a fate test
[00:45] <michaelni> ok, ill look into it
[01:10] <Compn> why vlc have 3 diff fourcc lists
[01:17] <cone-931> ffmpeg.git 03Michael Niedermayer 07master:57bdd67646cf: dirac: fix inverted check
[01:19] <ubitux> thx!
[01:19] <michaelni> ubitux, which dirac file should i put on fatesamples ? the file in the report is a bit big
[01:19] <michaelni> theres also http://diracvideo.org/download/test-streams/raw/vts/
[01:19] <michaelni> which has smaller files
[01:19] <ubitux> no idea, any would do
[01:19] <michaelni> ok
[01:20] <ubitux> you can also just take 1 or 2 seconds from the sample in the trac
[01:22] <michaelni> i uploaded a random one
[01:22] <ubitux> thx, will send a patch soon
[04:11] <llogan> i'm going to make a bug report to ubuntu to reupstream/rename/drop the transtional "ffmpeg" package
[04:11] <llogan> any suggestions? http://fakeoutdoorsman.com/ffmpeg/ubuntubug.txt
[04:23] <llogan> maybe reporting to debian would make more sense
[04:26] <Compn> llogan : isnt there already a bug on it ?
[04:26] <Compn> at their bugzilla i mean
[04:27] <Compn> oh you linked it already :)
[04:28] <llogan> that bug was about the message when users attempt to use the ffmpeg binary
[04:28] <llogan> this is aobut the package name itself
[04:29] <llogan> although i'm not happy about how the other bug ended up
[04:30] <Compn> you saw the response i got when i mailed debian-devel 
[04:30] <Compn> about mplayer2 / mplayer naming ?
[04:32] <llogan> no
[04:33] <Compn> http://lists.debian.org/debian-devel/2011/08/msg00418.html
[04:35] <Compn> llogan : imo, the only thing debian will listen to is a C&D from fabrice invoking ffmpeg trademark
[04:35] <Compn> unless you convince siretart or kierank (now i forgot which one) to rename the package
[04:36] <Compn> maybe they'll rename it to weaselmpeg :P
[04:36] Action: Compn lols
[04:37] <llogan> i don't think kieran has anything to do with that
[04:37] <Compn> ok i got them mixed up, my bad
[04:37] <Compn> couldnt remember who did the packages
[04:38] <llogan> the link in your link...the 2 appears to be the epoch number. unless i'm confused.
[04:39] <Compn> yeah, well try explaining that to users
[04:42] <Compn> confusion, the name of the game
[04:43] <Compn> llogan : you going to mail fabrice ?
[04:43] <llogan> i don't think i'll have to do anything like that.
[04:43] <Compn> ehe good luck then
[04:44] <Compn> your message looks ok
[04:44] <llogan> if i ask like a normal person and not make it a flame/complain fest then maybe something will come out of it.
[04:44] <Compn> i'm not sure if adding details of differences of the projects will help or just clutter the message
[04:45] <Compn> you werent around back in the day... debian did similar bullshit with mplayer :)
[04:46] <Compn> mplayer finally got all distros to stop distributing mplayer 
[04:46] <Compn> long long ago
[04:46] <Compn> so much drama
[04:46] <Compn> sorry for the noise llogan
[04:50] <llogan> where has Daemon404 gone?
[04:51] <Compn> llogan : i think we annoyed him too much with libav trolling
[04:51] <Compn> you can /q him
[04:52] <Compn> funman : highgod is the dxva patch author if you need to talk to him about his patch :)
[04:52] <Compn> i think
[04:56] <Compn> llogan : the iceweasel story is similar to ffmpeg/libav problem packaging name. did you read the history of it? also i think thats what iceweasel thing was same as ffmpeg/libav.
[04:56] <Compn> erg stupid paste
[04:57] <Compn> http://en.wikipedia.org/wiki/Mozilla_Corporation_software_rebranded_by_the_Debian_project
[04:57] <Compn> since debian freezes packages for years ... dont think they will rename voluntarily
[04:58] <Compn> i'm not against trying, of course :)
[05:02] <llogan> i didnt even mention trademark, but IANAL, and i don't know anything about that sort of thing
[05:06] <Compn> just pointing out the history of debian packages and naming schemes :)
[05:08] <llogan> i suppose more than anything a blurb about the legalese of the "ffmpeg" word would be useful for me.
[05:10] <michaelni> llogan, btw it would be great if we could get real debian packages in debian 
[05:10] <michaelni> ffmpeg packages that is ;)
[05:11] <llogan> michaelni: yes. that would be ideal.
[05:12] <llogan> we need a maintainer
[05:12] <michaelni> yes
[05:12] <llogan> ...but that is my first suggestion listed
[05:12] <michaelni> ok, i admit i didnt read ...
[05:12] <llogan> i wonder if jon severinsson would be interested.
[05:13] <michaelni> you should ask him
[05:14] <llogan> added to "to do" list.
[05:24] <llogan> is the ffmpeg name trademarked?
[05:26] <Compn> yes, fabrice owns the trademark
[05:26] <Compn> hes the creator of the project
[05:26] <cone-815> ffmpeg.git 03Michael Niedermayer 07master:f3c9d8d41bca: swr: use SIMD for dithering
[05:27] <Compn> oh thats why i get 403 on vlc forums. it thinks i'm a spammer or doesnt like my browser. great.
[05:29] <Compn> http://roundup.libav.org/ seems down
[05:39] <llogan> Compn: same
[06:08] <llogan> michaelni: message sent. i'll let y'all know what he says.
[06:37] <siretart> Compn: FYI: the package 'ffmpeg' is already gone for good in debian/ubuntu: http://anonscm.debian.org/gitweb/?p=pkg-multimedia/libav.git;a=blobdiff;f=debian/control;h=ba2760abac7efa6e4e8d3d4619a07e01b0bd10c3;hp=3ab1b81d2ebce43cb520300a3029de185b8f137d;hb=a7398cdc19fb2ea2a7716fb15b4b02f9ecf7c3ce;hpb=8cf50e974ef00b5cd050b139c7d9c00c09130f1e
[06:37] <siretart> Compn: cf. http://thread.gmane.org/gmane.linux.ubuntu.motu/7295 why it isn't in raring yet
[08:12] <ubitux> siretart: now it would be nice to rename the libraries
[08:12] <ubitux> ;)
[08:13] <ubitux> that's not the packager responsability though
[08:33] <siretart> ubitux: indeed
[08:33] <ubitux> i hope this "indeed" was to answer the nice renamine
[08:33] <ubitux> renaming*
[09:49] <cone-242> ffmpeg.git 03Michael Niedermayer 07release/1.1:ddb0317154ce: dirac: fix inverted check
[11:13] <durandal_1707> michaelni: you have typo on one of your signatures stuff you put at end of mail
[11:33] <cone-149> ffmpeg.git 03Kanglin 07master:ba8cb33273ae: hlsenc: Make the start_number option set the right variable
[11:33] <cone-149> ffmpeg.git 03Martin Storsjö 07master:92e354b65561: rtpdec_vp8: Set the timestamp when returning a deferred packet
[11:33] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:aed58f6aae0c: Merge commit '92e354b655613b88c3c202a7e19e7037daed37eb'
[11:34] <michaelni> durandal_1707, which one ? where ?
[11:34] Action: michaelni is surprised its just one
[11:36] <durandal_1707> There will always be a question for which you  ... awnser
[11:39] <cone-149> ffmpeg.git 03Martin Storsjö 07master:09ed8098ff66: rtpdec_vp8: Make sure the previous packet is returned
[11:39] <cone-149> ffmpeg.git 03Martin Storsjö 07master:9c80ed836a51: rtpdec_vp8: Avoid a warning about a possibly unused variable
[11:39] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:7b822b1de2ea: Merge commit '9c80ed836a511293f4cc3a858060969d32f2b1ce'
[11:41] <michaelni> durandal_1707, thx fixed
[11:47] <cone-149> ffmpeg.git 03Martin Storsjö 07master:42805eda554a: rtpdec: Store the dynamic payload handler in the rtpdec context
[11:47] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:8c3ae9ee66a2: Merge commit '42805eda554a7fc44341282771531e7837ac72b7'
[11:47] <durandal_1707> huh why iff demuxer depends on iff.o from lavc?
[11:49] <durandal_1707> michaelni: ff_find_pix_fmt is used by lavf
[11:49] <durandal_1707> it should be renamed to avpriv_... or av_ ...
[11:50] <durandal_1707> or moved to lavu
[11:56] <mateo`> i'd like to port vf_spp, however the page describing the algorithm is down: http://wayback.archive.org/web/*/http://citeseer.nj.nec.com/nosratinia99embedded.html. Anyone has a local copy or documentation ?
[11:58] <cone-149> ffmpeg.git 03Martin Storsjö 07master:86d9181cf41e: rtpdec: Support sending RTCP feedback packets
[11:58] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:34c1c08c6653: Merge commit '86d9181cf41edc3382bf2481f95a2fb321058689'
[12:01] <michaelni> mateo`, copy from /dev/brain: to deblock an image with 8x8 blocks you take the image 64 times each time shifted so a different pixel is at 0x0 and compress all these 64 by jpeg at similar parameters to the original
[12:01] <michaelni> then average all 64 and you have no more blocking artifacts
[12:02] <michaelni> this is slow so you can do it with a subset of 64 and you can skip full jpeg and just do (i)dct & (de)quant
[12:02] <mateo`> michaelni: thx :)
[12:03] <michaelni> np
[12:03] <Compn> mateo` : where did you find the link ?
[12:03] <Compn> maybe i can find a copy
[12:03] <mateo`> Compn: vf_spp.c (minus the webarchive thing)
[12:03] <Compn> oh :)
[12:04] <Compn> bonus points for anyone that checks all source for urls and archives them.
[12:07] <Compn> mateo` : is this it ? http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.50.5859
[12:09] <mateo`> Compn: i suppose michaelni can tell if it is this document
[12:09] <Compn> oh yeah thats it
[12:09] <Compn>    22  * This implementation is based on an algorithm described in
[12:09] <Compn>    23  * "Aria Nosratinia Embedded Post-Processing for
[12:09] <Compn>    24  * Enhancement of Compressed Images (1999)"
[12:09] <Compn>    25  * (http://citeseer.nj.nec.com/nosratinia99embedded.html)
[12:09] <Compn>    26  */
[12:09] <Compn> Embedded Post-Processing for Enhancement of Compressed Images (1999)
[12:10] <funman> Compn: hm he's gone
[12:10] <funman> (highgod)
[12:10] <Compn> by Aria Nosratinia
[12:11] <mateo`> Compn: awesome :)
[12:12] <Compn> funman : well you can always email :P
[12:14] <Compn> siretart : if you say so. http://packages.ubuntu.com/search?keywords=ffmpeg or http://packages.debian.org/sid/ffmpeg
[12:17] <Compn> siretart : thats a crazy amount of work. updating all packages depending on ffmpeg :D 
[12:17] <Compn> good luck
[12:19] <siretart> Compn: check experimental
[12:21] <cone-149> ffmpeg.git 03Martin Storsjö 07master:6f7244112009: rtpdec_vp8: Request a keyframe if RTP packets are lost
[12:21] <Compn> so for anyone not using experimental repo ?
[12:21] <cone-149> ffmpeg.git 03Justin Ruggles 07master:142c39097c30: doc/APIchanges: fill in missing dates and hashes.
[12:21] <cone-149> ffmpeg.git 03Justin Ruggles 07master:1fb8f6a44f06: x86: lavr: add SSE2 quantize() for dithering
[12:21] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:75afbe2ab4a1: Merge commit '1fb8f6a44f06e48386450fe0363aefc02583d24a'
[12:26] <Compn> siretart: so for anyone not using experimental repo ? they see the ffmpeg transistion package ?
[12:26] <siretart> Compn: yes
[12:28] <Compn> and the transistion package is named ffmpeg, but contains libav
[12:28] <Compn> ?
[12:28] <durandal_1707> because obviously it is targeted for idiots
[12:29] <Compn> siretart:  the transistion package is named ffmpeg, but contains libav?
[12:30] <Compn> durandal_1707 : say what? i dont understand 
[12:30] <siretart> Compn: are you trying to troll or are you too lazy to check yourself?
[12:31] <siretart> Compn: fyi: http://packages.debian.org/sid/i386/ffmpeg/filelist
[12:31] <Compn> its pretty confusing to me, not trying to troll
[12:31] <Compn> yeah that contains the tool ffmpeg , the source link points to libav
[12:32] <JEEB> libav provided a binary by the name of ffmpeg until after the 0.8 release methinks
[12:32] <Compn> Links for ffmpeg
[12:32] <Compn> Download Source Package libav:
[12:32] <Compn> http://packages.debian.org/sid/ffmpeg
[12:33] <JEEB> which was the ffmpeg it was forked off of, and then when elenril rewrote the settings stuff etc. it got renamed to avconv it seems
[12:33] <Compn> yeah i know the history of ffmpeg tool in libav
[12:33] <Compn> thats not my question
[12:33] <JEEB> k
[12:34] <Compn> lemme rephrase
[12:34] <Compn> siretart:  the transistion package is named ffmpeg, but contains libav's ffmpeg tool?
[12:36] <Compn> anyways, not my fight
[12:36] <Compn> siretart should be telling llogan :)
[12:37] <Compn> if i was fighting i'd be mailing fabrice :P
[12:38] <Compn> i wonder how many other packages debian and *buntu substitute some fork for the real project without notifying users
[12:38] <cone-149> ffmpeg.git 03Justin Ruggles 07master:a6a3164b1399: x86: lavr: add SSE2/AVX dither_int_to_float()
[12:39] <cone-149> ffmpeg.git 03Justin Ruggles 07master:59220d559b50: oggenc: add a page_duration option and deprecate the pagesize option
[12:39] <cone-149> ffmpeg.git 03Alexandra Khirnova 07master:d744801f1a7c: xan: Convert to bytestream2
[12:39] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:4765f635387c: Merge remote-tracking branch 'qatar/master'
[12:42] <durandal_1707> does get_buffer allocate little more than requested?
[12:46] <durandal_1707> the idea is to use bswap_buf directly from input to output
[12:46] <durandal_1707> and if input/output is not padded i would get overread/overwrites
[12:49] <michaelni> if its video buffers see avcodec_align_dimensions2()
[12:49] <durandal_1707> no it is audio
[12:49] <durandal_1707> and i use frame.data pkt->data directly
[12:50] <durandal_1707> i dont use temp buffer and do swaping on that
[12:54] <michaelni> you can set a larger nb_samples before allocation to ensure the bufer is large enough
[12:54] <michaelni> then decrease nb_samples to its true value
[12:57] <durandal_1707> that is for output only, and padding requirement for bswap_buf is nowhere documented
[12:58] <durandal_1707> it only mentions that it may write 1 extra item
[12:59] <michaelni> then thats the max that it (according to API docs) is allowed to over access
[12:59] <durandal_1707> and nowhere mentions input requirements, are they same ?
[13:00] <michaelni> i would have to read the implementation
[13:00] <durandal_1707> well i wanted to do bswap for pcm le to pcm be case
[13:01] <durandal_1707> but looks that only correct way is to add LE/BE SAMPLE_FORMAT variants like it is done in libswscale
[13:45] <pross-au> to boost membership of the google+ page, i suggest adding a (temporary) hyperlink in the #ffmpeg-.* channel topic
[13:45] <av500> g+?
[13:45] <av500> ff going social?
[13:46] <pross-au> socialist. yes.
[13:48] <av500> all equal and some equaler?
[13:48] <av500> +1
[13:48] <Compn> google has changed, cant support it anymore
[13:49] <durandal_1707> Compn: what?
[13:49] <av500> they are evil now
[13:49] <Compn> yep, they turned evil
[13:50] <durandal_1707> how?
[13:50] <pross-au> lets write our own
[13:50] <Compn> they made their bed in china and are sticking to it
[13:50] <pross-au> (i mean we did our our md5, aes, build system, test framework)
[13:50] <pross-au> FF+
[13:50] <Compn> pross-au : if you want me to put a link in topic i can
[13:51] <Compn> or did you mean #ffmpeg ?
[13:52] <Compn> dont think i have ops there
[13:52] <Compn> oh you have ops here too
[13:52] Action: Compn blind
[13:53] <pross-au> i am more of an ideas person, than implementor. 
[13:53] <pross-au> durandal_1707: which google link is best?
[13:55] <durandal_1707> it chicken-egg problem, to get short url to g+ one need certain number of members
[13:55] <durandal_1707> currently only link is long/ugly link
[13:57] <durandal_1707> i could request for shorter url, will see if get accepted
[13:59] <durandal_1707> firefox is really crap, it fails to dns 
[14:01] <durandal_1707> it ask for my phone number and organization address
[14:01] <durandal_1707> what is address for FFmpeg?
[14:02] <pross-au> Postal?
[14:02] <Compn> dont really have one
[14:02] <Compn> you can only g+ for physical addys ? lol
[14:03] <Compn> ironic considering google is an internet company
[14:03] <Compn> trying to design a social network via committee from people who are not internet savvy
[14:03] <pross-au> I now agree, its evil.
[14:04] <pross-au> such stupidity would *never* happen on facebook
[14:04] <durandal_1707> i could add address and numbers from whois for ffmpeg.org
[14:05] <durandal_1707> i need to enter my title: it is Senior Chief of Universe
[14:05] <Compn> facebook is evil for other reasons :D
[14:05] <durandal_1707> j-b: did you filled this for vlc?
[14:06] <Compn> but still similar reasons to google
[14:06] <j-b> durandal_1707: no
[14:07] <durandal_1707> lol, i submited n/a for such entries
[14:07] <pross-au> Popular destination, that N/A.
[14:10] <Compn> everyone loves 'North America' :P
[14:16] <cone-149> ffmpeg.git 03Peter Ross 07master:5b69c07d12b1: frmdec: tag 5 is AV_PIX_FMT_BGRA with reversed alpha (0=solid/opaque ..  255=transparent)
[18:25] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:82742294b7a8: swr: Implement Noise shaping dither
[19:46] <ubitux> saste: do you have an output example for the plotframes script?
[19:46] <ubitux> (yeah i'm lazy)
[19:49] <saste> ubitux, are you asking for a screenshot?
[19:49] <saste> no i'm lazy
[19:49] <ubitux> doesn't it output a png?
[19:49] <ubitux> ok :(
[19:50] <saste> ubitux, no it creates an interactive session, don't know how to save a graph from gnuplot
[19:51] <wm4> try a script with set output "plot.png"
[19:51] <wm4> the rest is like in the interactive session
[19:51] <ubitux> set term png size ...
[19:52] <ubitux> something like this
[19:54] <wm4> hm right
[19:56] <ubitux> saste: Failed to set value 'frame' for option 'show_entries'
[19:56] <ubitux> :(
[19:56] <ubitux> oh wait
[19:56] <ubitux> it's running ffprobe from my system]
[20:05] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:ba1314c2efc4: swr: update copyright years of a few files iam working on
[20:05] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:3ab1970612b5: swr: move dither related fields to their own context
[20:05] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:8514d5bcf780: swr: remove unused variable
[20:08] <ubitux> michaelni: missing docs ;)
[20:09] <ubitux> michaelni: undocumented features are meaningless, you should advertise them by at least mentioning them in the doc :p
[20:12] <michaelni> ok will fix
[20:15] Action: beastd is thinking loud
[20:15] <ubitux> nevcairiel: "both sides assume the other side just writes bad code" // if that was really the case we wouldn't have merge for 2 years
[20:16] <nevcairiel> that mail from last night reads differently, there is plenty comments on how buggy the code will be
[20:16] <ubitux> i believe there is some recognition of libav on ffmpeg side
[20:16] <ubitux> but of course, we are critic about their work :)
[20:17] <ubitux> but we can, since we actually are directly "victim" of it
[20:17] <ubitux> ;)
[20:17] <ubitux> we (at least i) appreciate a lot of their contributions
[20:19] <nevcairiel> so do i, but sometimes i get the feeling a lot of people would rather stop merging and diverge the projects
[20:19] <nevcairiel> while ffmpeg adds a bunch of new features, api improvements usually come from libav, which is also an important part
[20:19] <ubitux> i can think of paul
[20:19] <ubitux> but no one else
[20:20] <ubitux> maybe michaelni to some extent of course, but that's understandable given the work he did daily for years
[20:20] <saste> nevcairiel, api improvements come from libav -> that's why we tend to be very careful not to introduce more merge work
[20:20] <nevcairiel> i know
[20:20] <saste> nevcairiel, also, how much better would be if people who designed the code could also comment on it
[20:22] <saste> get to go now, bye
[20:22] <nevcairiel> communication, as always with everything, is the biggest problem :p
[20:23] <ubitux> i don't think there is a problem of communication
[20:23] <nevcairiel> a lack of?
[20:23] <nevcairiel> :)
[20:23] <ubitux> yeah there is some ears covered
[20:23] <ubitux> but that's all
[20:24] <ubitux> saying that 2 filters out of 50 are useful, and they are badly written is an obvious proof :p
[20:24] <ubitux> well, enough grambling for me
[20:25] <nevcairiel> yeah that was typical
[20:26] <nevcairiel> many projects and especially forks sadly suffer from NIH
[20:27] <nevcairiel> i can't say i don't have the same problem, often i prefer implementing a feature myself then trying to port it from somewhere else
[20:28] <ubitux> you don't cary the same responsibility in your own project
[20:28] <nevcairiel> of course not, but that doesnt change how people feel
[20:28] <ubitux> imo this behaviour is irresponsible at this level
[20:28] <ubitux> when you know how much users are involved, it's a nonsense to behave like that
[20:29] <ubitux> or just childish, but i'm not the best person to say that
[20:31] <nevcairiel> oh well back to coding it is .. i need to hack up my audio pipeline to do native planar finally...
[20:31] <ubitux> :)
[23:13] <cone-149> ffmpeg.git 03Carl Eugen Hoyos 07master:c20ce9ad699a: Add -movflags rtphint documentation.
[23:29] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:dc60ecdbeb11: doc: add noise shaping dithers
[23:29] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:3ef06f3415a4: swr: split swri_dither_init() out
[23:29] <cone-149> ffmpeg.git 03Michael Niedermayer 07master:6908ded20e80: swr: fix the noise sample type for noise shaping
[00:00] --- Thu Jan 10 2013