[Ffmpeg-devel-irc] ffmpeg-devel.log.20130212

[burek]

[00:30] <saste_> wow av_url_split() sucks
[00:35] <ubitux> unlimited free wifi at airport \o/
[00:35] <ubitux> i don't want to leave :(
[00:42] <cone-615> ffmpeg.git 03Michael Niedermayer 07master:3b1f037df56b: ffmpeg: fix streamcopy of lsf mp3.
[00:43] <michaelni> BBB-work, stuttering should be fixed
[00:43] <saste> ubitux, too bad for you
[00:43] <saste> better for us ;-)
[00:44] <ubitux> maybe i forgot how to write code
[00:44] <saste> ubitux, now you write code in kanji
[00:44] <ubitux> :)
[00:45] <wm4> ubitux: btw., what's the status of the subtitle stuff?
[00:45] <ubitux> untouched for one month?
[00:46] <wm4> maybe I'll feel the need to kill of mplayer's subreader.c on a whim or so in the near future
[00:46] <ubitux> wm4: i'll be back to it as soon as i'm home
[00:46] <wm4> there was something about codepages still
[00:47] <ubitux> but i need to catch up with one mo nth of semi hibernation
[00:47] <ubitux> wm4: yeah there is a pending patch for this
[00:47] <ubitux> codepage, char encoding etc
[00:47] <wm4> I found libass doesn't even read utf-16 natively :(
[00:47] <ubitux> it seems to be in need of some more thinking for the utf-16 shit
[00:47] <ubitux> wm4: yeah well of course
[00:48] <ubitux> SSA/ASS is utf-8 only afaik
[00:48] <wm4> well I've seen utf-16 files in the wild
[00:48] <wm4> http://code.google.com/p/libass/issues/detail?id=89
[00:48] <ubitux> lol
[00:48] <wm4> chinese stuff
[00:48] <wm4> fortunately not multibyte something :)
[00:57] <saste_> i wonder if there is some way to avoid all the MP3 application-level hacks
[01:02] <saste> michaelni: i summon you on "v4l2: bug #1570 and possible solution"
[01:03] <saste> after some thinking i can't find a better solution of the one proposed
[01:03] <saste> even if it is somehow hackish
[01:12] <J_Darnley> What the...  Who thought 99162f8d46db7ec02facae035c4ff573d2d8d612 was a good idea?
[01:22] <cone-615> ffmpeg.git 03Michael Niedermayer 07master:365270aec5c2: sanm: add forgotten check for decoded_size in old_codec37()
[01:24] <saste_> J_Darnley, I don't like it very much, but what's the problem with it?
[01:28] <J_Darnley> No real problem.  I just think it is ugly.
[01:28] <J_Darnley> Plus it missed one line.
[01:29] <J_Darnley> All that to silence one warning?  A warning that I thought an existing void pointer cast was there to hide.
[01:42] <BBB-work> michaelni, are timestamps assigned to all non-b-frames now? can you check that?
[01:44] <BBB-work> michaelni, we had issues in chrome where non-b-frames had AV_NOPTS_VALUE assigned to them for that video
[02:04] <BBB-work> michaelni, and new patch sent that should apply cleanly on head for that other patch
[02:14] <cone-615> ffmpeg.git 03Michael Niedermayer 07master:f0d9ccac8937: sanm: reset frame sizes on buffer destruction
[02:14] <cone-615> ffmpeg.git 03Michael Niedermayer 07master:4ee88db840b3: sanm: init sizes with the full frame size
[02:14] <cone-615> ffmpeg.git 03Michael Niedermayer 07master:5260edee7e5b: sanm: Use the correct height variable in the decoded_size checks
[02:23] <cone-615> ffmpeg.git 03Ronald S. Bultje 07master:7ff1a4b10f2d: Add add_pixels4/8() to h264dsp, and remove add_pixels4 from dsputil.
[02:34] <michaelni> BBB-work, you need "-fflags genpts" or equivalent for pts from avi with b frames
[02:34] <michaelni> this will add latency/delay though
[02:47] <BBB-work> michaelni, in decoder?
[02:49] <michaelni> latency in demuxer to delay packets till the next non B frame whichs dts is the previous frames pts
[02:57] <creep> »!
[03:41] <Plorkyeran> @ubitux> SSA/ASS is utf-8 only afaik <-- according to the spec it's ANSI, except for when you use \fe to override the encoding for a specific line
[03:41] <Plorkyeran> everyone ignores that these days because it's retarded
[04:32] <mfg> Hi, I have an mp4 file that wont play (its by Neulion) and I think they screw up the header on purpose to protected their content. Is there someone that can have a look at it?
[04:39] <llogan> mfg: wrong channel. try #ffmpeg.
[12:26] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=faac85005cadb5065d218b0563eeb3272a498073
[12:26] <ubitux> http://git.libav.org/?p=libav.git;a=commitdiff;h=fd1abf42693308ff99b9473c2387d5142f8b3f0e
[12:26] <ubitux> rofl
[12:27] <durandal_1707> you mean E1 thing?
[12:27] <ubitux> yeah
[12:27] <ubitux> E1 is obviously better
[12:28] <av500> it saves storage space :)
[12:28] Action: durandal_1707 blames nevcairiel
[12:29] <nevcairiel> clearly i should've suggested E1 instead of EXPAND
[12:48] <cone-102> ffmpeg.git 03Martin Storsjö 07master:ab8f1a698990: arm: Fall back to runtime cpu feature detection via /proc/cpuinfo
[12:48] <cone-102> ffmpeg.git 03Derek Buitenhuis 07master:375ef6528c9d: libfdk-aacenc: Actually check for upper bounds of cutoff
[12:48] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:3ccda2b02ba6: Merge commit '375ef6528c9dd2db7f9881e232cb0ec3aa16970d'
[13:07] <iive> nevcairiel: actually E7 is better, E is short of Exp (like in the float format), 7 is the same key as & that is user for AND, so e7 reads as EXPAND.
[13:08] <av500> iive: on your keyboard
[13:09] <iive> av500: on most keyboards :P
[13:10] <av500> iive: so maybe git post-checkout hook that modifies it for the locale :)
[13:11] <cone-102> ffmpeg.git 03Diego Biurrun 07master:6fb296e97edc: avcodec/rectangle: Remove nonsense assert
[13:11] <cone-102> ffmpeg.git 03Diego Biurrun 07master:62a43d30cc3b: crc: Move static data table from header to .c file
[13:11] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:e2b703f53612: Merge commit '62a43d30cc3b9b2460f2c926beb522df32493625'
[13:12] <durandal_1707> wasn't there patch on ml that fix VC1 interlaced?
[13:15] <burek> just to mention this, we could integrate something like this into our wiki: https://github.com/mpenkov/ffmpeg-tutorial
[13:15] <burek> A set of tutorials that demonstrates how to write a video player based on FFmpeg
[13:16] <burek> :D
[13:17] <burek> did you manage to get the list of cmds? :)
[13:21] <durandal_1707> burek: i have private correspodence with ffloger
[13:21] <durandal_1707> now if only it could write code ....
[13:22] <burek> :)
[13:25] <cone-102> ffmpeg.git 03Martin Storsjö 07master:d192ac3e03d1: swscale: Disallow conversion to GBRP16
[13:25] <cone-102> ffmpeg.git 03Martin Storsjö 07master:4c51fe48ba6d: h264: Copy h264chroma dsp context to slice thread copies
[13:25] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:fe54f6cb453c: Merge commit '4c51fe48ba6dde059360b7451db7f6bbf2f11db3'
[13:34] <cone-102> ffmpeg.git 03Daniel Kang 07master:b3f2a3fe3fe4: x86: mpeg4qpel: Make movsxifnidn do the right thing
[13:34] <cone-102> ffmpeg.git 03Diego Biurrun 07master:304b806cb524: build: Make library minor version visible in the Makefile
[13:34] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:7491356111e7: Merge commit '304b806cb524fb040f8e09a241040f1af2cb820b'
[14:05] <cone-102> ffmpeg.git 03Diego Biurrun 07master:69dca3a4f45e: openbsd: Add minor number to shared library install name
[14:05] <cone-102> ffmpeg.git 03Diego Biurrun 07master:56dea32f09e7: doc: developer: Fix wording in "naming conventions" section
[14:05] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:08db519e1275: Merge commit '56dea32f09e71460add14a1b6d0fbe6eb16f02a4'
[14:21] <cone-102> ffmpeg.git 03Diego Biurrun 07master:3658d2dfd48a: doc: developer: Add a note about reserved system name space
[14:21] <cone-102> ffmpeg.git 03Luca Barbato 07master:fd1abf426933: lavu: avoid clashing definition of E
[14:21] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:a63dc846b502: Merge remote-tracking branch 'qatar/master'
[14:42] <durandal_1707> why transpose filter does not work for h != v chroma subsampling?
[15:17] <divVerent> 14:42:28   @durandal_1707 | why transpose filter does not work for h != v chroma subsampling?
[15:17] <divVerent> because then transposing would be lossy
[15:17] <divVerent> and that's simply not implemented (one can always convert to 4:4:4 first though)
[15:34] <durandal_1707> ok -> no reply -> applying patches at will
[15:39] <durandal_1707> divVerent: ahh, because of spam i did not see you reply, sorry
[15:43] <cone-102> ffmpeg.git 03James Almer 07master:0957041221fa: lavc/tta: Add support for encrypted streams
[15:43] <durandal_1707> yes, that makes sense
[16:02] <cone-102> ffmpeg.git 03Daniel Verkamp 07master:1ecce266fa5b: WAV muxer: add RF64 support
[16:21] <ubitux> michaelni: "visible"
[16:22] <ubitux> (commit message)
[16:22] <av500> better than invisible commit message
[16:23] <durandal_1707> ubitux: what/where?
[16:24] <ubitux> "[FFmpeg-devel] [PATCH] developer.texi: replace "externally vissible" by public"
[16:25] <ubitux> just a typo
[16:25] <ubitux> 9h in the transit is a nightmare
[16:25] <ubitux> in about ~15 hours i'm back for new adventures
[16:26] <av500> going round the world?
[16:26] <durandal_1707> you need 24h of sleep to sync
[16:26] <ubitux> no, i'll be in sync directly
[16:28] <ubitux> i just need to stay away for about 8 extra hours, which is easy with all the movie stuff available
[16:29] <ubitux> i wonder what they use for the movies btw
[16:29] Action: durandal_1707 removed base64 from its system
[16:29] <ubitux> it's interlaced as hell
[16:29] <ubitux> also, they display multiple subtitles at once
[16:29] <ubitux> which is kinda interesting
[16:29] <ubitux> the system is a linux but that's common knowledge
[16:29] <ubitux> i wonder if someone has dumped a copy or two of these
[16:30] <ubitux> av500: no, just back from a month in a country at the other side of the planet
[16:31] <ubitux> and i'm waiting at the transit right now :P
[16:31] <av500> australia?
[16:31] <ubitux> transit is HK
[16:35] <ubitux> (fr->china->jpn and th other way around)
[16:36] <durandal_1707> hmm, why china?
[16:43] <durandal_1707> nut doesnt support qtrle by default
[17:00] <aballier> maybe i've missed something, but doesnt the AVFrame.channel change from int64 to int implies sizeof(AVFrame) decreases ?
[17:00] <aballier> and in this case, accessing AVFrame.pkt_size will break
[17:03] <aballier> emailed, better ask there directly
[17:04] <nevcairiel> these fields are not part of the public API, you are supposed to use the getter functions
[17:05] <aballier> ah ok, makes sense, i was missing that part
[18:30] <durandal_1707> saste: i may be stupid but what is way to pad silence at start of stream?
[18:30] <durandal_1707> sox have delay efffect which delays each channel in audio to some value
[18:49] <saste> durandal_1707, aresample=min_comp=0.001:min_hard_comp=0.1 is one way
[18:49] <saste> then there is a swr option which can be used to achieve the same effect
[18:49] <saste> and yes adding an example may be useful
[18:49] <saste> I don't know if apad can be used as well
[18:54] <durandal_1707> what that numbers means and what ammount of silence it adds?
[18:56] <saste> durandal_1707, wait, you want to *ADD* silence at the beginning of the stream?
[18:58] <durandal_1707> yes
[18:58] <saste> http://thread.gmane.org/gmane.comp.video.ffmpeg.user/43315/focus=43318
[18:59] <durandal_1707> that resets pts to some 0 value
[18:59] <saste> durandal_1707, plus, you can add an initial offset with asetpts=PTS+TIME_OFFSET/TB
[19:00] <saste> I didn't try the first_pts option yet
[19:00] <durandal_1707> well i would certainly need working command
[19:02] <saste> durandal_1707, e.g. if you want to add 100 seconds of silence, you do: asetpts=PTS+100/TB,aresample=min_comp=0.001:min_hard_comp=0.1
[19:02] <saste> I suspect first_pts may be a more robust solution
[19:03] <saste> apad could be extended to achieve the same effect
[19:03] <saste> also now the audio filters order is messed up
[19:04] <saste> what about factorizing documentation for all biquad filters?
[19:05] <saste> you could have section and a subsection for all variants
[19:05] <durandal_1707> saste: i'm more interested in writing new filters
[19:05] <saste> you could have *a main section* and a subsection for all variants
[19:05] <saste> i can do the work, just wanted to know your opinion
[19:05] <durandal_1707> also I really do not know doxygen
[19:05] <saste> texinfo i guess
[19:05] <saste> not very much to know
[19:09] <durandal_1707> saste: feel free to do whatever you feel is needed, my only opinion about documentation is that it should be present & correct and not cryptic
[19:11] <saste> durandal_1707, about mp and optims, yes they are disabled
[19:11] <saste> don't ask me why
[19:12] <durandal_1707> saste: i'm trying to write VV->V filter where both inputs are readonly and i create new output buffer, i get buffer queue overflow errors
[19:12] <durandal_1707> and nothing in output
[19:12] <durandal_1707> ffmpeg -f lavfi -i testsrc -f lavfi -i mptestsrc -filter_complex blend /tmp/out.png
[19:46] <Compn> http://torrentfreak.com/open-source-giant-vlc-mulls-bittorrent-support-130211/
[19:51] <durandal_1707> seed!!!
[19:57] <Compn> j-b talking to torrentfreak :D
[19:58] <durandal_1707> saste: or should i have another queue for filtered buffers?
[20:01] <durandal_1707> i really dunno what should I do and is that even possible
[20:07] <michaelni> it should work to dereference the input and create new output
[20:08] <durandal_1707> yes, but i'm clearly missing some code
[20:11] <durandal_1707> https://github.com/richardpl/ffmpeg/tree/blend
[20:28] <michaelni> I think you are missing a request_frame() implementation but i dont know if this is related to the problem
[20:36] <durandal_1707> michaelni: and i would request frame for each input?
[20:37] <durandal_1707> s/for/from
[20:38] <durandal_1707> i just need template and i did 99% of filter already :D
[20:47] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:1ac0fa50eff3: pngdec/filter: dont access out of array elements at the end
[20:57] <durandal_1707> yay
[23:52] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:6f9ae391deea: faxcompr: Dont read ref when the end has been reached in pass mode
[23:52] <cone-102> ffmpeg.git 03Michael Niedermayer 07master:e1219cdaf9fb: tiff: Check buffer allocation and pointer increment more carefully in shorts2str() and double2str()
[00:00] --- Wed Feb 13 2013