[Ffmpeg-devel-irc] ffmpeg-devel.log.20120928

[burek]

[03:02] <Daemon404> cbsrobot, cia is permadead
[03:02] <Daemon404> and github has its own irc bot, you only need to enable it in the project settings
[03:19] <michaelni> Daemon404, what do i need to enable exactly in github ?
[03:21] <Daemon404> its listed under service hooks
[03:21] <Daemon404> in project admin
[03:21] <Daemon404> as "IRC"
[03:38] Action: michaelni wonders why i486-linux-gnu-gcc-4.4 seems running a old seektest with new reference files
[03:45] <llogan> don't get your hopes/lols up, it's only a enhancement/wish and not an actual decoder.
[03:47] <Daemon404> from what i understand, aic isnt -that- hard to RE. just noone who can and wants to.
[03:48] <llogan> yeah, that's what i assumed. i had an AIC file from a client and made the ticket to procrastinate some meatspacings.
[03:50] <Daemon404> use mplayer's binary loader
[03:50] <Daemon404> i guess
[03:50] <ohsix> and they give you a comprehensible blob to do it, http://support.apple.com/downloads/Apple_Intermediate_Codec_1_0_1 lewl
[03:50] <llogan> eh..i just told them to give me almost any other format.
[03:50] <Daemon404> ohsix, better than "somewhere in iMovie"
[03:50] <ohsix> def
[03:51] <ohsix> why isn't there a tool like qemu-img for vm images, for .dmg images yet
[03:51] <Daemon404> you can
[03:51] <Daemon404> i mount them on linux all the time
[03:52] <ohsix> yea, but they aren't all images
[03:52] <ohsix> they're metafiles with a bunch of different things in them
[03:52] <Daemon404> ive never had a dmg i could not mount
[03:52] <Daemon404> :/
[03:55] <ohsix> do you just use mount -o loop?
[03:56] <Compn> ohsix : fusefs mounts stuff
[03:56] <Compn> dunno if dmg tho
[03:56] <ohsix> i'd probably just use hfsprogs instead of mounting :p
[03:56] <Daemon404> ohsix, if theyre comrpessed, there's a tool to uncomrpess it
[03:56] <Daemon404> then i just mount it as hfs+
[03:57] <Daemon404> -o loop yea
[03:57] <ohsix> ok, i need the name of the uncompress tool :p
[03:57] <Daemon404> http://vu1tur.eu.org/tools/
[03:57] <ohsix> ok
[03:58] <Compn> wonder what kostya is working on now
[03:58] <Compn> he is current RE master :P
[04:11] <ohsix> ah, dmg2img picked up ADC compression in 2009 ... it was probably that long ago :]
[04:36] <ohsix> it does look pretty straight forward to document AIC from that codec
[04:37] <ohsix> aside from fishing through vtables D:
[04:46] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/SUkE5A
[04:46] <KGB> [FFmpeg/master] mp3dec: ask for 8khz switch point mp3s - Michael Niedermayer
[04:53] <Daemon404> a tad lackluster
[04:54] <Daemon404> ohsix, vtables = kills RE
[04:54] <Daemon404> laziness
[04:54] <ohsix> shrug, yea
[04:55] <ohsix> i wanted to clone katafx (the original) some years ago, even with some effort no image of sanity was resolving itself
[04:56] <ohsix> all it is is some color band textures that rotate over eachother, but it had some magic
[05:17] <KGB> [FFmpeg] michaelni pushed 7 new commits to master: http://git.io/EIgYQg
[05:17] <KGB> [FFmpeg/master] avformat: Identify anonymous AVIO typedef structs. - Dale Curtis
[05:17] <KGB> [FFmpeg/master] Add Opus codec id and codec description. - Andrew D'Addesio
[05:17] <KGB> [FFmpeg/master] Add Opus support to the Ogg muxer. - Nathan Caldwell
[05:18] Action: michaelni takes note that the KGB can not count to 7
[05:19] <michaelni> Daemon404, that bot, how can it be made not to skip 4 out of 7 commits ?
[05:41] <Daemon404> michaelni, i have no idea
[05:44] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/FBogVA
[05:44] <KGB> [FFmpeg/master] configure: add missing dependancies for gpl tests - Michael Niedermayer
[07:37] <ubitux> yay 1.0.
[07:37] <ubitux> thanks michaelni :)
[07:37] <ubitux> btw, i added a no-gpl fate instance after seeing 4fefe91a
[07:38] <ubitux> and i also added the -DTRACE flag to the -DDEBUG box (which now fails but shouldn't)
[09:59] <ubitux> michaelni: got a rawvideo dump of the h264 race failing output
[10:16] <ubitux> it's fun, it's like there are some +1 at times
[10:17] <av500> [+1]
[10:54] <ubitux> michaelni: http://lucy.pkh.me/race.html
[10:54] <ubitux> here it is
[10:55] <ubitux> ±1 on most of the bytes afaict
[14:21] <KGB> [FFmpeg] michaelni pushed 10 new commits to master: http://git.io/80NRAg
[14:21] <KGB> [FFmpeg/master] get_bits: const correctness for get_bits_trace()/get_xbits_trace() arguments - Diego Biurrun
[14:21] <KGB> [FFmpeg/master] golomb: const correctness for get_ue()/get_se() function arguments - Diego Biurrun
[14:21] <KGB> [FFmpeg/master] h264_refs: Fix debug tprintf argument types - Diego Biurrun
[14:25] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/UpDsjA
[14:25] <KGB> [FFmpeg/master] nutdec: fix get_vb_trace() / -DTRACE compile - Michael Niedermayer
[14:30] <iive> i like how secretive the new bot is. shows up says something and then vanishes without a trace.
[14:41] <JEEB> I always disliked that in the github service somehow
[14:42] <av500> avoid getting kicked all day
[14:42] <av500> avoids
[14:48] Action: ubitux is wondering why libav didn't rename FATE into LATE </easy-taunt>
[15:27] <KGB> [FFmpeg] michaelni pushed 4 new commits to master: http://git.io/aljC8w
[15:27] <KGB> [FFmpeg/master] avcodec: Add missing include to version.h - jamal
[15:27] <KGB> [FFmpeg/master] postproc: Create version.h header - jamal
[15:27] <KGB> [FFmpeg/master] swresample: Create version.h header - jamal
[16:44] <Daemon404> [08:30] <@iive> i like how secretive the new bot is. shows up says something and then vanishes without a trace. <-- well it's name IS KGB
[16:45] <ubitux> :)
[16:49] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/osmNxg
[16:49] <KGB> [FFmpeg/master] ffprobe: remove unused "args" argument in writer init callback - Stefano Sabatini
[16:50] <nevcairiel> i kinda dislike these hit&run bots, but since i have joins/parts hidden now, it should just do what it does :p
[16:52] <Compn> someone try kicking / hugging him next time he comes in :P
[16:52] <Compn> or put dioxin in his soup
[17:09] <Daemon404> uh
[17:09] <Daemon404> ugh*
[17:10] <Daemon404> ffmpeg is still getting aac decoding wrong
[17:10] <Daemon404> tons of audio desync issues
[17:10] <Daemon404> all likely related to that patch from the xencoder dude
[17:10] <Daemon404> zencoder*
[17:10] <MP4_maniac> I
[17:11] <Daemon404> MP4_maniac, audio decoded from aac-in-mp4  is 1 or 2 video frames off
[17:12] <nevcairiel> what patch specifically?
[17:12] <nevcairiel> audio was spot-on in mp4 a while ago
[17:13] <Daemon404> not so now... (using ffmpeg cli)
[17:14] <Daemon404> i -thought- it was related to that patch by jon from zencoder, which i cant find atm
[17:14] <MP4_maniac> mov/mp4 demuxer specific?
[17:14] <Daemon404> but aac and mov/mp4 stuff gets touched so often i cant figure out what broke it
[17:14] <Daemon404> MP4_maniac, so far yes
[17:14] <nevcairiel> hm
[17:15] <MP4_maniac> it is related with assuming priming samples?
[17:15] <Daemon404> MP4_maniac, thats my guess
[17:15] <Daemon404> but it's only a guess
[17:15] <nevcairiel> good that i just ignore this audio delay sidedata crap =P
[17:19] <MP4_maniac> current ffmpeg's mov/mp4 demuxer assumes 1024 priming samples if priming samples is not specified in iTunSMPB
[17:19] <Daemon404> MP4_maniac, the current complain is that audio is 1 frame late
[17:19] <Daemon404> so i gues sthat would account for it?
[17:20] <MP4_maniac> aac 1 frame is 1024 samples
[17:20] <nevcairiel> why assume 1 frame when before this code was all added it was 0 and apparently fine?
[17:20] <Daemon404> it wasnt 0 before
[17:20] <Daemon404> it was 1024 before, and it didnt decode iTunSMPB
[17:20] <nevcairiel> previously there wasnt even code to deal with that
[17:20] <Daemon404> and before that it was 0
[17:21] <Daemon404> i startedg etting complains when this code was introduced
[17:26] <MP4_maniac> i dislike this ffmpeg's implicit 1024 priming samples. i believe there are people who consider encoder specific priming samples and set up edit list.
[17:28] <JEEB> IIRC the guy who did that was supposed to make it better, but never did if i remember correctly
[17:30] <MP4_maniac> i think current ffmpeg's assumption of implicit AAC priming samples conflicts with edit list...
[17:30] <JEEB> it does, more or less
[17:35] <Daemon404>     mov: Change default aac priming fallback from 2112 to 1024
[17:35] <Daemon404>     
[17:35] <Daemon404>     This fixes priming for faac
[17:35] <Daemon404> well gee
[17:35] <Daemon404> :/
[17:35] <Daemon404> one cruddy fix to another
[17:36] <Daemon404> i am of the optinion it should be 0
[17:36] <Daemon404> opinion*
[17:37] <JEEB> hahahaha
[17:37] <MP4_maniac> about lc-aac, apple: 2112, nero: 2624, faac: 1024, fghaacenc: variable :P
[17:37] <Daemon404> MP4_maniac, i am getting complains from tons of users (generally using professional editing softwazre) that its adding delay
[17:38] <Daemon404> aka should be 0
[17:38] <Daemon404> aka they set their stuff up properly
[17:38] <JEEB> Daemon404, yes the "correct" way would be to write it into the edit list, but not many things do that still :/ Otherwise it's a guessing game, and zero is as valid as it gets as well
[17:39] <Daemon404> a very vocal subset of users, may i add.
[17:39] <JEEB> yeah
[17:41] <MP4_maniac> mpeg's and apple's spec says explicit priming samples is specified by edit list. implicit priming samples makes sense only when the encoder is known
[17:42] <JEEB> ayup
[17:42] <MP4_maniac> otherwise, should be 0
[17:42] <JEEB> otherwise it just breaks others and creates herpy derpy results on others
[17:48] <michaelni> Daemon404, which ticket on trac describes this aac/mov/mp4 bug ?
[17:49] Action: michaelni likes to fix it ...
[17:49] <Daemon404> michaelni, it's quite simple:
[17:49] <Daemon404> teh default of 1024 is adding padding to things that shouldnt
[17:50] <Daemon404> apparently 1024 is faac's default
[17:50] <Daemon404> i dont htink there is a bug open
[17:50] <Daemon404> basically, we make assumptions about how to pad
[17:51] <michaelni> Daemon404, you assume things and some of this doesnt add up entirely so id like to know exactly whats not working so i can check if it really is the 1024
[17:51] <michaelni> we dont add 1024 padding btw, we remove 1024 
[17:52] <JEEB> yes
[17:52] <Daemon404> oh
[17:52] <JEEB> earlier 2112 was removed
[17:52] <Daemon404> well ive currently patched ffmpeg
[17:52] <Daemon404> and am tetsing
[17:52] <Daemon404> either way, using 1024 for everythingis still wrong.
[17:54] <JEEB> the amount of padding to be removed should be set via edit lists, which of course many things don't do. But still assuming things without knowing for sure is derp.
[17:55] <JEEB> there's a guy bisecting an a/v desynch with mp4 on #ffmpeg as well, we'll see if it's the same thing when he finishes it
[17:55] <michaelni> I dont have a single sample that specifies this via edit lists ... and iam happy to implement it if someone sends me a sample + a reference on how it should decode
[17:55] <Daemon404> JEEB, let me know
[17:56] <Daemon404> michaelni, i am of the opinion that it is technically correct to use 0 instead of 'trying' to compensate
[17:56] <MP4_maniac> michaelni: i can send a sample with pre-roll grouping (for decoder delay) and edit list (for encoder delay)
[17:57] <michaelni> MP4_maniac, please do (best in a trac ticket so i dont forget it if i work on it later)
[17:57] <MP4_maniac> oh, i have forgoten that iso publishes a sample :P
[17:58] <MP4_maniac> 14496-4 2004 Amd 24 2008 a9-aac-samplegroups-edit.mp4 <--- this
[17:58] <Daemon404> iso has test vecotrs? nifty
[17:58] <MP4_maniac> 14496-4 is conformance tesiting
[18:00] <michaelni> Daemon404, ill look into detecting faac and then making it 0 if its not faac
[18:00] <Daemon404> im not sure if it's "detectable"
[18:00] <Daemon404> this whole problem is really uglu
[18:00] <Daemon404> ugly*
[19:13] <saste> ubitux: yes fate-ffprobe was greatly useful
[19:35] <saste> ubitux: i'm going to push the whole ffprobe patchset in three days, if I read no comments
[19:58] <nevcairiel> speaking about ffprobe fate, what finally fixed the msvc tests? @Daemon404 ?
[20:00] <Daemon404> nevcairiel, switching to an input ffmeta file
[20:00] <Daemon404> instead of specifying utf8 on the cmd line
[20:00] <Daemon404> msys doesnt behave like normal windows wrt cmd line args
[20:00] <Daemon404> because MSYS Is Evil
[20:00] <Daemon404> etc
[20:03] <nevcairiel> because make is msys?
[20:03] <nevcairiel> i mean, ffprobe should be completely native
[20:05] <Compn> msys is evil.
[20:05] <Compn> you could always build mingw in cygwin... bwahaha
[20:06] <Daemon404> nevcairiel, msys passes args to stuff different than vis cmd.exe
[20:06] <Daemon404> wrt character encoding
[20:06] <nevcairiel> so it does happen because make runs everything in msys shells
[20:06] <Daemon404> yes.
[20:34] <KGB> [FFmpeg] michaelni pushed 2 new commits to master: http://git.io/bOfcbA
[20:34] <KGB> [FFmpeg/master] aacdec: detect faac and skip correct amount of startup samples - Michael Niedermayer
[20:34] <KGB> [FFmpeg/master] mov: remove default of 1024 sample skip for aac. - Michael Niedermayer
[20:36] <gnafu> Hehe.
[21:33] <Daemon404> i swear github's bot had color
[21:39] <gnafu> Daemon404: Maybe it's a configuration option./
[21:40] <ramiro> hi
[21:41] <ramiro> I have a video encoded with some smartphone that plays fine in vlc, but becomes rotates when uploaded to youtube.
[21:42] <ramiro> i imagine it's some kind of mp4 metadata ignored by vlc but taken into account by youtube (or the other way around), but I don't see any metadata from ffmpeg -i file.mp4
[21:42] <ramiro> any suggestions on how to find this out and make youtube play the file properly (short of re-encoding the file with ffmpeg)
[21:43] <Daemon404> ffprobe -show_streams file.mov
[21:43] <Daemon404> look for the rotate tag
[21:43] <Daemon404> if you dont want youtube to rotate it, kill the tag
[21:48] <MP4_maniac> rotation can be specified by matrix within tkhd atom
[21:51] <ramiro> Daemon404: hm, I see no rotate tag
[21:52] <ramiro> MP4_maniac: how do I see that without delving into the source code?
[21:55] <ohsix> >matrix
[21:59] <MP4_maniac> ramiro: search tkhd atom by 'tkhd' string in binary, and its last 8 bytes is width and height. matrix is placed directly before them. no rotation matrix shall be { 0x00010000,0x00000000,0x00000000,0x00000000,0x00010000,0x00000000,0x00000000,0x00000000,0x40000000 }
[22:00] <Daemon404> or use boxdumper?
[22:01] <MP4_maniac> yes. boxdumper can see the matrix. also dumpster can see and edit
[22:06] <MP4_maniac> uhg, mvhd atom also has matrix...
[22:06] <Daemon404> fwiw iphones and android phones use the rotate tag.
[22:10] <ods15> yes, and windows doesn't understand the rotate tag...
[22:10] <ods15> dunno about iphones, but yes for android
[22:37] <Daemon404> 'windows; >
[22:37] <Daemon404> ?
[22:56] <ramiro> saste: hi
[22:56] <saste> ramiro, hey
[23:26] <ubitux> saste: okay, will look at it in a moment :)
[23:28] <ubitux> saste: i'm a bit lost about the order
[23:29] <ubitux> can you give me a few hints? :)
[23:31] <saste> ubitux: http://pastebin.com/KwncDsCr
[23:31] <ubitux> thx :)
[23:33] <ubitux> saste: you didn't send "c2c95feeaa92b63d74c6c4d9c39e52bf79858124 ffprobe: drop SECTION_ENTRY macro" right?
[23:34] <ubitux> ah i see it
[23:34] <ubitux> sorry :)
[23:47] <saste> ubitux, thanks
[23:48] <ubitux> saste: rework/fix {flat,ini} writer are unchanged?
[23:48] <saste> yes
[00:00] --- Sat Sep 29 2012