[Ffmpeg-devel-irc] ffmpeg-devel.log.20130402

[burek]

[00:07] <durandal_1707> ubitux: can you run my aphaser branch under valgrind, i can get s32 to work (while at same time s32p and s16/s16p works)
[00:09] <ubitux> i'm on my laptop right now, so it will take some little time
[00:09] <ubitux> what command do you have in mind while it's building?
[00:10] <durandal_1707> the one that forces different sample formats
[00:10] <durandal_1707> exp. s32 as that one gives some strange output
[00:11] <durandal_1707> -af aformat=sample_fmts=s32,aphaser
[00:22] <ubitux> seems ok so far
[00:23] <durandal_1707> does is sound fine also?
[00:24] <ubitux> mmh i haven't my headphones so i can't tell
[00:24] <ubitux> valgrind is not complaining at least
[00:25] <durandal_1707> hmm what you see with ffplay?
[00:27] <ubitux> not sure what you mean
[00:27] <ubitux> same as you i guess?
[00:28] <ubitux> i have to go, ask me tomorrow
[00:28] <durandal_1707> if you compare spectrogram with s32 and s32p they differs here
[00:28] <durandal_1707> i have horizontal white periodical lines
[01:50] <cone-981> ffmpeg.git 03Michael Niedermayer 07master:7f4b588fdfab: ffplay: use seek per time for ogg by default
[05:23] <cone-509> ffmpeg.git 03Michael Niedermayer 07master:4dd31f88f513: cavs: fix "warning: initialization discards const qualifier from pointer target type"
[10:34] <saste> does someone know if hendrik leppkes is on irc?
[10:36] <JEEB> I think that's nev
[10:36] <JEEB> yup
[10:53] <saste> JEEB, nevcairiel ?
[11:02] <JEEB> saste, yes
[11:03] <saste> nevcairiel, you ok if i add your name to the wiki? btw do you have a multimedia wiki account?
[11:03] <saste> in case you don't, you can ask Mike Melanson for that
[12:33] <ubitux> got a new funny filter incoming
[12:34] <saste> ubitux, what??
[12:36] <ubitux> i just need to adjust a few things and i'll make a demo :p
[12:43] <ubitux> i need a sample now :p
[12:46] <ubitux> saste: http://lucy.pkh.me/vignette.webm 
[12:46] <saste> eheh :)
[12:47] <ubitux> i think i can do a james bond effet as well
[12:50] <ubitux> saste: vignette branch on my github if you want to have fun as well
[12:53] <ubitux> above video generated with -f lavfi -i mptestsrc -vf 'curves=vintage,vignette=edge_dist_factor=1.6:r1=(0.03*random(1)+0.79)*h/2:r1=(0.03*random(1)+0.79)*w/2'
[12:53] <ubitux> after this we'll need some generate grain and random vertical line to simulate the old camera effect
[12:55] <ubitux> erm it should have been r2, not r1, but well
[12:57] <__gb__> hi ubitux, I think the hwaccel task should be fine as is (split from the VA encoding tasks) -- I actually had some ancient code that performs that
[12:57] <ubitux> wow, blast from the past :)
[12:57] <ubitux> ok :)
[12:58] <__gb__> but this needs refreshing and improvements in some other parts
[12:58] <__gb__> for encoding, it should also be fine, most of the work already happens in the driver :)
[12:59] <__gb__> ubitux, oh, btw, I don't have the time right now, but at some point some people are also interested in merging the vaapi support to mplayer1
[13:00] <__gb__> in my vision, that should have been a streamlined version of what I had, i.e. only handle the VA/X11 case
[13:01] <__gb__> (sorry, reading backlog from before sick/holidays) :)
[13:04] <ubitux> i'm not active on mplayer1 anymore :p
[13:05] <__gb__> ah ok, I didn't follow :)
[13:06] <nevcairiel> saste: i dont have an account yet, but i'll request one
[13:07] <saste> nevcairiel, ok, thanks
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/0.10:8b72bcba713a: Write broken aac frames to mov files instead of skipping them.
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/0.11:2a30d4309f24: Write broken aac frames to mov files instead of skipping them.
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/0.9:38cd7f66b73a: Write broken aac frames to mov files instead of skipping them.
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/1.0:bd6e58d5cadd: Write broken aac frames to mov files instead of skipping them.
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/1.1:8ba3198549b9: Write broken aac frames to mov files instead of skipping them.
[13:09] <cone-798> ffmpeg.git 03Carl Eugen Hoyos 07release/1.2:0b6d5f27c811: Write broken aac frames to mov files instead of skipping them.
[13:12] <saste> nevcairiel, btw, maybe you want to co-mentor the other hwaccel task as well
[13:12] <nevcairiel> i figured someone would ask this, but i'm not really into encoding, but i suppose if noone else volunteers
[13:19] <ubitux> saste: http://lucy.pkh.me/gonenutty-vintage-vignette.webm  the encode quality is bad (default settings, i didn't care enough), but it somehow gives a better idea
[13:22] <ubitux> now i need to get rid of the ugly halo in the center
[14:17] <cone-798> ffmpeg.git 03Paul B Mahol 07master:1de11a7cab53: libmp3lame: allow joint stereo to be disabled
[14:24] <durandal_1707> ubitux: so, tried it again?
[14:44] <michaelni> Tjoppen, ping on: Matthieu Bouron (5.0K) [FFmpeg-devel] [PATCH] lavf/mxfdec: handle identification metadata
[14:57] <cone-798> ffmpeg.git 03Michael Niedermayer 07master:c09da45ffbeb: dsputil: Fix "warning: initialization discards const qualifier from pointer target type"
[16:10] <cone-798> ffmpeg.git 03highgod0401 07master:90793591417f: deshake opencl based on comments on 20130402 3rd
[16:24] <ungureanuvladvic> hello! i am trying to find the 0.8.12 version of ffmpeg but on the github there is 0.7.7 then it jumps to 0.8.13. any idea which hash has 0.8.12 ?
[16:25] <durandal_1707> ungureanuvladvic: why that old version?
[16:26] <ungureanuvladvic> durandal_1707: i am working on a auto-focus mechanism for an ip-camera and to take frames on rtsp from that specific brand of cameras it seems that only 0.8.12 works well. the last version of ffmpeg complains a lot about the stream and gives me malformed frames.
[16:27] <ungureanuvladvic> durandal_1707: i am also working on a RPi and on my device i have a cross-compiled version of 0.8.12 which works flawlessly but i cannot find the src for that one, have it like .deb..
[16:28] <durandal_1707> michaelni: grep avconv *.c inside source lists some strange stuff, like .avconv directory and preset files
[17:08] <durandal_1707> i think i asked it already but forget about it, why sox muxer dumps noise at end?
[17:16] <petar> ffmpeg_opt.c:get_preset_file_2() looks for *.avpreset files in some .avconv directory.  is that expected behaviour?  presets in ~/.ffmpeg don't work for me anymore.. at least not with 1.1.3.
[17:43] <ungureanuvladvic> anybody know the hash of 0.8.12 version? 
[17:50] <petar> ugh, scratch that, I was using -pre instead of -[vas]pre.  I guess -pre is dead?
[18:11] <michaelni> ungureanuvladvic, git show n0.8.12
[18:11] <michaelni> 858c3158b58eafee2fedd9d83651c06aa57ef217
[18:25] <cone-798> ffmpeg.git 03Michael Niedermayer 07master:356363c89830: ffmpeg: Fix forcing of the framerate for stream copy on input
[18:31] <ungureanuvladvic> michaelni: thank you. another question, maybe off-topic: why didn't you guys applied for Google Code-In this year also? last year i had so much fun working for you people .
[18:41] <michaelni> ungureanuvladvic, which gci do you mean ? 2012 or 2013 ?
[18:42] <ungureanuvladvic> 2012 i participated, this year i haven't seen ffmpeg in the organizations list.
[18:44] <michaelni> We where in gci 2011 together with videolan under videolan name if i remember corerctly
[18:45] <michaelni> in 2012 i think we wherent and 2013 is in december/january 2014 i suspect
[18:45] <ungureanuvladvic> michaelni: ah w8, yes 2011 that is right. i was there. i messed up the dates. 
[18:46] <ungureanuvladvic> michaelni: anyway great stuff what you guys are doing with FFMPEG, really helped me a lot in my work with ip cameras. 
[19:56] <thegeek> has anyone tested the new opencl stuff under msvc?
[20:08] <mateo`_> Tjoppen: can you review the mfxdec: handle identification metadata ?
[20:08] <mateo`_> +patch
[20:36] <ubitux> durandal_1707: got some time now
[20:36] <ubitux> so you want me to check if i hear some differences?
[20:42] <ubitux> durandal_1707: i ran for f in s32 s32p s16 s16p; do valgrind ./ffmpeg_g -i ~/samples/feathers.flac -af aformat=sample_fmts=$f,aphaser out-$f.wav; done
[20:42] <ubitux> valgrind didn't detect anything
[20:42] <ubitux> the distortion sounds the same in every sample
[21:51] <nevcairiel> so who broke Windows?
[21:56] <ubitux> nevcairiel: i'm sure it broke by itself
[22:08] <iive> nevcairiel: ballmer, who else? ;)
[22:13] <cone-798> ffmpeg.git 03Michael Niedermayer 07master:ccc25378bd1b: avformat: fix atomatic format selection for (s)segment
[22:29] <mateo`_> anyone to review the mxfdec patch ?
[23:01] <michaelni> mateo`_, the patch looks harmless, ill apply it unless someone wants to review
[23:12] <nevcairiel> ffmpeg.c(1889) : error C4013: 'elseif' undefined; assuming extern returning int
[23:13] <nevcairiel> i'm confused that this works in any compiler, is elseif a legal construct in some newer C version? :d
[23:14] Action: ubitux doesn't see any "elseif"
[23:14] <nevcairiel> where did the compiler find it then? <.<
[23:14] <ubitux> :)
[23:15] Action: michaelni has only elseif matches in arm asm
[23:16] <nevcairiel> that line contains a normal else if
[23:16] <nevcairiel> i wonder what happened
[23:16] <ubitux> did you check it's a 0x20?
[23:17] <nevcairiel> seems to be
[23:18] <ubitux> nevcairiel: maybe something broke in c99-to-c89?
[23:19] <nevcairiel> its really the only sane possibility, but its too late to start long investigations now, i'll try to remember for tomorrow
[23:19] <ubitux> i'm guessing it chokes on the AVRational
[23:20] <ubitux> (AV_TIME_BASE_Q)
[23:20] <nevcairiel> .. and then removes a space to screw with us?
[23:20] <ubitux> don't ask me what kind of magic c99-to-c89 does
[23:22] <nevcairiel> the code seems a bit odd as it is, wouldnt it be easier to rescale the "1" to the proper timebase and add it, and not scale value, add one, scale value back? :d
[23:23] <michaelni> it would be easier but it would accumulate rounding errors
[23:23] <ubitux> it looks like you are looking the source of the problem in the wrong place
[23:25] <nevcairiel> I was just looking at the commit that seems to have caused it and found the code complicated, but didn't think of rounding problems
[23:26] <nevcairiel> i'll check the converted source code tomorrow
[23:48] <Tjoppen> mateo`_: just got back from a long trip
[00:00] --- Wed Apr  3 2013