[Ffmpeg-devel-irc] ffmpeg-devel.log.20121004

burek burek021 at gmail.com
Fri Oct 5 02:05:02 CEST 2012


[00:02] <saste> michaelni, wbs: if it is too complicate to merge segment patches from martin, i'll try to merge them later
[00:03] <saste> my only problem is that i don't have much time to spend on ffmpeg in this period
[01:34] <KGB> [FFmpeg] michaelni pushed 1 new commit to master: http://git.io/QIEtPQ
[01:34] <KGB> [FFmpeg/master] movenc: force video timebase to be 0.1ms precisse at least. - Michael Niedermayer
[03:10] <KGB> [FFmpeg] michaelni pushed 2 new commits to master: http://git.io/YNs3jA
[03:10] <KGB> [FFmpeg/master] afq: update remaining samples variable. - Michael Niedermayer
[03:10] <KGB> [FFmpeg/master] afq: sanity assert on remaining_samples - Michael Niedermayer
[06:55] <Daemon404> >.>
[06:55] <Daemon404> i can't believe people are -supporting- a file named time.h
[09:11] <ubitux> wow it's crazy
[09:11] <ubitux> the ffmpeg h264 fate tests are suddenly waaaay faster in 
[09:11] <ubitux> helgrind
[09:13] <ubitux> or maybe it's a placebo
[09:17] <ubitux> derp, mistake.
[10:06] <ubitux> thresh: so you're here to bring an awesome git rainbow bot? :)
[10:06] <thresh> exactly
[10:06] <av500> thresh: why are you not in siberia?
[10:07] <av500> http://www.wired.co.uk/news/archive/2012-09/18/russian-diamond-smorgasbord
[10:07] <thresh> av500: those are technical diamonds, not fancy enough for us Moscovites
[10:08] <thresh> ubitux: got anything to push? :)
[10:08] <thresh> it should work now
[10:08] <ubitux> mmh i don't think so, but lemme check
[10:09] Action: ubitux is looking for a typo
[10:14] <ubitux> well i'll wait instead :)
[13:30] <ubitux> thresh: michael pushed some stuff
[13:30] <ubitux> doesn't seem to appear
[13:31] <thresh> yeah I've just seen.
[13:31] <thresh> michaelni: can you show me the output of git push?
[13:43] <michaelni> thresh, last git push http://pastebin.com/y00wR2w0
[13:44] <thresh> weird, but thanks
[13:45] <ubitux> maybe it's a problem with the dns redirect?
[13:45] <thresh> no, I've got it
[13:47] <thresh> I think I fixed it
[14:34] <ubitux> http://www.flickr.com/photos/flameeyes/7906103846/lightbox/ VDD12
[14:34] <ubitux> Daemon404: where are you btw? :)
[14:35] <ubitux> since i "missed" you then :p
[14:39] <KGB> [FFmpeg] michaelni pushed 25 new commits to master: http://git.io/3-62Aw
[14:39] <KGB> [FFmpeg/master] blowfish: Factorize testing into a separate function - Martin Storsjö
[14:39] <KGB> [FFmpeg/master] blowfish: Fix CBC decryption with dst==src - Martin Storsjö
[14:39] <KGB> [FFmpeg/master] blowfish: Add more tests - Martin Storsjö
[15:00] <j-b> thresh: cone?
[15:00] <thresh> j-b: it will join once there is a commit to show
[15:01] <thresh> you provide a channel to spam to in a hook, not in a daemon instance
[15:01] <j-b> oh, nice
[15:32] <michaelni> j-b, thresh http://pastebin.com/gimSnw4T
[15:32] <thresh> oups..
[15:32] <j-b> michaelni: well, well well
[15:33] <thresh> michaelni: re-push, please
[15:33] <cone-205> ffmpeg.git 3Michael Niedermayer 7libavcodec/mpeg12enc.c libavcodec/mpegvideo.h: mpeg2enc: support and use frame_rate_ext when needed
[15:36] <michaelni> done, thx
[15:36] <michaelni> ive also disabled the github KGB
[16:01] <durandal_1707> michaelni: frame.data[X] for planar format seems to not come one after another
[16:03] <durandal_1707> so instead of using single malloc function i need to use each one per channel
[16:04] <durandal_1707> issue doesn't seem to happen with 16bit but does for 24(32)
[16:04] <nevcairiel> why would you malloc the channels manually, there is get_buffer to do that
[16:05] <durandal_1707> i need that for < 24 bits per sample
[16:12] <kierank> huh
[16:12] <kierank> why does that make a difference
[16:13] <durandal_1707> in > 16 bit case i do not need extra buffer, i just use frame
[16:47] <saste> michaelni: where can I find a file with repeat_pict != 0?
[16:48] <michaelni> saste, some telecine stuff should use it
[16:49] <Daemon404> ubitux, you can only see my head
[16:49] <Daemon404> middle-ish...
[16:49] <saste> michaelni, so the question turns to: where can I find some telecine stuff? ;-)
[16:51] <michaelni> samples archive probably, i dont know a specific file ATM either :(
[16:51] <saste> uhm... well I'll have to do a brute force scan
[16:51] <saste> something in fate samples?
[16:52] <divVerent> sawe can narrow it down a little
[16:52] <divVerent> it's mostly used on DVDs
[16:52] <divVerent> and very unlikely in TS streams
[16:53] <divVerent> as these tend to be real interlaced norally ;)
[16:54] <michaelni> saste, maybe iive or rich know a sample
[16:54] <saste> uhm so in some DVD vobs?
[16:54] <divVerent> I'd expect it there, yes
[16:54] <divVerent> but even there it is unlikely
[16:54] <Daemon404> almost all anime dvds ever
[16:54] <Daemon404> for one
[16:54] <divVerent> you are looking for a NTSC DVD with both 24p and 30i content
[16:54] <iive> repeat_pict, that was used for software telecine?
[16:55] <divVerent> yes
[16:55] Action: saste realizes how useful -t option would be in ffprobe
[16:55] <Daemon404> protip: anything that isnt TIVTC is crap for ivtc
[16:55] <michaelni> we need a "ffprobe -find-the-sample-i-want" option :)
[16:55] <divVerent> hehe
[16:56] <iive> we have Compn for that.
[16:56] <divVerent> maybe we need a ffprobe index of all fate samples :P
[16:56] <saste> michaelni, i'm going to try something link ffprobe -t 2 -show_frames | grep pict_type | grep -v pict_type=0
[16:56] <saste> but i need to implement -t it will take forever
[16:56] <saste> *or it will ...
[16:57] <michaelni> saste, you could just av_assert0() on the value you want and then run ffmpeg -f null - over all and search for a Abort in the output
[16:57] <divVerent> ( ffprobe -show_frames | grep pict_type | grep -v pict_type=0 ) & sleep 1; killall ffprobe ;)
[16:58] <saste> also a stream selector would be useful (and maybe i should replace -t 2 with -n 20)
[16:59] <nevcairiel> saste: i can upload you some telecine dvb samples if you want
[16:59] <saste> nevcairiel, yes that would help
[16:59] <nevcairiel> soft telecine is what you want, right? so repeat flags in the stream
[17:02] <nevcairiel> saste: http://files.1f0.de/samples/Dext.mpg and http://files.1f0.de/samples/eviltrees.ts
[17:03] <nevcairiel> both contain repeat flags
[17:04] <saste> nevcairiel, thanks
[17:10] <ubitux> Daemon404: can you give some more hints, or show me somehow? :)
[17:11] <Daemon404> maybe is mspaint-circle it in a bit
[17:11] <Daemon404> ;p
[17:11] <Daemon404> s/is/ill/
[17:18] <ubitux> thx :)
[17:18] <ubitux> Daemon404: version it, so we can contribute by adding some nicks
[17:27] <Daemon404> isnt it easier to use something with a tagging ability
[17:27] <Daemon404> does flickr have that?
[17:28] <ubitux> no idea
[17:28] <av500> flickr has
[17:29] <Daemon404> youd have to contact diego to do that
[17:29] <Daemon404> hes not usually on irc, i think
[17:29] <Compn> Daemon404 : avtime.h ? :P
[17:31] <Compn> Daemon404 : dondiego is on irc right now (if you mean diego biurrun )
[17:32] <Daemon404> no i mean flameeyes
[17:36] <saste> ffprobe -select_streams or -specify_streams?
[17:37] <saste> or both through an alias...
[17:38] <Compn> oh
[17:38] <Compn> hehe
[17:41] <cone-205> ffmpeg.git 3Michael Niedermayer 7ffplay.c: ffplay: add a 3rd state to infbuf for autodetection
[17:41] <cone-205> ffmpeg.git 3Michael Niedermayer 7ffplay.c: ffplay: autodetect realtime streams and enable infbuf
[17:43] <ubitux> michaelni: what's with this indent? :DD
[17:53] <saste> michaelni, who owns the nut trademark?
[17:55] <av500> apple?
[17:59] <ubitux> :))
[18:31] <j-b> michaelni: cone-205 works?
[18:50] <michaelni> j-b, so far it seems it works
[18:51] <j-b> michaelni: cool. Anything else?
[18:55] <michaelni> atm it seems all fine
[21:35] <cone-205> ffmpeg.git 3Tim Nicholson 7libavformat/movenc.c tests/ref/fate/acodec-pcm-s24be tests/ref/fate/acodec-pcm-s32be: movenc.c: Add support for >16bit BE flavours
[21:53] <spaam> michaelni: http://guru.multimedia.cx/pp-vs-spp-filters/ does not work..
[21:54] <spaam> michaelni: i guess its related to http://multimedia.cx
[22:06] <spaam> oh fuck.. that mail was from first oct.. ;S
[22:06] <spaam> bah. fail
[22:08] <Daemon404> lul
[22:08] <ohsix> thine virginal earholes
[22:49] <llogan> burek: how do you envision ffmpeg would supply an official deb package? PPA?
[23:08] <burek> llogan, you mean this: https://help.launchpad.net/Packaging/PPA ?
[23:12] <burek> I was thinking more like doing it with our own infrastructure, but if launchpad is willing to build for those architectures itself (x86, AMD64 and LPIA), sure, why not
[23:12] <burek> even less trouble
[23:14] <burek> the goal is to provide the alternative to debian/ubuntu users, who still favor ffmpeg over avconv
[23:18] <llogan> burek: i don't know the capabilities of launchpad. i'm just wondering what would be easiest for users.
[23:21] <burek> well, imho, it is more convenient to have an apt repository (like launchpad), but even providing only a deb package (to be used with dpkg) is also good
[23:22] <burek> i've noticed many of my friends are moving from windows to linux world, choosing ubuntu mostly
[23:22] <llogan> would you be willing to set this up or maintain it?
[23:22] <burek> so, I guess it is very popular, hence the idea of deb package for ffmpeg
[23:23] <burek> well, of course, I just need some help because I never did something like that before
[23:25] <llogan> i think it is a good idea, but i too have no experience with this (i've only "maintained" PKGBUILDS)
[23:25] <Genia> hi all - new here - trying to use av_read_frame as input to apple's hw decoder. (yes, i know about vda.c, but for what i'm doing, i need to run vda externally). av_read_frame produces raw nals when reading .mp4, and start codes when reading HLS. Since the apple stuff wants a nal format, is there a filter that would convert start codes to nals?
[23:28] <llogan> burek: users might expect it to work as a libav replacement for any dependencies. not sure how you would address that.
[23:33] <burek> well we can just copy/paste the logic that was present before, while ffmpeg was in debian repo
[23:33] <burek> to be honest, im not sure how did they resolve dependancy problems with avconv
[23:33] <burek> i mean libavtools
[23:33] <burek> did they keep the same names and paths and stuff or what
[23:34] <burek> or all the projects had to manually change the deps to libav
[23:34] <llogan> i assume the first. plus keeping faux-ffmpeg for winff and other cli users.
[23:35] <llogan> i'm not sure how they will do it once they adopt upsteam that no longer contains ffake-ffmpeg.
[00:00] --- Fri Oct  5 2012


More information about the Ffmpeg-devel-irc mailing list