[Ffmpeg-devel-irc] ffmpeg-devel.log.20130626

burek burek021 at gmail.com
Thu Jun 27 02:05:02 CEST 2013


[00:03] <Daemon404> nevcairiel, i can NOT get relative samples paths to work
[00:03] <Daemon404> it just says "samples location not specified"
[00:06] <superware> I'm getting "Option mpegts_original_network_id not found." yet documentation says it exists
[00:11] <Daemon404> finally... trial and error \o/
[00:24] <Daemon404> nope still failed
[00:24] <Daemon404> because if the dir you run fate.sh from is dfferent than where make fate gets run
[00:24] <Daemon404> it doesnt work
[00:24] <Daemon404> what a shitshow.
[00:25] <Daemon404> nevcairiel, what dir do you run it from?
[00:49] <Daemon404> i wish carl would start putting hashes in ticket closes
[00:54] <wm4> why not set it up such that trac listens on git commit?
[01:08] <michaelni> wm4, lack of volunteer with free time (to investigate how to do it and to do it), also i should update trac to the latest version
[01:10] <michaelni> i read that trac 1.0 has something called tracgit
[01:10] <michaelni> but i dunno if thats related
[01:10] <michaelni> the feature list sounds royally useless
[01:11] <Daemon404> hey michaelni, a ffmpeg-specific change got clobbered in cllc.c
[01:11] <Daemon404> during the latest set of merges
[01:11] <Daemon404> (the ff_get_buffer thing)
[01:13] <michaelni> Daemon404, which merge / file /line /hash ...
[01:14] <Daemon404> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=9328ae484338b70a7f2dbcd420bbbac499df76cc
[01:14] <Daemon404> specifically
[01:14] <Daemon404> http://git.videolan.org/?p=ffmpeg.git;a=blobdiff;f=libavcodec/cllc.c;h=797d07b8eeff860bb3152e3f4e4cb388fa1d4510;hp=ab8d46aa52685dd8c23518aa5e33d15d2006637b;hb=9328ae484338b70a7f2dbcd420bbbac499df76cc;hpb=bad8f04b0454441edb0b7a1565e28c0b1a07f7ed
[01:14] <Daemon404> shouldnt have happened
[01:18] <michaelni> Daemon404, oops, will fix
[01:23] <michaelni> Daemon404, fixed
[01:24] <Daemon404> cool.
[01:26] <michaelni> j-b, was cone forgotten when the server was resurrected by the necromancers ?
[01:28] <j-b> michaelni: no idea. I can ask
[01:28] Action: j-b is in the USA, so I just landed and saw that the server was up
[01:29] <michaelni> j-b, thanks!
[01:30] <BBB-work_> yay j-b is alive
[01:30] <BBB-work_> j-b, can't me meet in the south bay tomorrow?
[01:30] <BBB-work_> sf is so far away
[01:32] <j-b> BBB-work_: see with Alex
[01:32] <j-b> he is the lead.
[01:32] <j-b> michaelni: asked
[01:32] <michaelni> j-b, thanks
[01:52] <llogan> what happened to the server?
[01:58] <Plorkyeran> there's a pretty standard trac plugin for updating tickets from commit messages that doesn't really need any configuration
[02:00] <Plorkyeran> I think I ended up hacking together my own thing for Aegisub though
[02:00] <Plorkyeran> since the standard plugin didn't set milestones based on what branch the commit is on
[02:00] <Plorkyeran> and interacted weirdly with the github post-receive hook
[03:17] <j-b> BBB-work_: well, I live in North SF, and the conf finishes late (7pm) and I am not motorized so, SF is way easier for me :)
[03:19] <Compn> cant just meet up over skype
[03:19] <Compn> have to meet in meatspace
[03:29] <kierank> drinking beer over skype isn't as fun
[03:42] <Compn> you ever done it ?
[03:42] <Compn> dont knock it til you tried it
[03:44] <kierank> yes
[03:44] <kierank> it involved someone spilling a can onto their laptop
[03:44] <Compn> sounds like fun :P
[03:45] <Compn> also probably why should move to smokeable intoxicants, esp in california
[03:46] Action: kierank doesn't do that
[03:46] <kierank> drugs are bad mmkay
[03:59] <Compn> deadly alcohol vs non toxic cannabis eh?
[04:06] <drv> drink cola instead ;)
[04:06] <kierank> both are good in moderation
[04:07] <BBB> j-b: I don't think I can be in SF at that time, I have work, y'know? :)
[04:26] <highgod> Hi, I want to ask a quesion.how can I get the number of slices in on frame? slice_group_count? thanks
[09:27] <superware> should ffplay experience difficulties with an h.264-ts-udp stream?
[09:27] <av500> with yours, yes
[09:28] <superware> if I use VLC to save the stream to a .ts file (raw input -> file), ffplay plays it nicely
[09:30] <superware> what does it mean?
[09:50] <superware> can someone please help me? http://pastebin.com/ZyNxavHJ
[10:01] <ubitux> -f h264 is wrong
[10:01] <av500> there is already -f mpegts
[10:02] <superware> well, it doesn't matter :(
[10:04] <superware> actually it does, please see http://pastebin.com/VjK8dVah - now it doesn't even detect the stream properties
[10:06] <superware> ubitux?
[10:06] <ubitux> i don't know
[10:08] <av500> Part of datagram lost due to insufficient buffer size
[10:09] <superware> av500: udp://0.0.0.0:1234?buffer_size=100000 same
[10:09] <av500> same error message?
[10:10] <superware> "Part of datagram lost due to insufficient buffer size" yes
[10:11] <superware> VLC plays it nicely, that's why I'm so frustrated
[10:11] <av500> you know
[10:11] <av500> ffmpeg could have a bug
[10:13] <superware> well, if I'm using VLC to stream an h264-ts-udp stream, then ffplay plays it
[10:13] <superware> I guess it's something specific to that stream
[10:13] <av500> indeed
[10:15] <superware> ok, so I thought maybe it's a problem with the format probing, so maybe it's possible to tweak options to make it work
[10:15] <superware> because when VLC saved the raw stream to a file, ffplay handles it nicely, meaning the problem has something to do with udp
[10:16] <michaelni> highgod, slice_num or current_slice
[10:23] <highgod> oh, thanks michaelni
[10:25] <highgod> I use if(code & 0x80) to detect whether this frame is start, if start, I push the previous slices in to queue
[10:25] <highgod> is it ok?
[10:25] <highgod> h264_probe
[10:26] <cone-855> ffmpeg.git 03Carl Eugen Hoyos 07master:35aed74fdc7b: Use AV_RN32 for an unaligned read in the mxg demuxer.
[10:26] <highgod> michaelni:our scale performance has been tested, I will submit the patch and the banchmark after I recheck it
[10:39] <michaelni> highgod, shouldnt there be a AVParser to split things into access units
[10:53] <highgod> michaelni:thanks, our app is get h264 slice bitstream from rtp stream, then send them to ffmpeg decoder. But I don't konw how many slices in on frames, so I have to detect the bit stream
[11:14] <nevcairiel> Daemon404: i cheated and created a second empty samples directory relative to the run path so the check succeeds
[11:15] <michaelni> highgod, the correct way to detect frame boundaries is described in the h264 spec, its a bit more complex than &0x80, but this correct detection should be implemented in the AVParser / shared with the AVParser
[11:16] <nevcairiel> but dont try to understand the code in the h264 parser, it'll just give you a headache
[11:43] <wm4> why does ffmpeg not use structs to define the contents of packet side data?
[11:44] <av500> what does it use?
[11:45] <wm4> stuff like AV_WL32
[11:46] <wm4> with a byte layout loosely defined in doxygen
[11:47] <av500> isnt packet side data opaque?
[11:47] <av500> aka just a blob?
[11:47] <wm4> in theory it is, but it appears it's mostly used for "internal" communication from demuxer to decoder
[11:47] <wm4> all within libav*
[11:48] <wm4> (of course not with all side data types)
[11:50] <av500> struct layout is compiler specific
[11:50] <av500> no?
[11:51] <wm4> I don't assume this form of packets is stored on disk or transferred between computers or even processes
[11:54] <wm4> e.g. AV_PKT_DATA_PARAM_CHANGE is generated in libavformat/utils.c and read in libavcodec/utils.c, and that's it
[11:54] <highgod> Hi, michaelni, can I use ffmpeg API to detect? is there any apis to implement this?
[11:54] <av500> wm4: lavc and lavf are not married
[11:54] <av500> one can use only one
[11:55] <wm4> still makes a rather awkward API
[11:56] <nevcairiel> yeah it was kinda silly to use
[11:56] <nevcairiel> i had to deal with it a bit, and it didnt really feel good
[11:57] <nevcairiel> but i'm done with that part now, so don't change it, or i'll have to touch it again!
[11:57] <nevcairiel> :)
[12:03] <superware> vlad_starkov: hi
[12:54] <wm4> lol
[12:54] <wm4> seriously
[12:54] <wm4> complaining about removing copyright notices... from examples?
[12:56] <cone-855> ffmpeg.git 03Carl Eugen Hoyos 07master:1235e91b3080: Require pthreads for compilation with OpenCL support.
[13:14] <wm4> does anyone know genpts actually does?
[13:14] <wm4> I've got a performance problem if pts jumps are involved (or at least that's what I expect)
[13:15] <wm4> in that case it seems to read the whole stream until EOF before returning the next packet, or something equally silly
[13:40] <nevcairiel> in my experience the pts it generates are not worth generating, they are usually pretty bad
[13:47] <cone-855> ffmpeg.git 03Michael Niedermayer 07master:87bc64893085: libavfilter/src_movie: fix which packet is reset
[13:47] <cone-855> ffmpeg.git 03Michael Niedermayer 07master:ee97982408c8: avfilter/src_movie: Fix handling of packet size for video
[13:48] <michaelni> genpts generates pts from dts so it requires correct dts on all frames
[15:15] <cone-855> ffmpeg.git 03Michael Niedermayer 07master:034b31df2cd0: swscale: Fix PAL8 input with alpha
[15:18] <ubitux> thank you michaelni for fixing all these bugs :)
[15:55] <j-b> michaelni: cone is back.
[16:25] <saste_>                     char tagbuf[32], cortag[32];
[16:25] <saste_> what is "cortag"??
[16:26] <ubitux> correct tag?
[16:28] <saste_> ubitux, good guess
[16:28] <ubitux> corguess
[16:28] <saste_> i'm writing a transcoding example
[16:29] <saste_> and fighting with various cryptic errors
[16:29] <saste_> how should work this high-level API?
[17:41] <michaelni> j-b, great, thanks
[18:26] <saste_> what's the purpose of the chomp bitstream filter?
[18:30] <saste> I mean in which cases is it useful?
[18:51] <Daemon404> [05:14] <+nevcairiel> Daemon404: i cheated and created a second empty samples directory relative to the run path so the check succeeds <-- this is exactly what i did
[18:51] <Daemon404> <_<
[18:58] <saste> how can i annotate a patch?
[18:58] <Daemon404> --annotate ?
[18:59] <saste> Unknown option: --annotate
[19:02] <Daemon404> it's part of git-send-email
[19:02] <Daemon404> personally, i always format-patch first and just use vim.
[21:30] <cone-855> ffmpeg.git 03Jean Delvare 07master:ff995e2b6f25: doc/filters: Fix texi syntax
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:3aa57e1582af: examples/muxing: remove useless instruction
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:c58d535b2f99: examples/Makefile: disable -O2 optimizations
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:47c9887ecaa7: lavc/utils: improve feedback in case of invalid packet size
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:08b99be7c474: lavf/mux: rename variable cortag -> tagbuf2 in init_muxer()
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:418b9454ff65: lavf/movenc: improve error feedback in case malformed AAC bitstream is detected
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:7b38c4c95f3b: doc/bitstream-filters.texi: add documentation for the aac_adtstoasc filter
[23:31] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:db7ebab5c360: doc/bitstream_filters: document the chomp filter
[23:39] <cone-855> ffmpeg.git 03Stefano Sabatini 07master:80b56a7bdd0f: examples/muxing: rename audio/video_pts to audio/video_time
[00:00] --- Thu Jun 27 2013


More information about the Ffmpeg-devel-irc mailing list