[Ffmpeg-devel-irc] ffmpeg-devel.log.20121227

burek burek021 at gmail.com
Fri Dec 28 02:05:02 CET 2012


[01:00] <saste> ubitux: AV_BASE64_SIZE() is used to compute the encoded data size
[01:01] <ubitux> oh
[01:02] <ubitux> i guess i'm a bit tired :))
[01:13] <llogan> saste: users might assume the usable range is 0-63, instead of 0-10, unless i am wrong about that.
[01:16] <llogan> why does the example not use -q:v/-qscale:v for simplicity/consistency/familiarity instead of -flags +qscale -global_quality X?
[01:16] <saste> llogan, because i don't want to use specific ffmpeg options
[01:17] <saste> in case the user is using the library, rather than the tool
[01:18] <llogan> ah, ok, but i think another example showing -q:v usage might clarify any confusion
[01:18] <saste> llogan, yes, patch welcome
[01:18] <saste> or even better we should add private options, but that's not as simple at the moment
[01:19] <saste> what's the difference between q and qscale?
[01:19] <saste> i always forget
[01:20] <llogan> q is an alias for qscale, IIRC
[01:20] <saste> yes
[04:26] <highgod> Hi all,I want to ask a question about ffmpeg video filters.If I use multi filters such as the command "ffmpeg -i ./testfile/xxx.m2ts -vf yadif,deshake./testfile/output.mkv",can deshake filer get the filter name before it? Or yadif can get the filter name after it?Thanks
[04:47] <michaelni> highgod, yes filters could inspect their connected filters and get their names
[06:00] <highgod> thanks michaelni.Can you say more details?Can I find any code in ffmpeg of this use?Should I inspect the structure AVFilterLink? Thanks
[07:08] <ubitux> in a demuxer, can i control the codec options?
[07:08] <ubitux> it seems av_opt_set(st->codec, ...) isn't able to find the option :(
[07:43] <an3k> i know you can in a muxer and splitter but tbh i dont see any good reason to overwrite the codec information if you want to demux
[10:30] <Compn> ubitux : this page (and similar pages) are useless to me http://ffmpeg.org/libavfilter.html
[10:30] <Compn> why are they there ?
[10:35] <ubitux> they can be useful to understand the purpose of the library
[10:35] <ubitux> some are more complete
[10:36] <ubitux> it's just an introduction/gate to the related documentation
[10:48] <Compn> ....
[10:49] <Compn> theres already a intro/gate to the docs, its the doc page http://ffmpeg.org/documentation.html
[11:29] <burek> is it useful to create a feature request for ffserver, which would suggest that ffserver implements something like "buffering from the last keyframe"
[11:30] <burek> which means that ffserver will buffer the stream in the feed file in such a way that the beginning of the buffer is always a keyframe
[11:30] <burek> so that every new client, requesting the stream, will be given the data buffered which can be played back immediately
[11:40] <michaelni> sounds usefull, i would prefer a patch or pull req over a fteature req though
[11:43] <burek> ok :)
[12:32] <cone-261> ffmpeg.git 03Rob Sykes 07master:f2957adb23a4: soxr_resample: Correct email address
[12:32] <cone-261> ffmpeg.git 03Rob Sykes 07master:03d38ee20752: soxr_resample: soxr API parameter rename
[12:32] <cone-261> ffmpeg.git 03Rob Sykes 07master:8d9a50331322: docs: add soxr documentation
[12:32] <cone-261> ffmpeg.git 03Michael Niedermayer 07master:b3f3fe3c28fb: MAINTAINERS: add libswresample
[13:45] <ubitux> i need to change the layout of the packet for a demuxer (and so the code in the decoder as well)
[13:45] <ubitux> what's the best way to deal with that situation?
[13:58] <michaelni> ubitux, is something using that codec or demuxer ? if so there may be API/ABI issues
[13:59] <michaelni> also could the packets have been stored in a container? if so there may be more compatibility issues
[14:01] <ubitux> i don't know, it's the microdvd demuxer/decoder
[14:02] <ubitux> and yes, i know it technically creates some api/abi issues
[14:02] <ubitux> my question is how i'm suppose to make such change
[14:04] <michaelni> on the decoder side you can just support both variants
[14:04] <michaelni> on the demuxer side a change means it wont work with decoders expecting the old format
[14:05] <ubitux> supporting both variants on the decoder side -> how can i know which format it is?
[14:05] <michaelni> do both decode correctly ?
[14:05] <ubitux> libavcodec major+minor version?
[14:07] <ubitux> michaelni: basically microdvd demuxer currently outputs "timing" (actually frames id) into the packets (it's the initial "{123}{456}" string at the beginning of the line)
[14:07] <michaelni> in general the decoder looks in the data, codec_tag, extradata and or anything else it has to figure out what variant it has if there are multiple
[14:07] <ubitux> (it's the old way of doing that stuff)
[14:07] <ubitux> currently the decoder skip them
[14:08] <ubitux> now i want to drop them from the packet, because i actually found have another format that could potentially output microdvd packets (microdvd markup)
[14:08] <ubitux> oh the codec_tag
[14:08] <ubitux> sounds like a potential workaround..
[14:08] <michaelni> detecting from packet.data is better than codec_tag if possible
[14:09] <ubitux> that's a bit delicate
[14:09] <michaelni> because codec_tags may get lost in converting around
[14:09] <ubitux> because just after the {123}{456} you can have some kind of markup store as... "{foobar}"
[14:09] <ubitux> so this kind of probing might get ugly
[14:10] <ubitux> not impossible though
[15:43] <burek> if anyone was working on ass filter lately, this might be a bug or something: http://ffmpeg.gusari.org/viewtopic.php?f=16&t=773&p=1375
[15:50] <Compn> ehe
[15:51] <cone-261> ffmpeg.git 03Stefano Sabatini 07master:750382d5c883: ffmpeg: remove annoying logs and improve reporting in case of unavailable bitrate
[15:51] <cone-261> ffmpeg.git 03Stefano Sabatini 07master:7e0676a7155a: doc/ffmpeg-formats: extend documentation for the probesize parameter
[15:51] <cone-261> ffmpeg.git 03Stefano Sabatini 07master:ae2b51acd2ac: doc/encoders: clarify the direction of the quality setting parameters for libtheora
[16:40] <burek> man.. i really had hard time these few days with ffmpeg and vlc.. i finally found out exact versions of ffmpeg and vlc and alsa and even kernel which works stable enough...
[16:42] <burek> the worst thing is that issue shows after streaming the video for some time (couple of hours maybe).. so it's not easily reproducible at all..
[16:42] <burek> and i can't even file a bug report on this :?
[17:46] <cone-261> ffmpeg.git 03Michael Niedermayer 07master:b61097fe0dc2: mjpegdec: reset restart_count on progressive scans
[19:04] <durandal_1707> why i cant specify number in hexdecimal?
[19:07] <J_Darnley> Specify where?
[19:11] <durandal_1707> for options
[19:15] <cone-261> ffmpeg.git 03Don Moir 07master:64f4fb75c4ce: gifdec: fix transparent background color
[19:26] <cone-261> ffmpeg.git 03Paul B Mahol 07master:10a3fa8f3181: build: fix g723_1 decoder dependency
[20:11] <burek> can someone advise something on this issue http://ffmpeg.gusari.org/viewtopic.php?f=12&t=762
[20:12] <burek> the guy wants to lower the latency when playing rtsp stream from a webcam
[20:12] <burek> to put in on the robot and remotely control/view it
[20:19] <Compn> have him force the rtsp format
[20:19] <Compn> hes losing latency in probe i'm guessing
[20:20] <Compn> force the demuxer ,i mean
[20:21] <Compn> tell him also to try a -ss 1
[20:22] <Compn> maybe it will skip some old data
[20:22] <Compn> who knows
[20:22] <Compn> maybe its playing at wrong fps
[20:22] <Compn> try specifying faster fps in ffplay 
[20:26] <burek> um, i think i'll just copy paste your suggestions :)
[20:26] <burek> thanks btw :)
[20:27] <michaelni> durandal_1707, ffmpeg ... -b 0xABCD ... works here
[20:50] <ubitux> wow& in cea-608/708 cc we can put& URLs!
[21:29] <tbarletz> I'm getting SIGILL when running ffprobe on a mips system. I'm using './configure --enable-cross-compile --arch=mipsel --disable-ffserver --disable-ffplay --target-os=linux --cross-prefix=mipsel-linux-'. Any ideas?
[21:35] <michaelni> see/compare to http://fate.ffmpeg.org/, there are some MIPS systems being tested there, configure options are listed there too
[21:35] <durandal_1707> valgrind, backtrace...
[21:35] <michaelni> yes that too
[21:45] <tbarletz> Thanks
[22:05] <tbarletz> It's crashing in sws_init_context, but when I --disable-swscale, I crash somewhere else
[22:10] <tbarletz> Does anybody use the Android binaries that fate.ffmpeg.org is building?
[22:11] <durandal_1707> it is used by fate and it does not crash otherwise it would be red and not green/yellow
[22:12] <tbarletz> For Android it appears to only build, not run the tests
[22:12] <tbarletz> I can build it fine too
[22:13] <durandal_1707> bad indeed
[22:14] <durandal_1707> it is probably disabled because it crash
[22:42] <ubitux> wtf i have an infinite loop with -nostats
[22:44] <ubitux> oh, get it.
[23:25] <cone-261> ffmpeg.git 03Michael Niedermayer 07master:69db85c30c91: mjpegdec: Support 16bit DQT
[00:00] --- Fri Dec 28 2012


More information about the Ffmpeg-devel-irc mailing list