[Ffmpeg-devel-irc] ffmpeg-devel.log.20130123

[burek]

[00:07] <saste> or.. i could extend if/ifnot to take an additional param...
[00:28] <Darkarnium> win 2
[00:29] <Darkarnium> Whoops
[04:54] <highgod> Hi,I want to ask a question. I get the mpeg2 video format(4:2:0 or 4:2:2) in the extra data. Is there any condition that the mpeg2 doesn't has the extra data? If under this condition, how can I get the video format? Thanks all.
[05:13] <michaelni> highgod, the extradata is set from the parser or avformat, a user application might use avcodec without these or they might both fail to set extradata
[05:13] <michaelni> in that case it would be in some frame (and data before that frame could not be decoded)
[05:20] <highgod> Thanks, michaelni, I submit the dxva2 patch, and I want to get the video format before decode start. If the video is not 4:2:0, I will stop the dxva2 process. I have get the format from extra data reference the mpeg2video decoder that read the extra data. in AVCodecContext. But the latest hendrik reply said that I should check the mpeg2video without the extra data.I don't know how to do 
[05:20] <highgod> it. Thanks 
[05:29] <wm4> I don't know too much about video formats, but I always thought you can't know anything before inspecting the first video packet
[05:29] <wm4> and even then, later packets may reconfigure the stream completely
[05:29] <cone-106> ffmpeg.git 03Michael Niedermayer 07master:e49c2aab80cb: truemotion2: clear the token array if its initialization fails.
[05:29] <cone-106> ffmpeg.git 03Michael Niedermayer 07master:713dea584b11: truemotion2: use av_mallocz()
[05:40] <highgod> wm4:I inspect the extra data.I just wang to get the format.Is there anyway to do this process?Thanks.
[05:42] <Darkarnium> Hey all, just a quick question regarding command line arguments to demuxers / muxers. It looks like it's all handled via the options 'field' of an AVClass for the (de)muxer. Is that correct?
[05:49] <cone-106> ffmpeg.git 03Xi Wang 07master:2f014567cfd6: rtpenc: fix overflow checking in avc_mp4_find_startcode()
[05:49] <cone-106> ffmpeg.git 03Xi Wang 07master:902cfe2f74d7: rtmp: fix multiple broken overflow checks
[05:49] <cone-106> ffmpeg.git 03Xi Wang 07master:4e692374f796: rtmp: fix buffer overflows in ff_amf_tag_contents()
[10:54] <durandal_1707> uh, why LJPG in riff have two different codecs?
[12:23] <cone-16> ffmpeg.git 03Paul B Mahol 07master:92d874cb787d: flacenc: include correct header
[12:23] <cone-16> ffmpeg.git 03Paul B Mahol 07master:6f465c6219b6: dcaenc: remove unused header
[12:24] <cone-16> ffmpeg.git 03Paul B Mahol 07master:be38aaba4a8e: ffv1enc: include correct header
[12:24] <cone-16> ffmpeg.git 03Paul B Mahol 07master:a2dd988bf83b: jpeglsenc: add missing put_bits.h header
[12:24] <cone-16> ffmpeg.git 03Paul B Mahol 07master:205b2d2a9822: adcpmenc: remove unused header
[12:39] <durandal_1707> lol why ffplay still does not support -af, wasnt there some patch for this?
[13:07] <Compn> durandal_1707 : iirc theres jpegls and ljpeg
[13:08] <Compn> different lossless jpeg formats
[13:08] <Compn> but i dont know if its correct in riff, i didnt look
[13:59] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:55aa03b9f8f1: floatdsp: move vector_fmul_add from dsputil to avfloatdsp.
[13:59] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:b1b870fbd718: Merge commit '55aa03b9f8f11ebb7535424cc0e5635558590f49'
[14:11] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:42d324694883: floatdsp: move vector_fmul_reverse from dsputil to avfloatdsp.
[14:11] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:6e6e1708984e: Merge commit '42d324694883cdf1fff1612ac70fa403692a1ad4'
[14:22] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:5959bfaca396: floatdsp: move butterflies_float from dsputil to avfloatdsp.
[14:22] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:24604ebaf85b: Merge commit '5959bfaca396ecaf63a8123055f499688b79cae3'
[14:36] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:d56668bd8007: floatdsp: move scalarproduct_float from dsputil to avfloatdsp.
[14:36] <cone-16> ffmpeg.git 03Diego Biurrun 07master:73b704ac609d: arm: Add some missing header #includes
[14:36] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:8102f27b5b3d: Merge commit '73b704ac609d83e0be124589f24efd9b94947cf9'
[15:59] <cone-16> ffmpeg.git 03Tim Walker 07master:99ccd2ba10ea: mlp: store the channel layout for each substream.
[15:59] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:fa36270c4c68: Merge commit '99ccd2ba10eac2b282c272ad9e75f082123c765a'
[16:11] <cone-16> ffmpeg.git 03Tim Walker 07master:3ffcccb4fbaa: mlpdec: TrueHD: use Libav channel order.
[16:11] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:cd6a8618b1f5: Merge commit '3ffcccb4fbaae4d5ad775506f1f2761f2029affa'
[16:34] <cone-16> ffmpeg.git 03Tim Walker 07master:1fd2deedcc64: mlpdec: set the channel layout.
[16:34] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:1cf6f6f3daa9: Merge commit '1fd2deedcc6400e08b31566a547a5fac3b38cefb'
[16:41] <cone-16> ffmpeg.git 03Tim Walker 07master:b0c7e3ffd09f: mlp_parser: account for AVCodecContext.request_channels when setting the channel layout.
[16:41] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:86b892af199d: Merge commit 'b0c7e3ffd09f68022429920da483ad7abed84aa1'
[16:49] <cone-16> ffmpeg.git 03Tim Walker 07master:ed1b01131e66: mlp: implement support for AVCodecContext.request_channel_layout.
[16:49] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:826188d27acd: Merge commit 'ed1b01131e662c9086b27aaaea69684d8575fbea'
[16:57] <cone-16> ffmpeg.git 03Tim Walker 07master:40cf1add12fa: mlp_parser: cosmetics: re-indent.
[16:57] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:0ee8293a77a6: vp3dsp: don't do aligned reads on input.
[16:57] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:0ff21c316534: mlpdec: remove disabled code, leftover from merge
[16:57] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:516f30ed2082: Merge commit '0ee8293a77a6afad161e91ce1d43c4a57ce33a6a'
[17:04] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:2e4bb99f4df7: vorbisdsp: convert x86 simd functions from inline asm to yasm.
[17:04] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:b90ab2b99365: Merge commit '2e4bb99f4df7052b3e147ee898fcb4013a34d904'
[17:15] <cone-16> ffmpeg.git 03Paul B Mahol 07master:a793a587df05: lavfi/swapuv: support all planar yuv pixel formats
[17:25] <antonello> hallo , I am doing an audio transcode , Can help me ?
[17:28] <durandal_1707> antonello: join #ffmpeg and ask there
[17:29] <antonello> ok txt
[17:36] <durandal_1707> michaelni: sox have bunch of biquad filters, i want to port to lavfi
[17:37] <durandal_1707> question is should i put filters in separate files? , probably not
[17:38] <durandal_1707> all filters actually use same "biquad function" with differen parameters...
[17:39] <michaelni> if they differ just in parameters then same file 
[17:39] <michaelni> unless the code stays very similar (then multiple could make sense for easier updating)
[17:40] <durandal_1707> well i do not want to use structure in function as that makes is slower
[17:48] <cone-16> ffmpeg.git 03Diego Biurrun 07master:88bd7fdc821a: Drop DCTELEM typedef
[17:48] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:ac8987591f55: Merge commit '88bd7fdc821aaa0cbcf44cf075c62aaa42121e3f'
[17:57] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:32ff6432284f: dsputil: remove avg_no_rnd_pixels8.
[17:57] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:6d1e9d993a55: Merge commit '32ff6432284f713e9f837ee5b36fc8e9f1902836'
[18:01] <durandal_1707> does somebody know important reasons of current dsputil split surge?
[18:02] <wm4> apparently it's a big clusterfuck
[18:03] <durandal_1707> so only cosmetics?
[18:04] <durandal_1707> yes it is really big clusterfuck....
[18:05] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:2bef1a83d7a3: dsputil: remove some never-assigned function pointers from the struct.
[18:05] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:70270f27538f: Merge commit '2bef1a83d7a3ca60b161ab90bf4d4338e5def5e7'
[18:08] <wm4> durandal_1707: I think they want to decouple decoders from what they perceive as big interlocked clusterfuck
[18:08] <wm4> and then delete the clusterfuck
[18:09] <wm4> but I don't know if it's more than just a cleanup
[18:09] <durandal_1707> it should provide smaller binary if you disable most of code
[18:11] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:2612c4dc78e1: dsputil: remove 9/10 bits hpel functions.
[18:11] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:f900fd5d31dc: Merge commit '2612c4dc78e1750c4653bf5a9f2cbe95f7b4ed6e'
[18:17] <saste> michaelni, about the image2 updatefirst option, I cannot understand what's useful for
[18:22] <michaelni> saste, if you have a webserver and serve a image from it, with update first a ffmpeg running on the server could update that image 
[18:22] <michaelni> IIRC
[18:25] <saste> michaelni, i don't think i understand the code, or the code in img2enc.c is wrong
[18:25] <saste> why should av_get_frame_filename() fail?
[18:25] <cone-16> ffmpeg.git 03Ronald S. Bultje 07master:baf35bb4bc4f: dsputil: remove one array dimension from avg_no_rnd_pixels_tab.
[18:25] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:1e7a92f219a8: Merge commit 'baf35bb4bc4fe7a2a4113c50989d11dd9ef81e76'
[18:25] <saste> also in that case we have no guarantee that "filename" contains a valid string
[18:27] <michaelni> it fails because theres no %d in it
[18:27] <michaelni> when you update always the same file, you dont need/want a %d
[18:28] <saste> michaelni, i see
[18:28] <saste> thanks
[18:29] <cone-16> ffmpeg.git 03Martin Storsjö 07master:2e4130037cf3: float_dsp: Include config.h for redefining restrict
[18:29] <cone-16> ffmpeg.git 03Xi Wang 07master:3cff53369acd: rtmp: fix multiple broken overflow checks
[18:29] <cone-16> ffmpeg.git 03Xi Wang 07master:ecb918e5f0a4: rtmp: fix buffer overflows in ff_amf_tag_contents()
[18:29] <cone-16> ffmpeg.git 03Xi Wang 07master:cf29f49d8ae0: rtpenc: fix overflow checking in avc_mp4_find_startcode()
[18:30] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:8b8ac78ef3a1: Merge remote-tracking branch 'qatar/master'
[18:33] <cone-16> ffmpeg.git 03Stefano Sabatini 07master:ca68acd7ac37: lavf/segment: mark #EXT-X-ENDLIST in M3U8 only at the end of stream
[18:33] <cone-16> ffmpeg.git 03Stefano Sabatini 07master:f4e4e5b84162: lavf/segment: drop now pointless segment_list_close() function
[18:56] <cone-16> ffmpeg.git 03Stefano Sabatini 07master:bb539eb13ab0: lavf/segment: avoid failure in case -flags +live and -segment_times/frames are specified
[18:56] <cone-16> ffmpeg.git 03Stefano Sabatini 07master:e3197257171c: lavf/segment: exit immediately in case of invalid stream specifier
[19:23] <durandal_1707> michaelni: aselect filter that select samples based on time interval is exactly what i'm looking for, i just want that timestamps are keept intact
[19:30] <cone-16> ffmpeg.git 03Micah Galizia 07master:d4e5d8d4c264: lavf/hls: broker HTTP options
[19:35] <durandal_1707> no wonder nobody cares for lavfi when even creators do not understand it
[19:35] <saste> durandal_1707, what's the matter?
[19:36] <wm4> lol
[19:36] <durandal_1707> i want to silence portion of audio
[19:36] <durandal_1707> using aselect and volume for example
[19:36] <durandal_1707> other parts of audio should be intact
[19:37] <saste> durandal_1707, in short, you can't
[19:37] <Compn> cant mute ?
[19:37] <saste> durandal_1707, adding support to dynamic expression in volume should be easy
[19:37] <durandal_1707> so some combination of aselect,avolume,abuffer,asplit
[19:37] <Compn> like mplayer's edit decision list ? :P
[19:37] <saste> then you can script stuff .e.g. with sendcmd (or a complicated expression)
[19:37] <Compn> " edit decision list "
[19:37] <durandal_1707> saste: adding extra option for every filter is boring task
[19:38] <wm4> Compn: that's quite a special case
[19:38] <Compn> you want to be able to insert and remove filters at will
[19:38] <durandal_1707> you mean i wate 30 minutes to send cmd?
[19:38] <Compn> wm4 : what? edl ? what about filtering a logo that only appears on screen for 5 minutes every 15 minutes ? 
[19:39] <durandal_1707> adding extra option in each filter that will parse command is overkill
[19:39] <wm4> well, edl allows you to mute - but AFAIK that's a hardcoded special case, and you can't do anything else in edls
[19:39] <Compn> you can skip video in edl 
[19:39] <saste> durandal_1707, it's not that hard if you get used
[19:39] <Compn> seek that is
[19:39] <Compn> seek and mute
[19:39] <durandal_1707> saste: it is wrong....
[19:40] <Compn> a very simple edl system. but other edl applications have more options. this is just mplayer limitation....
[19:40] <wm4> I think you need to provide a stable and simple interface to lavfi
[19:40] <wm4> so at least _users_ can use it
[19:40] <durandal_1707> it is not hard to clusterfuck on big pile of hack ..
[19:41] <saste> durandal_1707, you add more options, or you add more filters
[19:41] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:6747b0be9b2a: mlp: fix channel order.
[19:41] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:8071288601cc: mlpdec: move ch_assign setup code down to after ch layout is set.
[19:41] <Compn> there was another use case for turning filters on at random, but i've forgotten it now.
[19:41] <wm4> Compn: well, in theory you could design an EDL system that allows execution arbitrary slave mode commands, and that's as far as mplayer could get, probably
[19:41] <saste> sometimes it is easier to add options, or filters, depending on various constraints (efficiency, ease of use, etc.)
[19:41] <wm4> personally, I think a lua shell for lavfi would solve this
[19:42] <wm4> you could do things like slicing audio with Lua functions, and explicitly filter them, then concat the result
[19:42] <saste> durandal_1707, for example i completely forget that i wanted to implement audio fading through expression evaluation
[19:42] <durandal_1707> saste: so can you explain what new filter would do?
[19:42] <saste> (using volume)
[19:42] <durandal_1707> saste: good, because it is bad idea
[19:42] <michaelni> the correct solution is to have a filter that can enable or disable another
[19:42] <michaelni> if thats what is wanted
[19:43] <saste> michaelni, how?
[19:43] <Compn> -af spigot:volume
[19:43] <wm4> sounds like a clusterfuck solution
[19:43] <Compn> lol
[19:43] <wm4> I mean, filter disabling other filters?
[19:43] <saste> wm4, can you help with implementing a lavfi lua scripting filter?
[19:43] <durandal_1707> there is asplit- it clones 
[19:44] <durandal_1707> there is aselect - it select frames/samples
[19:44] <Compn> saste : would you commit lua shell if he did ?
[19:44] <wm4> saste: actually I'm somewhat interested, but I have other things to do, and it'd probably break in a month due to API changes
[19:44] <Compn> wm4 is here to troll, not to develop ;P
[19:44] <wm4> Compn is here to troll, not to develop ;P
[19:45] <Compn> nice comeback , your mom!
[19:45] <michaelni> like -vf switch(volume=0,ts>5,ts<9)
[19:45] <wm4> Compn: no u
[19:45] <gnafu> "So's your face."
[19:45] <michaelni> but the whole lavfi parser is flawed, it cant even parse anything but flat syntax
[19:45] <saste> durandal_1707, asplit[in1][in2]; [in1]aselect=EPXR[out1];[in2]aselect=!EXPR[out2]; [out1][out2]ainterleave
[19:46] <durandal_1707> i dont want ainterleave i want acat
[19:46] <michaelni> saste, good example why its not useable
[19:46] <michaelni> i mean the syntax
[19:46] <saste> durandal_1707, or you extend *select to send the samples to another "else" output
[19:47] <saste> michaelni, about flat syntax -> not true
[19:47] <saste> it's a mess because escaping, but nothing i can't do about
[19:47] <michaelni> yes
[19:47] <saste> we can still design an alternative language (for file scripting)
[19:47] <michaelni> its flat or not useable
[19:47] <michaelni> yes
[19:47] <michaelni> we should
[19:47] <michaelni> also doesnt need to be just for files
[19:47] <wm4> no, you should not design another language
[19:47] <saste> about the switch thing: we (almost) already have it, it's the sendcmd
[19:48] <michaelni> wm4, language is a bad term here
[19:48] <wm4> the commands stuff is even worse, because it allows only flat strings, and unless with graph parsing, everyone is absolutely forced to use it
[19:48] <saste> alternatively: we provide scripting, and the user builds the filtergraph using the wrapping language
[19:49] <durandal_1707> atime=30,40,60,80[a1][a2][a3][a4][a5],][a1]afade[out0],....
[19:49] <saste> all solutions imply someone doing work, of course
[19:49] <michaelni> i just want filter(foo(bar=5)) instead of filter=\\\foo\\\\\:\\\\;\\\\\;;;\\\\\:bar\\\\\\=\\\\\\\5\\\\\whatever
[19:49] <wm4> anyway, people inventing their own powerful ad-hoc languages almost always ends in tears
[19:49] <michaelni> wm4, YES
[19:49] <michaelni> i want the insane syntax fixed not a new language
[19:50] <durandal_1707> so use C syntax
[19:50] <wm4> add clang as ffmpeg dependency!
[19:50] <durandal_1707> wm4: so what you propose?
[19:51] <wm4> durandal_1707: if you really want to allow complex filter graphs, use an external, existing, well-proven scripting language, and make lavfi available to it
[19:52] <durandal_1707> lua, now i'm not sure I'm right person for it...
[19:52] <durandal_1707> what would get exported?
[19:54] <wm4> I suppose you would use the scripting language to build a filter graph
[19:54] <wm4> so, just that
[19:56] Action: wm4 suggests looking at avisynth/vsynth
[19:57] <saste> durandal_1707, also now that i think at it, afade has the same limitations of fade, that's why i wrote (and never finished) multifade
[19:58] <michaelni> lua and scripting is nice and good,yet this doesnt solve the problem that a generic filter that disables/enables a filter that is given as its argument is a huge mess ATM escaping wise.
[20:00] <wm4> ah, I see, even when using av_opt, filters still get a flat argument string
[20:00] <wm4> so maybe export the options directly
[20:00] <saste> wm4, having a consistent av_opt interface for all filters is the first step in that direction
[20:01] <wm4> definitely
[20:05] <durandal_1707> saste: i already said my opinion on hacking
[20:07] <durandal_1707> multiselect,multivolume,multidelogo,multiyadif,multicolor,multilife and so on
[20:09] <durandal_1707> when integer dsputil stuff will be  moved to lavu?
[20:18] <durandal_1707> the earwax filter does different thing from sox effect
[20:20] <durandal_1707> and above scalarproduct comment in earwax is bogus because there is extra shift...
[20:21] <durandal_1707> actually function name is just wrong
[20:23] <saste> durandal_1707, earwax is broken as well, i filed a ticket
[20:23] <saste> patches are very welcome...
[20:26] <durandal_1707> saste: i'm aware of both statements very well and I work on stuff i understand, I also avoid clusterfuck code
[20:27] <durandal_1707> i just posted small cleanup patch for earwax
[20:31] <durandal_1707> now only me and michaelni are left :(
[20:32] <durandal_1707> lol, user on ml list posted long filter_complex example
[20:33] <wm4> durandal_1707: that's the direction it'll take
[20:33] <durandal_1707> wm4: what?
[20:34] <wm4> users coming up with more and more complex scripts
[20:35] <durandal_1707> ahh good, i was thinking you are leaving too...
[20:38] <durandal_1707> having useful scripting in lavfi is critical for its success
[20:41] <Compn> ehe you guys are piling up on saste :P
[20:41] <Compn> gotta help him create the scripting stuff
[20:41] <Compn> its a difficult thing to create an api by yourself
[20:42] <durandal_1707> nope it should be simple stupid
[20:47] <llogan> has any recieved any "ffmpeg-* has been disabled due to excessive bounces" messsages?
[20:47] <llogan> someone is complaining in ffmpeg-user and is "sure" that it isn't his fault.
[20:48] <llogan> i blamed him, but i wanted to double check
[20:57] <durandal_1707> llogan: i think i did not get first mail in thread
[20:58] <durandal_1707> well not that thread but some other .....
[20:59] <llogan> i was going to say if it is from "reindl" then you're just missing a troll
[21:03] <Compn> llogan : any reply from ubuntu about ffmpeg package ? or is it settled now ?
[21:03] <Compn> llogan : you saw siretart's explanation ?
[21:09] <llogan> Compn: yes. i saw his explanation. as i was unaware of upstream activity at the time i dropped my previous plan. beastd is working it it right now, and i have another plan if that doesn't go through.
[21:09] <llogan> *so i dropped...
[22:21] <Compn> llogan : ok , good.
[22:50] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:665f2d432ccd: hls: add missing checks for accessing avoption fields
[22:50] <cone-16> ffmpeg.git 03Michael Niedermayer 07master:31cd1e20bbd5: eval: add function to access the current "wallclock" time.
[23:54] <saste> michaelni: ping on eval docu patches
[00:00] --- Thu Jan 24 2013