[Ffmpeg-devel-irc] ffmpeg-devel.log.20130416

[burek]

[00:03] <Lil-|^Red> Hi there, I downloaded a creative commons mp4 H264-Mpeg-4 (part10) (avc1) with Audio AAC Audio (mp4a) and would like to put this onto a viewable PS3 DVD using VLC or some other program. Is there a website/forum/or Guide to make this process slightly easier (as I only attempt this every 2 yrs- & usu. fail) that you are aware of and can tell me ??
[00:44] <cone-986> ffmpeg.git 03Martin Storsjö 07master:c0dcf89887f7: bfin: Match DEFUN_END macros to the right functions
[00:44] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:398b9ed28f58: Merge commit 'c0dcf89887f7fbdd7a8502bc7c715f50157562cb'
[00:49] <cone-986> ffmpeg.git 03Ronald S. Bultje 07master:c46819f2299c: x86: Move constants to the only place where they are used
[00:49] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:d5c31403aa13: Merge commit 'c46819f2299c73cd1bfa8ef04d08b0153a5699d3'
[00:55] <cone-986> ffmpeg.git 03Martin Storsjö 07master:5941978e71d2: vp3: bfin: Transpose the IDCTs
[00:55] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:05bea443935a: Merge commit '5941978e71d2c3a8e2a7e87951e081e0b2e77da9'
[01:01] <cone-986> ffmpeg.git 03Ronald S. Bultje 07master:015821229f96: vp3: Use full transpose for all IDCTs
[01:01] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:d98c3adef2ef: Merge commit '015821229f96bf7e677f2a711a58dbea3009f574'
[01:08] <cone-986> ffmpeg.git 03Martin Storsjö 07master:6add6272dac7: fate: Set the bitexact flag for vp3/vp5/vp6 tests
[01:08] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:7ed691e2db78: Merge remote-tracking branch 'qatar/master'
[04:02] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:7e20f80d2cd4: fate: fix histogram test, try #2
[05:10] <DooMMasteR> is there a way to use the parametric stereo feature of libaacplus via ffmpeg?
[05:11] <DooMMasteR> and why is this mode not supported? libaacplus: bad aac setting: br:32000, AACch:1, AACsr:24000 
[05:11] <DooMMasteR> using anything else aside from -ac 2 -ar 44100 will break
[05:11] <DooMMasteR> ahh not anuthing, mono also works
[05:11] <DooMMasteR> but parametric stereo is never used -.-
[06:09] <cone-986> ffmpeg.git 03Michael Niedermayer 07master:003bf337477f: sws: fix alignment of rgb2yuv
[10:47] <cone-507> ffmpeg.git 03Paul B Mahol 07master:97b2865aa27c: libaacplus: remove FF_API_OLD_ENCODE_AUDIO cruft
[11:43] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:2fa11a94a312: sws/fill_rgb2yuv_table: avoid 2 divisions
[11:43] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:bcf643486f46: sws/fill_rgb2yuv_table: increase precission to 32bit
[11:43] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:f05cfb654fe9: sws/fill_rgb2yuv_table: use ROUNDED_DIV
[12:20] <cone-507> ffmpeg.git 03Paul B Mahol 07master:82082695de3b: libaacplus: set supported channel_layouts
[12:20] <cone-507> ffmpeg.git 03Paul B Mahol 07master:0b06bd1c7a1d: libaacplus: set supported profiles
[13:25] <durandal_1707> michaelni: some 0RGB/0BGR is missing alpha in pixdesc.c
[13:26] <nevcairiel> isnt the point of the 0RGB things that they dont have alpha?
[13:27] <durandal_1707> yes but without that plane they are same as rgb24
[13:27] <durandal_1707> s/plane/comp stuff in pixdesc ...
[13:32] <michaelni> durandal_1707, which function uses the 4th plane when nb_components is 3
[13:32] <michaelni> ?
[13:34] <durandal_1707> michaelni: afaik, it should not matter, i just noticed that some 0rgb have it but others do not
[14:28] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:1314f534e54d: avutil/pixdesc: fix rgb0 descriptors
[14:28] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:20e99a9c10cd: avutil/pixdesc: support for self-checking the descriptors
[15:23] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:f807d6d2009b: avutil/pixdesc: check step/depth against each other
[15:23] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:2d8a3cf4c168: avutil/pixdesc: check for component overlap
[15:25] <durandal_1707> how you can draw pts with drawtext?
[15:26] <ubitux> %{pts} ?
[15:27] <durandal_1707> [Parsed_drawtext_0 @ 0x29fb90c0] Stray % near 'pts}'=
[15:27] <ubitux> 'drawtext=text=%{pts}:fontcolor=yellow:fontsize=30:x=30:y=30'
[15:27] <ubitux> works fine here ^
[15:27] <durandal_1707> yea, pebkac
[15:44] <durandal_1707> ubitux: why you did not added support for rgba/bgra/etc to curves filter?
[15:45] <ubitux> because i was lazy
[16:16] <durandal_1707> michaelni: i get this: [swscaler @ 0x29c20000] 20bpp not supported by yuv2rgb
[16:22] <ubitux> durandal_1707: so you want to add pixfmt to curves?
[16:23] <durandal_1707> maybe
[16:25] <durandal_1707> so what ivtc filters are for removal?
[16:26] <ubitux> yes
[16:27] <ubitux> want to drop them?
[16:27] <ubitux> there is not much remaining filters :)
[16:27] <ubitux> pullup (really?), eq2, ow, spp ?
[16:28] <durandal_1707> list what to drop, not what to keep....
[16:28] <ubitux> mmh
[16:28] <ubitux> just a sec then
[16:28] <durandal_1707> ubitux: pullup: does it works and is it faster than native filter?
[16:29] <ubitux> i don't know
[16:29] <ubitux> mp={detc,divtc,filmdint,pullup,ivtc}
[16:29] <ubitux> this is what i noted last time
[16:29] <ubitux> to be dropped
[16:35] <cone-507> ffmpeg.git 03Paul B Mahol 07master:65efa1dbd935: lavfi/interlace: remove now unused av_opt_free() call
[16:41] <durandal_1707> dint filter?
[16:42] <durandal_1707> ubitux: you forgot mcdeint?
[16:45] <ubitux> mcdeint is doing deinterlacing iirc
[16:45] <ubitux> not inverse telecine
[16:45] <durandal_1707> and is it good?
[16:47] <iive> it is slow and uses wavelets (snow)
[16:48] <durandal_1707> i tried it, and its not that slow but is full of artifacts
[16:49] <iive> it doesn't do deinterlace on its own, you need to insert yadif before it.
[16:51] <durandal_1707> hmm it encodes video ...
[17:11] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:e701cd96c2d5: avutil/avutil_version: Run checks just once
[17:11] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:6b657ac78897: avcodec/c93: Check for block overlap.
[17:22] <durandal_1707> michaelni: read your commit log message at least twice
[17:37] <michaelni> durandal_1707, i always do
[17:38] <michaelni> i read them again, now still sound good
[17:44] <michaelni> durandal_1707, you meant "checls" ?
[17:44] <michaelni> seems i missed that ...
[17:57] <ubitux> ffplay -f lavfi 'amovie=http\\://lucy.pkh.me/samples/spectrum.mp3, asplit [out0], showspectrum=size=640x256:slide=1:scale=cbrt [out1]'
[17:57] <ubitux> thanks to divVerent ^
[18:00] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:13bbfb8fda21: sws: Skip ff_yuv2rgb_c_init_tables() when the output is YUV
[18:14] <cone-507> ffmpeg.git 03Paul B Mahol 07master:657553917691: libaacplus: support for float sample format
[18:23] <BBB-work> j-b: ping
[18:27] <funman> dibs on pinging him
[18:51] <j-b> BBB-work: pong
[20:51] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:bec402b502df: mjpegdec: fix overlapping memcpy with upscale_v
[21:22] <maruthi> hi all.. am trying to build ffmpeg with stereo3d filter
[21:23] <maruthi> i enabled it during configure, but it is not listed in enabled filters
[21:24] <ubitux> maruthi: please ask #ffmpeg
[21:24] <ubitux> i'm going to help you there
[21:24] <ubitux> actually i already have the answer to your problem, but i want you to ask there anyway
[21:25] <maruthi> ok...moving to #ffmpeg
[21:28] <cehoyos> ubitux: Hi, the description for the fieldmatch filter is "fieldmatch       |->V       Field matching for inverse telecine." - shouldn't this be V->V ?
[21:28] <ubitux> the number of inputs is variable
[21:28] <ubitux> that listing should use the DYNAMIC flags recently introduced
[21:28] <ubitux> to make a difference between sources and variable inputs
[21:29] <ubitux> variable *number of* inputs
[21:32] <ubitux> cehoyos: http://pastie.org/7610609
[21:32] <ubitux> this should do the trick
[21:33] <ubitux> mmh not exactly, just a sec
[21:36] <ubitux> cehoyos: http://pastie.org/7610796 ’ http://b.pkh.me/filters
[22:17] <saste> can someone review my select patches?
[22:18] <ubitux> "add support to branch option" "factorize options definition between select and aselect" and "make select_frame() return void"?
[22:18] <ubitux> (what order?)
[22:19] <cone-507> ffmpeg.git 03Stefano Sabatini 07master:ba32afd70eaa: lavfi/split: use AVFILTER_DEFINE_CLASS to define class
[22:19] <cone-507> ffmpeg.git 03Stefano Sabatini 07master:30926a1f1ff5: lavfi/aspect: fix weird copyright notice
[22:19] <cone-507> ffmpeg.git 03Stefano Sabatini 07master:dc11acdf2d19: doc/filters: fix sentence in setsar/setdar docs
[22:20] <saste> ubitux: 1. lavfi/select: make select_frame() return void
[22:20] <saste> 2. lavfi/select: add support for dynamic number of outputs
[22:20] <saste> 3. lavfi/select: factorize options definition between select and aselect
[22:21] <ubitux> ok
[22:47] <durandal11707> michaelni: why probably ok for av_log change?
[22:57] <durandal11707> take2: what mp=dint does, is it usefull?
[22:57] <durandal11707> it "drops" interlaced frames
[23:02] <ubitux> saste: anything else to review?
[23:03] <ubitux> durandal11707: mmmh
[23:03] <ubitux> durandal11707: -vf idet,select could reproduce the same i guess?
[23:03] <ubitux> but idet might not be as effective as dint
[23:09] <cehoyos> ubitux: I find "N->V" difficult to understand, what's wrong with "V->V" (even if it does not show exactly the possibilities of the filter)?
[23:10] <ubitux> cehoyos: it can't be guess
[23:10] <ubitux> dynamic i/o = number AND type are unknown
[23:11] <cehoyos> durandal11707: I believe something went wrong when you committed 8208269 - please fix;-)
[23:11] <durandal11707> cehoyos: really? what?
[23:11] <cehoyos> ubitux: I don't understand: Does filedmatch support somthing else than video?
[23:11] <ubitux> durandal11707: you define channel_layouts[] but don't use it
[23:12] <ubitux> cehoyos: no but it support 1+ input
[23:12] <durandal11707> cehoyos: lol
[23:12] <ubitux> cehoyos: the number is undefined, and thus the types become unknown as well
[23:13] <ubitux> cehoyos: ffmpeg -h filter=fieldmatch
[23:13] <ubitux>     Inputs:
[23:13] <ubitux>         dynamic (depending on the options)
[23:13] <cehoyos> I understand now - it doesn't look optimal
[23:13] <ubitux> we could add various other random flags
[23:14] <ubitux> with ± use-cases
[23:14] <cehoyos> ± ?
[23:14] <ubitux> but that may be overkill, and often forgotten
[23:14] <ubitux> ± useful use-cases
[23:14] <cehoyos> I certainly is overkill, I just wanted to point out that I thought it does not look optimal currently.
[23:15] <durandal11707> you can make it bold
[23:15] <ubitux> :)
[23:15] <durandal11707> or additional flags like codecs/decoders/encoders
[23:16] <ubitux> wat
[23:16] <durandal11707> or icons
[23:16] <ubitux> gif
[23:16] <ubitux> but we need to fix our encoder
[23:16] <ubitux> and no one wants to do it :(
[23:17] <ubitux> mmh
[23:18] <ubitux> i'm tempted to improve it now that i have little time
[23:18] <cone-507> ffmpeg.git 03Paul B Mahol 07master:b05d8d4cf4af: lavc: avcodec_open2(): pass context to av_log()
[23:18] <cone-507> ffmpeg.git 03Paul B Mahol 07master:9429408cc189: sgirledec: remove unused headers
[23:18] <cone-507> ffmpeg.git 03Paul B Mahol 07master:83253431b716: lavfi/stereo3d: remove pointless store
[23:18] <cone-507> ffmpeg.git 03Paul B Mahol 07master:c1fe41ed472e: libaacpluc: remove unused code
[23:18] <J_Darnley> I once considered lokking at it but then I realised I didn't know the first thing about reducing to 256 colours
[23:19] <ubitux> J_Darnley: if you want to do it, i can still stop, right now
[23:19] <durandal11707> anything is better than current rgb8 solution
[23:23] <J_Darnley> ubitux: Not particularly.  I would either need to "borrow" code to make the palette with or do some serious reading
[23:24] <durandal11707> there is already code (elbg iirc)
[23:25] <durandal11707> the thing is palettization is out of scope of proper gif encoder, so you should not concentrate on it that much
[23:26] <J_Darnley> What?
[23:26] <ubitux> i'm not trying to fix the palette thing first
[23:26] <J_Darnley> You want it there so you can only make a new palette when it is worthwhile
[23:27] <durandal11707> aren't there good png->git converters?
[23:27] <ubitux> there are indeed some stuff to do with the colors too
[23:27] <durandal11707> even with source code and sane license?
[23:27] <ubitux> maybe i should start with that.
[23:27] <ubitux> but the color convert should be done properly in swscale
[23:28] <ubitux> actually, it might even be able to do that properly&
[23:28] <ubitux> anyway, gonna dig that a little
[23:29] <ubitux> ./ffmpeg -i tests/lena.pnm -vf format=rgb8,format=rgb24 -y lena.gif && feh lena.gif
[23:29] <ubitux> vs:
[23:29] <ubitux> ./ffmpeg -i tests/lena.pnm -y lena.gif && feh lena.gif
[23:29] <ubitux> i think requesting rgb8 only might actually fix the main problems
[23:30] <ubitux> and internals + swscale will do the necessary
[23:30] <durandal11707> rgb8 only is ugly and sucks
[23:30] <durandal11707> also you need to make gif muxer not to be encoder ...
[23:30] <ubitux> rgb8 makes a way better palette
[23:31] <ubitux> it can be improved in swscale, but it's mostly fine
[23:31] <cone-507> ffmpeg.git 03Rafaël Carré 07master:5b27c307e753: flvenc: do not mux more than one stream per type
[23:31] <cone-507> ffmpeg.git 03Anton Khirnov 07master:7f1fd9763668: cmdutils: Fix build with lavfi disabled
[23:31] <ubitux> (libcaca doc talks about that topic fairly well)
[23:31] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:0ea97a1c5b45: Merge remote-tracking branch 'qatar/master'
[23:31] <ubitux> anyway, i'll dig that and send some cool patches later
[23:31] <durandal11707> rgb8 is rgb24 with 8 bits - its not real pal8 that gif can actually make use of
[23:32] <ubitux> yeah right pal8 sorry
[23:36] <michaelni> ffmpeg -i tests/lena.pnm -y -vf scale=flags=0x806002,format=rgb8 lena.bmp
[23:37] <durandal11707> rm -rf tests/lena.pnm
[23:37] <durandal11707> use real images
[23:37] <ubitux> michaelni: looks nice
[23:37] <ubitux> what is this magic flag?
[23:37] <durandal11707> secret
[23:38] <ubitux> oh well
[23:38] <ubitux> format=pal8 gives the same :)
[23:38] <durandal11707> its just dithering that sucks less
[23:39] <durandal11707> instead dithering should be avoided at all cost....
[23:40] <michaelni> dd if=/dev/zero of=no_dither_at_all_cost
[23:40] <ubitux> the scale flags indeed make things better
[23:42] <durandal11707> ffmpeg -f lavfi -i smptehdbars -y -vf scale=flags=0x806002,format=rgb8 dither.bmp
[23:42] <michaelni> flags=error_diffusion+bicubic... or biliner+...    can also be used instead of the hex numbers
[23:42] <ubitux>  ./ffplay tests/lena.pnm -vf 'split[a][b]; [a]format=pal8,pad=iw*2[x]; [b]scale=flags=0x806002,format=pal8[y]; [x][y]overlay=w'
[23:43] <ubitux> the difference is interesting
[23:44] <ubitux> durandal11707: hehe
[23:44] <ubitux> durandal11707: try with the video
[23:44] <ubitux> it actually... changes (wtf)
[23:44] <ubitux> (./ffplay -f lavfi -i smptehdbars -vf scale=flags=0x806002,format=rgb8)
[23:45] <ubitux> oh well, dithering, random..
[23:45] <cehoyos> I tested compilation with PGC (Portland compiler) and apart from the fact that it does not pass fate, it fails during compilation:
[23:45] <ubitux> it's fun how the middle changes more often than the borders
[23:47] <cehoyos> It does not like line 232 in libavcodec/srtenc.c: ".colour            = srt_color_cb"
[23:47] <saste> durandal11707, what algo are you using in colorbalance?
[23:47] <cehoyos> Sorry, it fails with ".color            = srt_color_cb", after renaming to colour it worked.
[23:47] <ubitux> cehoyos: english/american nazi compiler?
[23:48] <cehoyos> And it fails with "anonymous structs" (if that is correct): libavcodec/pngenc.c line 61 and 
[23:48] <durandal11707> saste: the one in gimp
[23:48] <cehoyos> libavformat/flvdec.c line 747 (both from michaelni I suspect)
[23:48] <ubitux> easy to change
[23:49] <cehoyos> Should we try to fix compilation or is it "broken compiler)?
[23:49] <saste> durandal11707, link to code?
[23:49] <ubitux> cehoyos: i'm fine with replacing those (int[]){...} with a static const int ... array
[23:49] <durandal11707> saste: why?
[23:49] <ubitux> cehoyos: but that color/colour nazism sounds weird
[23:49] <cehoyos> ubitux: I wanted a comment to srtenc from you;-)
[23:49] <ubitux> cehoyos: i don't understand the problem
[23:50] <ubitux> the field is named "color", not "colour"
[23:50] <saste> durandal11707, so I know what the code does (I can't figure out from the code)
[23:50] <cehoyos> ubitux: Yes, I had to rename the other occurences as well;-)
[23:50] <cehoyos> (The others do not fail compilation)
[23:50] <ubitux> cehoyos: the compiler seriously doesn't like "color"?
[23:51] <cone-507> ffmpeg.git 03Paul B Mahol 07master:9bd1b1c1e756: lavfi/earwax: remove pointless store
[23:51] <cone-507> ffmpeg.git 03Paul B Mahol 07master:295ce83e2f06: lavfi/il: remove pointless store
[23:51] <ubitux> what is it complaining about in particular?
[23:51] <cehoyos> I am not sure I understand the question....
[23:51] <cehoyos> PGC-S-0060-color is not a member of this struct or union (libavcodec/srtenc.c: 240)
[23:51] <ubitux> cehoyos: i thought you said "colour" was ok but not "color"
[23:52] <cehoyos> Yes colour (and probably supercalifragelisticexpialigetic) is ok, color fails with above error message
[23:52] <ubitux> any idea why?
[23:53] <cehoyos> Actually, I don't care - the question is if we workaround a problem with a compiler that costs money and does not pass fate
[23:53] <cehoyos> (and does not support inlined asm properly, only 1024 characters)
[23:53] <cehoyos> or if we declare the compiler as "unsupported"
[23:53] <ubitux> i actually care, it might hides a real bug
[23:53] <durandal11707> how much costs?
[23:54] <cehoyos> (configure claims we support it)
[23:54] <cehoyos> Students: https://www.pgroup.com/pricing/aews.htm
[23:59] <michaelni> i think we should try to support it if its easy, it could uncover bugs in our code
[00:00] --- Wed Apr 17 2013