[Ffmpeg-devel-irc] ffmpeg-devel.log.20130417

[burek]

[00:00] <michaelni> also we should setup a fate instance once it (mostly) passes fate
[00:00] <michaelni> a list of all warnings and errors that it generates when compiling might be interresting too
[00:02] <durandal11707> but it requries s/color/colour etc....
[00:05] <saste> what's the best way to select a single channel component to output?
[00:05] <saste> for example if I want to convert alpha to a gray image?
[00:05] <durandal11707> alphaextract
[00:07] <cehoyos> michaelni: Patch sent, I wanted to leave the renaming to the subtitles maintainer.
[00:07] <saste> durandal11707, and if i want to select another component?
[00:08] <michaelni> in "void (*color)(void *priv, unsigned int color, unsigned int color_id);" <-- does it also fail if that int color is just int ?
[00:08] <durandal11707> saste: use geq to move channel to alpa and than alhaextract?
[00:09] <saste> durandal11707, yes, but that's not very efficient
[00:09] <saste> indeed i wonder if we should generalize alphaextract
[00:09] <saste> something like rgbextract
[00:11] <cone-507> ffmpeg.git 03Christophe Gisquet 07master:1a4007964c10: x86: float dsp: butterflies_float SSE
[00:11] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:92218aad007f: butterflies_float: replace 2 lea by 2 add
[00:13] <cehoyos> =-)
[00:13] <cehoyos> Removing the second "color" from the line in ass_split.h fixes the compilation problem.
[00:14] <durandal11707> saste: https://git.gnome.org/browse/gimp/tree/app/base/color-balance.c?id=1f7cf2dd6b3d4a93a9c01a66da0320f8f05673b1
[00:23] <ubitux> cehoyos: nit: static const enum AVMediaType
[00:24] <ubitux> cehoyos: so a clash between the function name and the argument?
[00:24] <cehoyos> ubitux: Yes, apparently
[00:25] <ubitux> lol
[00:25] <cehoyos> Could you suggest a commit message?
[00:25] <ubitux> cehoyos: -unsigned int color +unsigned int /* color */
[00:26] <ubitux> and commit message saying what compiler is stupid, and what version eventually
[00:27] <ubitux> michaelni: you should stop using "que" for "queue"; this word doesn't exist :p
[00:27] <ubitux> i'm not even sure it's valid slang
[00:28] <cehoyos> ubitux: My question was: Is the construct I remove really called "anonymous struct" or is that something else?
[00:30] <ubitux> afaict you're just removing an argument from a function pointer in a struct
[00:30] <ubitux> argument *name*
[00:30] <cehoyos> The other patch...
[00:30] <ubitux> seems i brainfart a lot tonight
[00:31] <ubitux> it's an anonymous array
[00:32] <ubitux> could be considered a struct somehow, but i'm not sure of the vocabulary
[00:32] <cehoyos> Of course, thank you!
[00:33] <michaelni> ubitux, "slang, language peculiar to a particular group ...", can we say que is used really alot by the ffmpeg developers, then it would qualify as slang ?
[00:33] <ubitux> you're the only one to use it :p
[00:34] <michaelni> thats not my fault or is it ?
[00:34] <durandal11707> saste: i really do not like [-2, 2], it should remain 200....
[00:36] <saste> durandal11707, why?
[00:38] <durandal11707> i need to type . all the time
[00:39] <cehoyos> michaelni: pgc compilation fixes are on github ready to merge
[00:41] <ubitux> michaelni: i mean you don't represent a group by yourself :)
[00:45] <michaelni> ubitux, please see: https://en.wikipedia.org/wiki/Trivial_group, it describes all properties of the trivial group with one element that is me
[00:47] <durandal11707> axiom: michaelni does not represent group by himself
[00:49] <ubitux> michaelni :(
[00:50] Action: michaelni goes back to working on ffmpeg
[00:54] <cehoyos> Nobody wants to comment on this mail? (Stefano?) http://thread.gmane.org/gmane.comp.video.ffmpeg.libav.user/10532
[00:54] <cone-507> ffmpeg.git 03Carl Eugen Hoyos 07master:9802f56684fe: ass_split: Do not use the function name as a parameter name in a declaration.
[00:54] <cone-507> ffmpeg.git 03Carl Eugen Hoyos 07master:fe1de12fafce: Remove two anonymous arrays.
[00:57] <durandal11707> cehoyos: your xv does not support rgb?
[01:01] <saste> cehoyos, we should pay someone to write a tutorial
[01:02] <saste> and we should add more examples covering more real-world use cases
[01:02] <saste> usual stuff about crowd funding etc. etc.
[01:03] <durandal11707> saste: aren't you in ffmtech?
[01:03] <saste> durandal11707, no
[01:03] <saste> I left more than one year ago
[01:05] <ubitux> wtf, the gif encoder is... never used?
[01:05] Action: ubitux wonders wtf is going on
[01:06] <ubitux> something looks definitely wrong
[01:06] <cehoyos> durandal11707: Yours does?
[01:06] <ubitux> both encoder and muxer are doing somehow the same with data
[01:06] <ubitux> oh god the format takes rawvideo
[01:06] <cehoyos> ubitux: There is a muxer that uses the rawvideo encoder and an encoder used by image2
[01:06] <cehoyos> Or wasn't that the question?
[01:07] <durandal11707> cehoyos: is that affirmative/negative answer?
[01:07] <cehoyos> The muxer likes rgb24, the encoder pal8
[01:07] <ubitux> cehoyos: why doing it like this?
[01:07] <ubitux> cehoyos: why not make the muxer just write encoded gif data?
[01:07] <cehoyos> durandal11707: So far, I was sure that nobody's xv supported anything else (that is the reason I already asked the other day about your xv)
[01:07] <cehoyos> ubitux: historical reasons?
[01:08] <ubitux> cehoyos: ok
[01:08] Action: durandal11707 looks like ubitux forget glory days when gif decoder where written in demuxer
[01:08] <ubitux> ._.
[01:08] <cehoyos> durandal11707: What does xvinfo report for you?
[01:09] <cehoyos> (It also supports yuy2 and yuyv here, but the number of FFmpeg decoders is limited, so I don't think they are so useful)
[01:09] <durandal11707> Xvideo image format: 0x41424752 (RGBA) packed and so one from mpv....
[01:09] <cehoyos> That is what xvinfo reports?
[01:09] <durandal11707> i dont heave xvinfo installed
[01:09] <durandal11707> mplayer should give similar output with verbose flag
[01:13] <saste> AVDictionary **avpriv_frame_get_metadatap(AVFrame *frame);
[01:13] <saste> why it isn't const *frame?
[01:13] <cehoyos> durandal11707: ATI?
[01:14] <durandal11707> cehoyos: radeon
[01:15] <cehoyos> That seems to be the explanation: I am on nvidia, and google seems to confirm that ati supports rgb via xv.
[01:16] <cehoyos> Good night!
[01:21] <kierank> michaelni: how can you guarantee monotonic dts
[01:22] <kierank> might be possible to do what you want to do with additional latency to normal
[01:24] <kierank> i need to think about it a bit
[01:25] <kierank> i'm pretty sure there's a reason you can't do that
[01:34] <michaelni> kierank, theres code between the user app and the muxer that sorts packets per dts
[01:34] <michaelni> it can be disabled per muxer if the muxer doesnt want it
[01:34] <michaelni> or be replaced by some other rule
[01:35] <kierank> but what if i write a packet with a dts less than the packet the muxer has already worked on
[01:35] <kierank> i.e how does it know when to start leaking the buffer
[01:36] <michaelni> dts cant move backward in each stream
[01:36] <michaelni> so it just needs one packet in each stream
[01:37] <michaelni> the details are a bit more complex IIRC
[01:37] <michaelni> but thats the prnciple
[01:37] <michaelni> see ff_interleave_packet_per_dts() and surrounding code
[01:49] <kierank> might work if it can deliver an interleaved dts across all streams reliably
[02:01] <durandal11707> saste: need to send another iteration?
[02:01] <saste> durandal11707, yes please
[02:02] <durandal11707> saste: for both?
[02:03] <saste> yes, colorchannel is trivial, while I'd like to check better the colorbalance patch
[02:03] <saste> *colorchannelmixer
[02:05] <durandal11707> saste: for what?
[02:05] <saste> durandal11707, for the transform
[02:05] <durandal11707> its same
[02:06] <saste>   double mid = 2.0 / 3.0 * (1 - ((i - 127.0) / 127.0) * ((i - 127.0) / 127.0));
[02:06] <saste> also i suspect readability could be improved a bit
[02:08] <durandal11707> it is already readable
[02:15] <durandal11707> ffplay -f lavfi -i testsrc -vf histeq=0.9 segv here
[02:19] <cone-507> ffmpeg.git 03Michael Niedermayer 07master:161dee43213d: rmdec: dont return uninitialized data
[02:25] <cone-507> ffmpeg.git 03Paul B Mahol 07master:b8edf91657ad: lavfi/histeq: fix out of array write
[02:26] <durandal11707> histeq filter nees some serious love
[03:12] <ubitux> yay, got it working
[03:49] <ubitux>  3 files changed, 25 insertions(+), 200 deletions(-)
[03:49] <ubitux> and adding a feature.
[03:49] <ubitux> actually, fixing that mess mainly
[03:53] <ubitux> and size of files on a quick test divided by 3
[11:12] <cone-483> ffmpeg.git 03Clément BSsch 07master:9db6aaeb2c1e: ffplay: use format title metadata to set window caption.
[11:21] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:f2619cbd6148: examples/filtering_audio: enable ref counted frames
[11:21] <cone-483> ffmpeg.git 03Pavel Koshevoy 07master:e5644313c581: examples/filtering_audio: get rid of AVABufferSinkParams
[11:58] <cone-483> ffmpeg.git 03Clément BSsch 07master:a5ebd2d987c3: cmdutils: slightly better -filters output.
[12:08] <durandal_1707> last time to object on mp inverse telecine filters removal
[12:19] <durandal_1707> ubitux: adding direct variable is really not needed
[12:19] <ubitux> one mem access instead of two?
[12:20] <ubitux> it also gives some meaning to the code so i kind of like it
[12:21] <durandal_1707> i find it completly pointless ....
[12:59] <ubitux> when is the gif demuxer going to honor the loop flag?
[12:59] <ubitux> :(
[12:59] <durandal_1707> ?
[13:00] <ubitux> if a loop is set it should seek back to the start and resend img packets
[13:01] <durandal_1707> well that is trivial to add if you really like demuxer to loop forever
[13:01] <ubitux> for files with a loop i'd like it indeed
[13:02] <durandal_1707> then send patch
[13:02] <ubitux> will do
[13:02] <ubitux> but i want to improve the encoder a little first :)
[13:20] <mateo`> Hello guys, i plan to support PKL files from DCP packages in a dedicated demuxer, thus ffmpeg will be able to play a DCP without manually specify the MXF files. Do you think this is a good idea (and could be accepted) ?
[13:22] <durandal_1707> mateo`: why would it be 'bad' idea?
[13:23] <mateo`> I don't know, i just want to make sure that this work could be integrated in ffmpeg :)
[13:23] <durandal_1707> then how i would know if i do not know anything about PKL/DCP
[13:25] <durandal_1707> DCP are rar files?
[13:26] <cbsrobot> dcp is basicly a folder with xml and mxf files and the xml files are describing (with hash and size check) what the mxf are used for
[13:27] <durandal_1707> and pkl files in dcp are?
[13:27] <cbsrobot> pkl is a sort of xml files
[13:27] <cbsrobot> there are also cpl files
[13:27] <Compn> mateo` : is that something that needs some kind of playlist support ?
[13:28] <ubitux> i hope you can have .jar into it, otherwise it's disappointing
[13:28] <Compn> mateo` : if you have a small example file, i suggest uploading and posting to ffmpeg-devel so people can figure out how they want to tackle it
[13:28] <cbsrobot> I think the pkl is used to define all the assets
[13:28] <cbsrobot> the cpl is used to define what media files go togehter
[13:28] <mateo`> Compn: i don't know enough about the CPL/PKL format
[13:29] <cbsrobot> you can have multiple cpl files in a dcp folder
[13:29] <mateo`> but i just planned to support the PKL files, so no playlist support i guess
[13:30] <mateo`> afaik it just references the mxf files to use
[13:38] <Compn> mateo` : i think someone added support for chained ogg files. so that sounds kinda similar. might want to copy how that was done
[13:39] <Compn> i think its just a demuxer addition
[13:40] <mateo`> my idea was to copy the hls demuxer, parse the description file, open demuxers ...
[13:46] <durandal_1707> ubitux: why ffmpeg -f filter=format list 0 as inputs and outputs?
[13:47] <ubitux> 0 is the pad number
[13:47] <ubitux> input[0] called "default"
[13:47] <ubitux> output[0] called "default"
[13:50] Action: durandal_1707 gonna remove all filters
[13:51] <ubitux> no :(
[13:54] <durandal_1707> "convert stereo movie from top-bottom to left-right field" <- what that means?
[13:57] <durandal_1707> hmm gonna implement this in stereo3d and remove this filter too
[13:59] <durandal_1707> this filter is bad, it dumps half of pixels
[13:59] <durandal_1707> and same funcionality is already in stereo3d
[14:00] <durandal_1707> except its broken, probably because of ubitux
[14:00] <ubitux> :(
[14:02] Action: durandal_1707 finished removing filters, now _pushing_ ...
[14:02] <cone-483> ffmpeg.git 03Paul B Mahol 07master:f0a149e53827: lavfi: remove bad inverse telecine filters
[14:06] <durandal_1707> i removed 3k lines of code
[14:06] <ubitux> :)
[14:13] <ubitux> i need to keep a reference to the previous frame in an encoder
[14:13] <ubitux> but the frame is passed as const
[14:14] <ubitux> so it seems i can't use av_frame_clone, ref, or similar
[14:14] <ubitux> (since the source is affected)
[14:14] <ubitux> how am i supposed to do that?
[14:16] <durandal_1707> ubitux: you return cloned frame
[14:16] <durandal_1707> and use reget_buffer
[14:16] <durandal_1707> thats how many codecs doo
[14:16] <durandal_1707> lol this is for decoders ...
[14:17] <ubitux> yes for decoders there is plenty of examples
[14:17] <ubitux> but for encoders..
[14:17] <durandal_1707> return copied frame
[14:18] <ubitux> how do you copy the frame?
[14:18] <ubitux> av_frame_clone wants a non-const for the source
[14:18] <durandal_1707> data stuff
[14:19] <ubitux> erk.
[14:19] <ubitux> that sucks a bit..
[14:22] <cone-483> ffmpeg.git 03Paul B Mahol 07master:b725202546da: lavfi/stereo3d: fix output width for sbsl/sbs2l
[14:23] <ubitux> afaict it means duplicating the buffers (so manual alloc for each plane)
[14:23] <ubitux> sounds a bit braindead to me
[14:26] <durandal_1707> ubitux: but how would you make sure that right data is muxed?
[14:27] <ubitux> what do you mean?
[14:29] <durandal_1707> use av_frame_clone with cast
[14:30] <ubitux> yeah i guess.
[14:43] <ubitux> hehe size/2 for an encode of testsrc
[14:43] <ubitux> more to come.
[14:49] <durandal_1707> what you added?
[14:49] <ubitux> top and bottom cropping
[14:51] <durandal_1707> how you detect it?
[14:51] <ubitux> memcmp with the previous frame :p
[15:05] <ubitux> -f lavfi testsrc,vflip
[15:05] <ubitux> crash.
[15:07] <durandal_1707> not here
[15:09] <ubitux> works fine when dropping get_video_buffer from vflip
[15:10] <durandal_1707> but does it still vflips?
[15:10] <ubitux> sure.
[15:10] <ubitux> it's done in filter_frame
[15:10] <ubitux> there is no reason for this get_video_buffer insanity
[15:10] <ubitux> unless i'm missing something?
[15:12] <ubitux> negative linesize, yum yum
[15:13] <cone-483> ffmpeg.git 03Anton Khirnov 07master:aafed1175df7: af_channelmap: sanity check input channel indices in all cases.
[15:13] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:74e86d3147cc: Merge commit 'aafed1175df76603e94c99a7748968780d6548d2'
[15:24] <cone-483> ffmpeg.git 03Anton Khirnov 07master:c0771a1ac6da: af_channelmap: cosmetics, use an extra local variable to shorten the code
[15:24] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:f60ba6c52418: Merge commit 'c0771a1ac6da697f86e3b10c8fe5dbc2ee92e347'
[15:42] <cone-483> ffmpeg.git 03Anton Khirnov 07master:b1bb8fb860b4: svq1dec: check that the reference frame has the same dimensions as the current one
[15:42] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:251f283b9d1b: Merge commit 'b1bb8fb860b47e90dd67f0c5740698128fc82dcc'
[15:57] <cone-483> ffmpeg.git 03Anton Khirnov 07master:ecff5acb5a73: svq1dec: clip motion vectors to the frame size.
[15:57] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:0a946599f881: Merge commit 'ecff5acb5a738fcb4f9e206a12070dac4bf259b3'
[16:10] <cone-483> ffmpeg.git 03Anton Khirnov 07master:34f87a58532e: qdm2: check that the FFT size is a power of 2
[16:10] <cone-483> ffmpeg.git 03Anton Khirnov 07master:69f6f03cd16a: rv10: cosmetics, reformat
[16:10] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:8cff853b2336: Merge commit '69f6f03cd16a09e4e848b4e82bae4993bc361cd0'
[16:15] <ubitux> here we go, cropping done
[16:20] <cone-483> ffmpeg.git 03Anton Khirnov 07master:bac8d38c0a71: rv10: return meaningful error codes.
[16:20] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:4916e220a60a: Merge commit 'bac8d38c0a716c8bdfdecf0ae545015c68f5df36'
[16:24] <ubitux> durandal_1707: any trick in mind to increase compression a bit more?
[16:25] <ubitux> playing with palettes could help, but except with the palette stuff maybe there is some other stuff
[16:26] <durandal_1707> ubitux: disposals
[16:27] <durandal_1707> also there are converters...
[16:27] <ubitux> yeah i could have a look to them but well
[16:28] <ubitux> less fun :)
[16:38] <Snaggle> !seen highgod
[16:49] <cone-483> ffmpeg.git 03Anton Khirnov 07master:01d376f598fe: rv10: check that extradata is large enough
[16:49] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:e5dc5095f9fa: Merge commit '01d376f598fe95478036f5d1e3e5e14ffe32d4bf'
[16:56] <cone-483> ffmpeg.git 03Anton Khirnov 07master:66531d634e75: indeo3: switch parsing the header to bytestream2
[16:56] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:4029a5ebc2f9: Merge commit '66531d634e75b834e89e4a6a0f7470ca018712a1'
[17:03] <cone-483> ffmpeg.git 03Anton Khirnov 07master:34e6af9e204c: indeo3: fix data size check
[17:03] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:2787f7b188f5: Merge commit '34e6af9e204ca6bb18d8cf8ec68fe19b0e083e95'
[17:19] <cone-483> ffmpeg.git 03Paul B Mahol 07master:716588188d87: lavfi/stereo3d: check input width & height
[17:21] <durandal_1707> Snaggle: you can also contact him by email...
[17:22] <Snaggle> durandal_1707: I did. No response yet.
[17:35] <cone-483> ffmpeg.git 03Anton Khirnov 07master:a0a872d0733f: indeo3: check motion vectors.
[17:36] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:7f2253078d3e: Merge remote-tracking branch 'qatar/master'
[18:00] <durandal_1707> what it best way to write filter that reads 2 frames(from single input) and outputs one frame?
[18:08] <ubitux> just like decimate?
[18:09] <ubitux> but since there is only one frame to keep, a pointer to the prev frame in the context is enough
[19:02] <durandal_1707> ubitux: what is this nonsense macro in decimate filter?
[19:04] <ubitux> which one?
[19:05] <durandal_1707> PF() magic
[19:05] <durandal_1707> its very funny
[19:05] <ubitux> oh well the list was too painful to expand manually
[19:05] <ubitux> is there a problem with it?
[19:06] <durandal_1707> except grep fail, no
[19:28] <cone-483> ffmpeg.git 03Clément BSsch 07master:3db3b278f2dc: lavfi/lut: add direct path.
[19:28] <cone-483> ffmpeg.git 03Clément BSsch 07master:45f5bf917b54: lavfi/lut: reindent after previous commit.
[19:28] <cone-483> ffmpeg.git 03Clément BSsch 07master:4c6fa4ef45b9: lavfi/lut: simplify nested component stepping.
[19:28] <cone-483> ffmpeg.git 03Clément BSsch 07master:7ee2daed9b73: lavfi/interlace: remove exec permissions.
[19:29] <ubitux> anyone for the vflip crash?
[19:30] <durandal_1707> ubitux: i cant reproduce, does it happen with lavfi only?
[19:31] <saste> ubitux, i'm debugging it
[19:31] <saste> is it a regression?
[19:31] <ubitux> yes
[19:31] <ubitux> durandal_1707: i think so
[19:31] <saste> ubitux, since when?
[19:32] <ubitux> no idea, it works with 1.2
[19:32] <ubitux> so most likely one of the recent libav bomb
[19:32] <saste> weird
[19:32] <ubitux> (i can reproduce with the fork)
[19:32] <saste> ubitux, otoh, how can you use -f lavfi with the fork?
[19:33] <ubitux> -filter_complex testsrc,fvlip
[19:33] <saste> and seems a filter code bug to me, like that it is assuming a positive linesize
[19:33] <ubitux> vflip*
[19:33] <ubitux> gtg, bbl
[19:34] <saste> ubitux, cyl
[19:34] <saste> gty
[19:34] <cone-483> ffmpeg.git 03Paul B Mahol 07master:1de7dcb4574d: lavfi/stereo3d: fix interleave row output modes
[19:36] <durandal_1707> i gonna remove down3dright
[21:32] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:45741dd81f41: vf_lut: fix simplification / off by 1 error
[21:32] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:cc6f848dba89: avutil/intfloat_readwrite: avoid comparission with INFINITY, use isinf()
[21:32] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:3220083c112f: ra144: Try to fix int16/uint16 warnings from pgc
[21:35] <ubitux> oups
[21:37] <saste> ubitux, the testsrc bug is weird
[21:37] <saste> it should affect libav as well, and older versions
[21:37] <cone-483> ffmpeg.git 03Nicolas George 07master:e3e6aa7afb82: lavfi/buffersrc: fix failed requests count.
[21:37] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:0b3bd78714ab: Merge remote-tracking branch 'cigaes/master'
[21:38] <JEEB> talking about weird, I love these Japanese dual mono airings, that have a stereo AAC track that is actually supposed to be two separate mono outputs that the user can switch between
[21:38] <saste> so, who's going to continue with xvideo?
[21:38] <JEEB> I should check one day in the ARIB specs how it is actually signaled
[21:45] <ubitux> saste: it affects libav
[21:45] <saste> ubitux, and older versions?
[21:45] <saste> the bug is there since the first commit
[21:45] <ubitux> i wasn't able to reproduce with ffmpeg 1.2
[21:46] <ubitux> maybe a lucky run
[21:46] <saste> but i'm pretty sure i tested that combination before
[21:46] <saste> it should happen *always* if I understand the problem
[21:46] <saste> now debugging mptestsrc,vflip
[21:48] <saste>  memset(picref->data[0], 0,   picref->linesize[0] * h);
[21:48] <saste> nice
[21:51] <durandal_1707> aww
[21:54] <durandal_1707> there is bunch of such casese in other filters
[21:55] <durandal_1707> showwaves,showspectrum
[22:00] <durandal_1707> ubitux: someone have problems with fieldmatch
[22:00] <ubitux> that's not a surprise
[22:01] <ubitux> where?
[22:01] <durandal_1707> on ffmpeg-user
[22:02] <saste> durandal_1707, you going to pick the xvideo patch?
[22:02] <durandal_1707> saste: should I?
[22:02] <saste> seems already in good shape, assuming it works
[22:02] <saste> durandal_1707, only if you want
[22:03] <durandal_1707> i'm already full of patches floating around
[22:04] <saste> i wonder if there is a better way to iterate over the lines to memset them to 0
[22:06] <durandal_1707> use FFABS
[22:08] <durandal11707> saste: i gonna push colorchannelmixer unless you object
[22:08] <saste> durandal11707, let me review
[22:08] <saste> i'm gonna do soon
[22:08] <saste> blame ubitux for finding new bugs all the time
[22:18] <durandal11707> saste: review what?, i already improved filter by adding alpha and 16bit support
[22:24] <ubitux> saste: i'm also adding some ± discreetly so you never get bored
[22:24] <ubitux> so vflip is the new filter breaker?
[22:26] <durandal11707> its not breaker, you should never ever assume linesize is positive
[22:27] <ubitux> i believe we assume it quite a lot
[22:27] <ubitux> though, i still don't get that get buffer thing
[22:29] <durandal11707> i asked saste for explanation....
[22:29] <saste> ubitux, did you test your patch with fate?
[22:30] <saste> i don't think it can work
[22:31] <saste> it is a bit tricky, try to imagine what happen with a padded buffer and your approach
[22:36] <ubitux> saste: i believe i did
[22:37] <ubitux> saste: oh, with pad mmmh 
[22:52] <durandal11707> saste: you gonna improve histeq?
[22:52] <saste> durandal11707, what's the problem with it?
[22:53] <saste> i really have no much time, feel free to create a ticket so i don't discard
[22:54] <saste> how much time does it take make fate on your machines?
[22:54] <saste> here it is more than 20 minutes, but then my laptop has serious heating design issues
[23:00] <durandal11707> saste: buy new one?
[23:01] <saste> when the weather is warmer is much worse
[23:02] <saste> in winter is better, probably i should move to northern europe (or to the other emisphere)
[23:03] <nevcairiel> michaelni: cc6f848dba89 broke msvc dll builds
[23:03] <nevcairiel> hm possibly static builds too, those didnt run yet
[23:05] <nevcairiel> let that damn pgc compiler warn, but dont break msvc :P
[23:05] <durandal11707> saste: preferably somewhere without internet connection?.... :evil
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:565c50ac7b51: lavfi/select: make select_frame() return void
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:dcc1b3236168: lavfi/select: add support for dynamic number of outputs
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:8bfd8781519b: lavfi/select: factorize options definition between select and aselect
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:fe5ca6b91371: lavfi/aspect: restore ratio parsing
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:5229b7837164: lavfi/aspect: extend verbose feedback, show old SAR and DAR values
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:7ab631261033: lavfi/testsrc: use int in place of unsigned
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:3933963d7ba0: lavfi/testsrc: fix style
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:bbf003e94970: lavfi/mptestsrc: reindent option constants
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:db84088e99f5: lavfi/mptestsrc: use AV_OPT_TYPE_DURATION
[23:07] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:e252460c9be7: lavfi/mptestsrc: remove unused size field
[23:08] <ubitux> saste: any new comment on the timeline btw?
[23:08] <saste> ubitux, did you send a new patch?
[23:09] <ubitux> i don't remember
[23:09] <ubitux> i remember answering your question though
[23:09] <ubitux> but my memory might be in a quantum state
[23:11] <ubitux> i guess i'll have to convince nicolas
[23:19] <ubitux> saste: you forgot a #if FF_API_OLD_FILTER_OPTS around a .init
[23:19] <ubitux> (you did it for dar only)
[23:21] <michaelni> nevcairiel, i dont really care about pgcs warnings i did these chnages in hopes to fix pgc build and pgc idependant bugs. ill fix msvc asap
[23:22] <michaelni> s/pgc build/pgc fate/
[23:24] <cone-483> ffmpeg.git 03Michael Niedermayer 07master:c25224737cba: avutil/intfloat_readwrite: include common.h for isinf()
[23:27] <michaelni> saste, you broke  -vf setdar=16:9
[23:27] <michaelni> example: .(ffmpeg -i matrixbench_mpeg2.mpg -vf setdar=16:9   tmp.avi
[23:27] <michaelni> ./ffmpeg
[23:31] <durandal11707> hmm, it looks like wrong stuff is applied....
[23:40] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:f212c2380a11: lavfi/setsar: remove spurious #if FF_API_OLD_FILTER_OPTS around init field
[23:40] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:f04ec05f0ed0: lavfi/setdar: fix num/den swapping in log message
[23:40] <cone-483> ffmpeg.git 03Stefano Sabatini 07master:1efcab02b6c7: lavfi/aspect: apply hot fix for NUM:DEN syntax
[23:44] <saste> durandal11707, i don't like arbitrary rescaling factors which complicates understanding for questionable gain
[23:44] <saste> "mathematically cleaner" in this case it means a simpler expression of the output value
[23:45] <durandal11707> i don't need suck kind of explanations
[23:47] <saste> durandal11707, explanation ? no it's rather a "justification"
[23:50] <durandal11707> saste: your "justification" for colorbalance is same, note that scaling is hiden there
[00:00] --- Thu Apr 18 2013