[Ffmpeg-devel-irc] ffmpeg-devel.log.20150201

[burek]

[03:46] <jamrial> 158730 decicycles in sao_edge_filter_8, 32768 runs, 0 skips
[03:46] <jamrial> 6422 decicycles in ff_hevc_sao_edge_filter_32_8_ssse3, 32764 runs, 4 skips
[04:08] <cone-465> ffmpeg.git 03Christophe Gisquet 07master:cb335b3af202: dvenc: fix 2x4x8 (interlaced) weight table
[05:47] <Timothy_1u> jamrial: WTF gcc
[12:33] <cone-594> ffmpeg.git 03Paul B Mahol 07master:01a3c5dae0e0: avcodec/put_bits:  remove unneeded #include, there are no assert()
[12:33] <cone-594> ffmpeg.git 03Paul B Mahol 07master:021c0c5f5246: avcodec/motion_est: remove unneeded #include, there are no assert() only av_assert*
[13:34] <cone-594> ffmpeg.git 03Luca Barbato 07master:f726fc21ef76: ogg: Provide an option to offset the serial number
[13:34] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:9bacb576bcdf: Merge commit 'f726fc21ef76a8ba3445448066f7b2a687fbca16'
[13:57] <JEEB> ubitux, I guess those dithering algos could be of use with this kind of stuff as well https://twitter.com/marcan42/status/561736363259600896
[14:24] <wm4> weeaboos everywhere
[14:34] <cone-594> ffmpeg.git 03Rémi Denis-Courmont 07master:60392480181f: mem: fix pointer pointer aliasing violations
[14:34] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:fa73358c9bfd: Merge commit '60392480181f24ebf3ab48d8ac3614705de90152'
[14:56] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:e912f45b8165: avformat/oggenc: Simplify by using OFFSET and PARAM
[14:59] <dhead666_> hi, I'm trying to add to support to set the service type of mpegts, and was asked to use consts, what do you think about this? https://github.com/dhead666/FFmpeg/commit/048fb759cee70783d5e69f58df7d092a2e4e5b33 
[15:00] <dhead666_> p.s. I'm not sure how to add the list properly to doc/muxers.texi
[15:42] <ubitux> JEEB: maybe :)
[15:42] <ubitux> dhead666_: you can use an enum
[15:42] <ubitux> so that specs only define 3>
[15:42] <ubitux> ?
[15:43] <ubitux> "Set service_type field."  you might want a better description for each field
[15:43] <ubitux> the limits are not correct anymore
[15:44] <dhead666_> ubitux: no the specs defines too many, but I think these are the most useful, the rest can still be send by just using the hex value
[15:44] <ubitux> "too many"? how much exactly? 20, 30?
[15:45] <wm4> it's fine, ffmpeg produces broken and invalid files all the time anyway
[15:45] <dhead666_> this is my updated commit https://github.com/dhead666/FFmpeg/commit/1f56b68bf6d808eaabd798d5b2d7dd519909410f
[15:45] <ubitux> dhead666_: you can remove the "Set service_type field as " for values
[15:46] <ubitux> "digital_radio", "Digital Radio" is fine for these constants
[15:46] <ubitux> dhead666_: you can check output with ./ffmpeg -h muxer=mpegts
[15:48] <ubitux> dhead666_: if you want an example for the documentation, look for "The following values are recognized" into that doc file
[15:49] <dhead666_> ubitux: :)
[17:06] <dhead666_> ubitux: how about this? https://github.com/dhead666/FFmpeg/commit/f07eee96c076c1f13d081e91d49c00445200fbdc
[17:10] <ubitux> dhead666_: s/default 0x01, max 0xff/default @var{digital_tv}/
[17:10] <ubitux> no more comment from me
[17:12] <dhead666_> ubitux: what is this, sed notation ?
[17:13] <ubitux> yes
[17:13] <ubitux> replace "default 0x01, max 0xff" with "default @var{digital_tv}"
[17:15] <dhead666_> haha, I can't believe I knew that, linux truely affects me, I should have said that I'm not really a developer ;)
[17:15] <ubitux> sorry
[17:16] <dhead666_> ubitux: no biggy, thank you very much for all your help :)
[17:31] <cone-594> ffmpeg.git 03Carl Eugen Hoyos 07master:f018b1f3a0f1: Support YA16 in jpeg 2000.
[17:39] <durandal_1707> so ffmpeg was not present on this year fosdem?
[17:48] <cone-594> ffmpeg.git 03Paul B Mahol 07master:d0045312c345: avfilter/vf_dctdnoiz: Use av_malloc_array()
[18:03] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:0182e5a740ff: avcodec/nellymoserenc: replace assert by av_assert1()
[19:25] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:61aca123fb30: avcodec/h261dec: Fix exported MVs for skipped MBs
[19:25] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:0ff8fe10f2c3: avcodec/h261enc: Fix number suffix
[19:25] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:4b15bba2aec9: avformat/utils: Fix number suffixes in tb_unreliable()
[19:46] <cone-594> ffmpeg.git 03James Almer 07master:65e6ab0c5a9f: hevcdsp: remove unused parameter from sao_band_filter
[19:46] <cone-594> ffmpeg.git 03James Almer 07master:2929e5600613: hevcdsp: replace the SAOParams struct parameter from sao_band_filter
[20:12] <wm4> is there a way to stop libavformat doing "approximate" seeks in raw files, like estimating time by bitrate?
[20:13] <wm4> I'm seriously thinking about working around the shit seek API by doing my own incremental indexing on top of it and then using byte seeks, but that would suck
[20:13] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:c77cc2c17666: swresample/dither: Cleanup number suffixes
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:b4ad2853c50d: avcodec/dxtory: Use LL instead of L number suffix
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:fc3cdb00d084: avformat/matroskadec: Fix number suffixes
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:465f3705b1ef: avformat/smacker: Fix number suffix
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:f1f7f5903ab4: avformat/omadec: fix number suffix
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:592ba6ec1062: avcodec/h264_cabac: use int instead of long for mbb_xy
[20:14] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:4531e2c489d2: avcodec/mpegvideo_enc: Fix number suffixes in rc_buffer_size calculation
[20:23] <akira4> ubitux, could you have a look at this diff? http://pastebin.com/Liwdzp3K. 
[20:25] <ubitux> does it work?
[20:28] <ubitux> did you try hardcoding a style first? something like char **list[] = {"FontName=Times","FontSize=70",NULL} to see if at least this part behaves correctly?
[20:28] <ubitux> make a loop to print your list
[20:28] <ubitux> to make sure it's constructed correctly
[20:28] <ubitux> (assuming it doesn't work)
[20:28] <akira4> it works.
[20:28] <akira4> but the font color doesnt change
[20:29] <akira4> for eg if Itry ./ffmpeg -i test.avi -vf "subtitles=testing.srt:force_style='FontName=Arial,Underline=-1,Italic=-1,FontSize=29,PrimaryColor=255'"  out.avi
[20:29] <akira4> I don't see any change in the font color but the other things are implemented.
[20:30] <ubitux> what's 255?
[20:30] <ubitux> look at the ASS on how to specify a color
[20:30] <ubitux> it's something like &Hsomething
[20:30] <ubitux> iirc
[20:31] <ubitux> http://docs.aegisub.org/manual/ASS_Tags
[20:33] <akira4> I've tried it. Doesnt work.
[20:33] <ubitux> does it make a difference if it's not the last entry?
[20:34] <ubitux> what did you try exactly?
[20:34] <akira4> ./ffmpeg -i ~/test.avi -vf "subtitles=$HOME/testing.srt:force_style='PrimaryColor=&HFF000000'"  ~/out.avi
[20:34] <ubitux> ah btw, it's Colour, not Color
[20:34] <akira4> :'(
[20:34] <ubitux> ;)
[20:34] <akira4> my bad. Thanks
[20:35] <ubitux> can't you have only 1 call to av_strtok?
[20:36] <ubitux> same question for realloc (if you always have 1+x)
[20:37] <akira4> hmm. 
[20:37] <akira4> yeah. I'll try using one. Its better that way.
[20:37] <ubitux> btw, we have av_dynarray* functions
[20:38] <ubitux> which might simplify this
[20:38] <akira4> Oh. I'll check them out.The font color changes now :D
[20:38] <ubitux> :)
[20:39] <akira4> I'll try making the code more optimized then. Thanks :)
[22:45] <ubitux> we don't have any ktree implementation in ffmpeg, right?
[22:47] <cone-594> ffmpeg.git 03Michael Niedermayer 07master:36adfa32748f: tests/utils: Remove unneeded L suffix
[22:49] <ubitux> kdtree* sorry
[22:49] <ubitux> (for k=3 in particular... :P)
[23:34] <kurosu__> michaelni, jamrial: Fabrice Bellard's patch for sao buffers that I adapted: https://github.com/kurosu/ffmpeg/commit/717da8896deccf163a8428c6684b5f6c5b71479b
[23:34] <kurosu__> the only noteworthy change is adapting it to our sao dsp and abusing an already existing buffer
[23:38] <jamrial> regarding the "XXX" comment in sao_edge, that's already taken care of in the sao_edge asm patchset i'm going to send later
[23:38] <kurosu__> I could rewrite the commit message to actually explain what it does, seeing how it is already a hotchpotch
[23:38] <kurosu__> ok
[23:39] <kurosu__> I'll submit the above when the patchset you sent is applied, as it depends on it
[23:48] <jamrial> michaelni: is the patchet in the ml ok to apply?
[23:54] <kurosu__> good night
[23:56] <michaelni> jamrial, should be ok
[00:00] --- Mon Feb  2 2015