[Ffmpeg-devel-irc] ffmpeg-devel.log.20120630

[burek]

[00:50] <michaelni> durandal_1707, dsputils supports sereral blocksizes
[00:50] <michaelni> which depends on the function as well
[00:52] <durandal_1707> in sanm case pixel in blocks take 2 bytes
[00:53] <durandal_1707> so copying block of size 2x2 is actually copying block of size 2v x 4h
[01:06] <michaelni> in >8bit h264 pixels are also 2 byte
[01:08] <durandal_1707> but does dsp supports 2x4 blocks?
[01:23] <michaelni> sure, h264 chroma needs 2x2, 2x4 4x2 ...
[01:24] <michaelni> too
[01:24] <durandal_1707> and what it calls?
[01:25] <michaelni> i think though if you need just copy simply reading / writing int64_t might be more appropriate
[01:31] <michaelni> h264 uses put_h264_chroma_pixels_tab if iam not mistaken
[01:31] <michaelni> but as said i doubt its a good idea if you just need tiny blocks and just copy
[01:31] <michaelni> with no subpixel stuff
[08:57] <marcosmlopes> hello, could someone help me? I want to implement an streaming server in node.js, and i want to know how ffmpeg transmit video through http. Can someone point me an article or something about this? Thanks
[11:24] <ubitux> michaelni: https://github.com/ubitux/FFmpeg/compare/swr  any objection?
[13:02] <michaelni> ubitux, LGTM
[13:03] <ubitux> michaelni: thx, will push in a moment
[13:20] <CIA-41> ffmpeg: 03Clément BSsch 07master * r3491c2a909 10ffmpeg/libswresample/x86/resample_mmx.h: 
[13:20] <CIA-41> ffmpeg: swr: use __asm__ instead of __asm.
[13:20] <CIA-41> ffmpeg: For consistency only.
[13:20] <CIA-41> ffmpeg: 03Clément BSsch 07master * r022cbb6791 10ffmpeg/libswresample/x86/resample_mmx.h: swr: small align cosmetic.
[13:20] <CIA-41> ffmpeg: 03Clément BSsch 07master * rca612a27ae 10ffmpeg/libswresample/x86/resample_mmx.h: swr: fix make checkheaders.
[13:36] <saste> https://co.clickandpledge.com/advanced/default.aspx?wid=56226
[13:36] <saste> donations enabled
[13:43] <michaelni> \o/
[13:48] <saste> michaelni: I'm going to send a news patch soon
[13:55] <ubitux> i'm adding a checkheaders instance to FATE, please ignore the failures :)
[13:57] <ubitux> btw, i still don't understand why the coverage stuff stopped working
[13:59] <CIA-41> ffmpeg: 03Nicolas George 07master * r20a8dca57f 10ffmpeg/ffmpeg.c: 
[13:59] <CIA-41> ffmpeg: ffmpeg: fix parsing of -force_key_frames option.
[13:59] <CIA-41> ffmpeg: Identical to commit 4f6a1c9 by Reimar Döffinger,
[13:59] <CIA-41> ffmpeg: that was lost after the merge of 19ad567.
[13:59] <CIA-41> ffmpeg: 03Nicolas George 07master * rc9c4835f51 10ffmpeg/tests/ (Makefile fate/options.mak ref/fate/options-force_key_frames): 
[13:59] <CIA-41> ffmpeg: fate: add a test for the -force_key_frames option.
[13:59] <CIA-41> ffmpeg: It broke twice in the last three months due to other changes.
[14:22] <CIA-41> ffmpeg: 03Nicolas George 07master * r0689d5e17a 10ffmpeg/libavfilter/ (audio.c audio.h avfilter.c avfilter.h): 
[14:22] <CIA-41> ffmpeg: lavfi: implement samples framing on links.
[14:22] <CIA-41> ffmpeg: Links can be set up to group samples into buffers of
[14:22] <CIA-41> ffmpeg: specified minimum and maximum size.
[14:22] <CIA-41> ffmpeg: 03Nicolas George 07master * r1f5c1333a7 10ffmpeg/libavfilter/ (buffersink.h sink_buffer.c): sink_buffer: implement fixed frame size.
[14:22] <CIA-41> ffmpeg: 03Nicolas George 07master * r6f5e7e8690 10ffmpeg/ffmpeg.c: ffmpeg: use the more powerful version of abuffersink.
[15:15] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rb12e61ac7f 10ffmpeg/ffplay.c: 
[15:15] <CIA-41> ffmpeg: ffplay: factorize expression in audio_decode_frame()
[15:15] <CIA-41> ffmpeg: Possibly improve readability.
[16:14] <ubitux> (checkheaders seems to be working now)
[17:23] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r9d230ba924 10ffmpeg/libavformat/nutdec.c: 
[17:23] <CIA-41> ffmpeg: nutdec: check that filesize is valid before using it.
[17:23] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:23] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re2c95e6bd8 10ffmpeg/libavformat/nutdec.c: 
[17:23] <CIA-41> ffmpeg: nutdec: remove assert that is wrong for some new duration code.
[17:23] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:23] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r3233ad4b44 10ffmpeg/libavformat/nutdec.c: 
[17:23] <CIA-41> ffmpeg: nutdec: estimate duration from last syncpoint
[17:23] <CIA-41> ffmpeg: Previously nut used the fallback of estimation by bitrate.
[17:23] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:05] <burek> if -vf yadif is so good, why not replace -deinterlace with it?
[18:12] <ubitux> hey, fate almost green *_*
[18:31] <michaelni> burek, because you didnzt sent a patch
[18:31] <burek> :D
[18:31] <burek> :beer: :)))
[18:32] <ubitux> hey burek, want something to add in the wiki? :)
[18:32] <ubitux> i think a hardsub section would be nice
[18:33] <burek> ubitux, I'll copy tutorials from forum
[18:33] <burek> there is (I think) a post about hard subs too
[18:33] <ubitux> so something like "get your subtitles in ASS: ffmpeg -i input.mkv out.ass, or ffmpeg -i input.srt out.ass, check your ffmpeg is build with --enable-libass, hardsub with ffmpeg -i input.avi -vf ass=out.ass output.avi"
[18:33] <ubitux> ok :)
[18:34] <burek> "get your subtitles in ASS" :D
[18:34] <burek> what a wonderfull topic :D
[18:34] <ubitux> :))
[18:35] <ubitux> get your subtitles in ASS, and shift "your" by two words
[18:35] <ubitux> to the right
[18:35] <burek> :))
[19:44] Action: Daemon404 sees michaelni is odifying nutdec
[20:40] <burek> can ffserver do -vcodec copy (restream the live input to output)
[20:40] <burek> http/mpegts is in question
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rac2d3b0f6e 10ffmpeg/libavformat/nutdec.c: 
[20:48] <CIA-41> ffmpeg: nutdec: set duration_estimation_method
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re704ab3d5f 10ffmpeg/libavcodec/wmaenc.c: 
[20:48] <CIA-41> ffmpeg: wmaenc: change another assert to av_assert
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r799df083f3 10ffmpeg/libavcodec/wmaenc.c: 
[20:48] <CIA-41> ffmpeg: wmaenc: fix off by 1 error in quantization loop
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r5968a07622 10ffmpeg/libavcodec/wmaenc.c: 
[20:48] <CIA-41> ffmpeg: wmaenc: rewrite 2nd stage quantization code
[20:48] <CIA-41> ffmpeg: this is faster and more correct
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r0755033f8a 10ffmpeg/libavcodec/wmaenc.c: 
[20:48] <CIA-41> ffmpeg: wmaenc: remove alternative quantization loop
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:48] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r7001eee1f4 10ffmpeg/libavcodec/utils.c: 
[20:48] <CIA-41> ffmpeg: avcodec_encode_audio2: print error messages for errors
[20:48] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:12] <saste> ubitux: how am I supposed to test the rss script?
[21:15] <ubitux> with http://feed2.w3.org/#validate_by_input i guess
[21:15] <ubitux> (htdocs/main.rss)
[21:16] <ubitux> or just manually check in the file if the entry is there
[21:20] <saste> seems fine
[21:20] <saste> ubitux: btw what about http://ffmpeg.org/trac/ffmpeg/ticket/844
[21:20] <saste> should we close it?
[21:21] <saste> i suppose the rss is already working, or not?
[21:21] <ubitux> extracting the content is not easy
[21:21] <ubitux> you can close saying it will be too much effort and unreliable to extract the content of the news and put it in the rss
[21:22] <ubitux> (you need to parse html with sed, and then likely escape it)
[21:25] <saste> uhm... can you suggest a tool/program to test the rss feed?
[21:25] <ubitux> offline i don't know any
[21:26] <ubitux> oh you mean just for viewing?
[21:26] <ubitux> i use newsbeuter
[21:26] <ubitux> but it's a curse app
[21:27] <saste> what about writing an input protocol for ffmpeg? ;-)
[21:28] <saste> (i always wondered why we don't support ftp i/o)
[21:29] <ubitux> there are a lot of input protocol to add :)
[21:30] <ubitux> archives (tar, zip, rar, ...), ftp, samba, ...
[21:30] <saste> nfs, samba, to name a few
[21:30] <ubitux> (i wonder how this would work with a ftp://.../foobar.tar)
[21:31] <ubitux> maybe archives would be a special case
[21:32] <saste> afk
[21:34] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r51d2578348 10ffmpeg/libswresample/resample.c: 
[21:34] <CIA-41> ffmpeg: swr: fix MMX resample code, add emms
[21:34] <CIA-41> ffmpeg: Fixes Ticket1495
[21:34] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:41] <burek> saste, validator.w3.org/appc/
[21:41] <burek> http://validator.w3.org/appc/
[21:49] <CIA-41> ffmpeg: 03Nicolas George 07master * rbe24f85176 10ffmpeg/ffmpeg.c: 
[21:49] <CIA-41> ffmpeg: ffmpeg: re-set frame_size after filter graph reset.
[21:49] <CIA-41> ffmpeg: The filter contexts are completely cleared,
[21:49] <CIA-41> ffmpeg: so the frame size can not be kept.
[22:47] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rf2fd167835 10ffmpeg/libavcodec/x86/ (vc1dsp_mmx.c vc1dsp_yasm.asm): (log message trimmed)
[22:47] <CIA-41> ffmpeg: x86: vc1: fix and enable optimised loop filter
[22:47] <CIA-41> ffmpeg: The problem is that the ssse3 psign instruction does the wrong
[22:47] <CIA-41> ffmpeg: thing here. Commit ea60dfe incorrectly removed a macro emulating
[22:47] <CIA-41> ffmpeg: this instruction for pre-ssse3 code. However, the emulation is
[22:47] <CIA-41> ffmpeg: incorrect, and the code relies on the behaviour of the macro.
[22:47] <CIA-41> ffmpeg: Specifically, the psign sets destination elements to zero where
[22:47] <CIA-41> ffmpeg: 03Ronald S. Bultje 07master * r5aff37d28d 10ffmpeg/libavformat/os_support.c: (log message trimmed)
[22:47] <CIA-41> ffmpeg: os_support: Include all the necessary headers for the win32 open function
[22:47] <CIA-41> ffmpeg: io.h is required for open and _wopen, and fcntl.h is required for
[22:47] <CIA-41> ffmpeg: the O_CREAT flag. On mingw, fcntl.h is included by os_support.h (and
[22:47] <CIA-41> ffmpeg: the mingw fcntl.h includes io.h), but include it explicitly here
[22:47] <CIA-41> ffmpeg: since this implementation requires it.
[22:47] <CIA-41> ffmpeg: Also move the #undef open up. open must not be defined to ff_win32_open
[22:47] <CIA-41> ffmpeg: 03Martin Storsjö 07master * rcdee08e365 10ffmpeg/ (configure libavformat/os_support.h): 
[22:47] <CIA-41> ffmpeg: network: Check for struct pollfd
[22:47] <CIA-41> ffmpeg: We need to include winsock2.h here, to make sure we have the
[22:47] <CIA-41> ffmpeg: real pollfd struct definition, if one exists, before defining the
[22:47] <CIA-41> ffmpeg: fallback poll function.
[22:50] <DGMurdockIII> http://lib-ray.org/
[22:50] <DGMurdockIII> check that out
[22:51] <ohsix> no limits huh, how do you put together a hardware player then
[22:51] <Daemon404> lib-ray was that freetard thing
[22:51] <Daemon404> wasnt it
[22:52] <ohsix> overcomplicated implies there's a level of complication that is acceptable, but that isn't treated
[22:52] <Daemon404> inb4 sony sues for logo plagarism
[22:52] <Daemon404> a la mru
[22:54] <ohsix> replicate all the patentend technology that isn't aacs and all the drm bits they complain about lul
[22:55] <ohsix> oh boy who wouldn't want to start up their web browser to watch a video
[22:55] <DGMurdockIII> http://www.kickstarter.com/projects/2144275086/lib-ray-non-drm-open-standards-hd-video-format
[22:56] <DGMurdockIII> there a video there
[23:08] <nevcairiel> too bad H.264 is just so much better then VP8 for HD content
[23:09] <nevcairiel> VP8 is ok for webstreaming, not for HD movie distribution
[23:09] <durandal_1707> isn't jpeg2000 best?
[23:10] <Daemon404> jpeg2000 is best "for some types of content"
[23:10] <Daemon404> like large flat areas
[23:11] <Daemon404> but jpeg2000 is so bad in all current forms mostly
[23:11] <nevcairiel> i wonder if implementing it is so hard or just no-one cares for a ffmpeg/libav decoder
[23:11] <nevcairiel> a good one anyway
[23:12] <Daemon404> some "pros" might care
[23:12] <Daemon404> red's "raw" format is a bastardization of jpeg2000
[23:12] <JEEB> yeah, in the "professional" area jpg2k is quite used
[23:12] <JEEB> in various, more or less bastardized, forms
[23:13] <nevcairiel> wasnt prores already a bastardization of mjpeg
[23:13] <JEEB> something similar IIRC, yeah
[00:00] --- Sun Jul  1 2012