[Ffmpeg-devel-irc] ffmpeg-devel.log.20120616

[burek]

[00:13] <ubitux> michaelni: how is generated the doc on the website?
[00:18] <ubitux> simply a cron which pull ffmpeg, run a ``make documentation`` and then cp doc/*.html htdocs/ ?
[00:29] <ubitux> strangely syntax.html isn't available
[00:32] <llogan> saste: FancyFilteringExamples now has a link on the main page (and I'm curious about your mirror example).
[00:32] <saste> llogan: yes 20 minutes
[00:33] <saste> i also discovered a leak in split/asplit
[00:35] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * rf0394f28c3 10ffmpeg/libavfilter/avcodec.c: 
[00:35] <CIA-119> ffmpeg: lavfi/avcodec: add consistency checks in avfilter_copy_buf_props()
[00:35] <CIA-119> ffmpeg: The function will abort through an assert if the source is not defined,
[00:35] <CIA-119> ffmpeg: or if the internal state of the source is inconsistent (e.g. type = AUDIO
[00:35] <CIA-119> ffmpeg: && !src->audio).
[00:35] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * r6808759a58 10ffmpeg/libavcodec/rawdec.c: 
[00:35] <CIA-119> ffmpeg: lavc/rawdec: fix various bogus error codes and add some logs
[00:35] <CIA-119> ffmpeg: Help debugging.
[00:35] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * r3cc1a89881 10ffmpeg/libavcodec/utils.c: 
[00:35] <CIA-119> ffmpeg: lavc/utils: in avcodec_decode_video2() return proper error code instead of -1
[00:35] <CIA-119> ffmpeg: Return AVERROR(EINVAL) in case of invalid coded size.
[00:35] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * r635a25954e 10ffmpeg/libavfilter/video.c: 
[00:35] <CIA-119> ffmpeg: lavfi/video: set output frame w/h properties in default_start_frame()
[00:35] <CIA-119> ffmpeg: Use the same values of the video output link.
[00:35] <CIA-119> ffmpeg: Avoid the need to override the default_start_frame() with an ad-hoc
[00:35] <CIA-119> ffmpeg: start_frame() callback.
[00:36] <llogan> saste: no hurry. i was just mentioning it.
[00:36] <CIA-119> ffmpeg: iteration.
[00:36] <CIA-119> (4 lines omitted)
[00:44] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * ra1411eec21 10ffmpeg/ffprobe.c: 
[00:44] <CIA-119> ffmpeg: ffprobe: add writer_print_rational()
[00:44] <CIA-119> ffmpeg: Improve overall consistency, allow some factorization.
[00:57] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * re20530add7 10ffmpeg/libavfilter/split.c: 
[00:57] <CIA-119> ffmpeg: lavfi/split: give a more accurate description to split
[00:57] <CIA-119> ffmpeg: The filter now supports a configurable number of outputs.
[00:57] <CIA-119> ffmpeg: 03Stefano Sabatini 07master * r2b1fc5621d 10ffmpeg/ (6 files in 3 dirs): 
[00:57] <CIA-119> ffmpeg: lavfi: add asetnsamples audio filter
[00:57] <CIA-119> ffmpeg: This filter changes the number of samples on single output operation.
[00:57] <CIA-119> ffmpeg: Based on a patch by Andrey Utkin <andrey.krieger.utkin at gmail.com>.
[01:21] <saste> llogan: ffplay -i INPUT -vf "crop=iw/2:ih:0:0,split[tmp],pad=2*iw[left]; [tmp]hflip[right]; [left][right] overlay=W/2"
[01:21] <Compn> would everyone else be comfortable with a dictator calling donations approvals ? voting has never been ffmpeg's strong point
[01:21] <Compn> saste : and i'm guessing it would make SPI's job easier
[01:21] <saste> Compn: actually two dictators ;-)
[01:22] <saste> what i mean is that from what we discussed the requests must be approved by two different persons
[01:22] <saste> liaison and pm
[01:22] <Compn> yeah liason figured in there too
[01:22] <llogan> saste: dude. hand me the bong.
[01:23] <Compn> saste : i'm just remembering how many dumb votes this project has had and how most of the votes just go horribly wrong. people making new options for the polls and stuff mid-vote etc
[01:24] <Compn> which in itself isnt a problem
[01:24] <Compn> but then someone else says the vote is invalid
[01:24] <Compn> nothing but trouble
[01:24] Action: Compn afk
[01:25] <saste> Compn: as I said we're experimenting
[01:25] <saste> note that in the current "process" there are no votes
[01:25] <saste> a fund request is approved if *both* the pm and the liaison approve it
[01:26] <saste> we can change it later, but the idea of this "power split" was to avoid to concentrate too much power on a single person (the pm)
[01:26] <saste> and also to relieve michael from some organizational overhead
[01:28] <iive> the idea is good, but you should have some rotation in mind.
[01:30] <iive> e.g. changing one of the dictators each year.
[01:31] <saste> iive: liaison role has a fixed duration (yet to decide, one/two years shall be good)
[01:31] <saste> also we can force rotation every N mandates
[01:31] <iive> sounds good.
[01:43] <saste> michaelni: ffplay matrixbench_mpeg2.mpg -vf "crop=iw/2:ih:0:0,pad=2*iw:ih:0:0:red"
[01:43] <saste> maybe you want to debug it...
[01:44] <saste> something to deal with lavc buffer permissions
[01:47] <ubitux> saste: and here you have the pic for illustrating that example: http://blackhole.ubitux.fr/ffplay-mirror.jpg
[01:48] <saste> ubitux: feel free to add an entry in the wiki
[01:48] <ubitux> yup ok
[01:48] <saste> with mpeg codecs it presents artifacts though
[01:48] <ubitux> dunno where to add pics...
[01:50] <durandal_1707> our j2k decoder and encoder is not lossless
[01:51] <ubitux> saste: can you update your tree with the showwaves filter? :)
[01:52] <saste> ubitux: what do you mean? i just sent it to the ML
[01:52] <ubitux> yes, but on your public tree
[01:52] <ubitux> so it's easier to deal with branches directly ;)
[01:52] <ubitux> (yeah i'm lazy :D)
[01:52] <saste> it applies cleanly to latest master
[01:53] <saste> but yes i'll do but not now, my brain screams to sleep
[01:53] <ubitux> ok ok no problem i'll use the patch on the ml :)
[01:53] <ubitux> thanks for working on it! :)
[01:59] <llogan> ubitux: what character is that?
[02:00] <durandal_1707> michaelni: was j2k decoder ever correctly decoding?
[02:00] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r46c50b1726 10ffmpeg/libavfilter/vf_pad.c: 
[02:00] <CIA-119> ffmpeg: vf_pad: check that we have write permission before writing in the buffer.
[02:00] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:00] <ubitux> page updated
[02:01] <ubitux> llogan: Gone Nutty
[02:01] <ubitux> extra video for ice age
[02:01] <ubitux> https://en.wikipedia.org/wiki/Gone_Nutty
[02:01] <ubitux> page updated: https://ffmpeg.org/trac/ffmpeg/wiki/FancyFilteringExamples
[02:01] <ubitux> saste: do you have a better example for cellauto?
[02:02] Action: llogan is pop culturally ignorant.
[02:02] <saste> yes, there are many cool rules, but 110 is "special" in the sense that it was demonstrated that automaton was turing-complete
[02:02] <ubitux> llogan: http://samples.ffmpeg.org/avi/GoneNutty.avi :)
[02:03] <ubitux> saste: ok :)
[02:03] <michaelni> durandal_1707, decoding as in 5/3 lossless bitexact, i think ive never tested, i just tested lossy
[02:03] <saste> also it weirdly reminds of feynman schemes, which describe interactions of elementary particles
[02:05] <saste> ubitux: cellauto=rule=18
[02:05] <ubitux> oh.
[02:05] <ubitux> looks like some kind of plateform game universe
[02:05] <durandal_1707> michaelni: i'm working on adding other pix_fmts to encoder and encountered that bug
[02:06] <saste> 22, 54
[02:06] <ubitux> it's a tree!
[02:07] <saste> these are somehow interesting: 9,18,22,26,30,41,45,50,54,60,62,73,75,77,82,86,89,90,97,99,101,102,105,107,109,110,124,126,129,131,133,135,137,145,146,149,150,151,153,154,161,167,169,181,182,183,193,195,210,218,225
[02:08] <durandal_1707> so what's up with showwaves filter? (or any A->V filter)
[02:11] <saste> durandal_1707: sound visualization, useful for debugging or just fun
[02:13] <durandal_1707> i want them
[02:13] <ubitux> or gives services like youtube something to show when uploading sound-only files
[02:13] <ubitux> and this is great :)
[02:30] <durandal_1707> michaelni: and broken yuv4XY on encoder side is really irritating me
[02:50] <CIA-119> ffmpeg: 03Clément BSsch 07master * rc7bdfbe79f 10ffmpeg/MAINTAINERS: MAINTAINERS: add my GPG fingerprint.
[03:03] <durandal_1707> and decoder side is also broken for yuv4XYp....
[09:22] <burek>     if (first_index == 5)
[09:22] <burek>         goto fail;
[09:23] <burek> :D I haven't seen goto in years :D
[09:23] <ohsix> single exit is nice if you can do it :p
[09:23] <burek> fail:    return -1;
[09:23] <burek> :D
[09:24] <burek> ok ok, I don't want to start a flame or anything similar :) I'll shut up :D
[09:28] <nevcairiel> goto has its uses, especially when providing a single function exit with resource cleanup
[09:29] <ohsix> the harmful stuff was from people basically doing what they did in assembly, in c
[09:30] <ohsix> a bunch of cleanup sites when you avoid goto might indicate you need to split up the function too :p
[13:03] <CIA-119> ffmpeg: 03Giorgio Vazzana 07master * rc78cf00bd8 10ffmpeg/libavutil/md5.c: (log message trimmed)
[13:03] <CIA-119> ffmpeg: md5: consistently use uint32_t instead of unsigned int
[13:03] <CIA-119> ffmpeg: Basically to make code clearer and adherent to the
[13:03] <CIA-119> ffmpeg: standard. RFC 1321, on page 2 states
[13:03] <CIA-119> ffmpeg: Let the symbol "+" denote addition of words (i.e., modulo-2^32
[13:03] <CIA-119> ffmpeg: addition). Let X <<< s denote the 32-bit value obtained by circularly
[13:03] <CIA-119> ffmpeg: shifting (rotating) X left by s bit positions.
[16:34] <CIA-119> ffmpeg: 03Paul B Mahol 07master * rfd8def9248 10ffmpeg/ (2 files in 2 dirs): 
[16:34] <CIA-119> ffmpeg: lavfi/super2xsai: fix fate test on bigendian
[16:34] <CIA-119> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[16:38] <ubitux> do i need to use FF_INPUT_BUFFER_PADDING_SIZE for codec extradata when i know the codec access will be safe?
[16:38] <ubitux> or there is weird stuff going on that extradata outside the codec?
[16:39] <nevcairiel> is there a reason not to do it? its not like 8 bytes of padding or what its size now will hurt
[16:40] <ubitux> because it's simpler not to in my case
[16:40] <ubitux> basically i just have to do st->codec->extradata = av_strdup(buf.str); st->codec->extradata_size = buf.len+1;
[16:40] <ubitux> (it's a subtitles codec)
[16:43] <durandal_1707> are we gonna follow replacing public api functions in lavfi?
[16:45] <ubitux> durandal_1707: you mean making them private?
[16:45] <durandal_1707> yep
[16:45] <durandal_1707> like what is qatar doing
[16:46] <ubitux> if not it will be a pain to follow them when they will break the prototypes
[16:46] <ubitux> even with public api they don't care very much, so they won't hesitate to change them in the private api (this is normal)
[16:59] <durandal_1707> ubitux: i'm doing that boring job right now ...
[17:21] <ubitux> what job?
[17:21] <durandal_1707> replacing functions in lavfi
[17:22] <ubitux> making our public ones not present on qatar private?
[17:22] <durandal_1707> no
[17:22] <durandal_1707> replacing deprecated functions
[17:22] <ubitux> oh, you mean update the callers?
[17:23] <ubitux> yeah that would be nice
[17:34] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r809d71d650 10ffmpeg/libswscale/swscale_unscaled.c: 
[17:34] <CIA-119> ffmpeg: sws/swscale_unscaled: use av_assert
[17:34] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:34] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * re003da7139 10ffmpeg/libswscale/utils.c: 
[17:34] <CIA-119> ffmpeg: sws/utils: use av_assert
[17:34] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:16] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r3000074495 10ffmpeg/libavcodec/mpegvideo_common.h: 
[18:16] <CIA-119> ffmpeg: mpegvideo_common.h: use av_assert
[18:16] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:16] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r0354412aa3 10ffmpeg/libavformat/ffmenc.c: 
[18:16] <CIA-119> ffmpeg: ffmenc: use av_assert
[18:16] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:02] <durandal_1707> i have 34 patches, can i send them on ml for review?
[19:02] <ubitux> no, better wait until you have 50 patches
[19:03] <nevcairiel> lol
[19:05] <durandal_1707> incoming spam
[19:09] <Compn> 33, no more, no less
[19:09] <Compn> one of your patches will have to be destroyed
[19:09] <Compn> 31 is right out!
[19:09] <Compn> and 35 is just silly
[19:10] <Compn> 32 would be ok, but its not tuesday, so out of the question
[19:10] <ubitux> sounds complicated
[19:10] Action: Compn returns to the ministry of silly walks
[19:11] <Compn> monty python comedy, still funny
[19:13] <Compn> http://www.youtube.com/watch?v=xOrgLj9lOwk
[19:16] <burek> HHG rulez :D
[19:18] <ubitux> :)
[19:33] <durandal_1707> why some AVCLasses from lavfi do not have .version set?
[19:43] <durandal_1707> and bunch of filters have not .category...
[19:43] <ubitux> ah i said i should take some time to add some .category where missing
[19:44] <ubitux> sorry i still didn't do it
[19:44] <durandal_1707> i can do it, count more than 50 patches
[19:44] <ubitux> do it in one patch?
[19:45] <durandal_1707> ok, without sending to ml?
[19:45] <ubitux> do you believe in you?
[19:46] <durandal_1707> a believe in fate and compiler
[19:46] <ubitux> :)
[19:46] <ubitux> mmh it seems i could make something much better for the weird xml parsing stuff in SAMI
[19:47] <ubitux> at least the way i did it for RT is way better...
[19:47] <ubitux> looks like i'll update the SAMI soon :)
[20:10] <durandal_1707> why ffmpeg does not list filters options?
[20:21] <Compn> no one wanted to make that bikeshed on how ffmpeg output should be changed? :P
[20:21] <Compn> maybe no one thought about it :)
[20:22] <durandal_1707> otherwise i'm force to read source code or broken documentation
[20:30] Action: ubitux also always read the source to get list of options..
[20:33] <Alkarex> Hello, I am trying ffmpeg on Raspberry Pi with Debian Wheezy. I get an error "Illegal instruction" just by starting ffmpeg. strace: http://pastebin.com/7RHCr3B2 . Any idea?
[20:33] <Alkarex> I test the image from http://www.raspberrypi.org/phpBB3/viewtopic.php?f=50&t=8071 and ffmpeg works fine on Raspberry Pi with the official image based on Debian Squeeze
[20:34] <Compn> durandal_1707 : patch welcome :)
[20:35] <Compn> Alkarex : it means that the ffmpeg binary has some cpu instruction that your raspberry does not
[20:35] Action: Compn states the obvious
[20:36] <Alkarex> Compn: Ok. Do you know a way to tell which instruction more precisely?
[20:37] <Compn> after reading your pastebin, no , i do not
[20:37] <Compn> how did you compile ffmpeg ?
[20:37] <Compn> Alkarex : did you copy the configure option from the official ffmpeg build to build your ffmpeg ?
[20:37] <Compn> that would be smart
[20:37] <Alkarex> Not compiled. "sudo apt-get install ffmpeg" just like I did on Debian Squeeze
[20:37] <Compn> heh
[20:37] <michaelni> Alkarex, gdb will tell you the instruction
[20:37] <michaelni> but it probably wont help :)
[20:38] <michaelni> you will have to compile it yourself anyway
[20:38] <Alkarex> ok
[20:38] <Compn> Alkarex : you ran that on your raspberry ?
[20:38] <michaelni> with the correcr --enable* --cpu --arch options
[20:38] <burek> i want raspberry :(
[20:38] <Compn> yes, i suggest copying the compile line from the official ffmpeg build
[20:38] <Compn> and dont forget to install the toolchain required :)
[20:39] <Compn> i'll take a raspberry too if anyone has extras ;P
[20:39] <Alkarex> So the processor requirements have possibly changed since Debian Squeeze
[20:39] <michaelni> i guess debian just disables all optimizations so the code can run everywhere but iam just guessing
[20:39] <nevcairiel> apparently not, as it doesnt work :d
[20:39] <Alkarex> Thanks for your feedback
[20:39] <Compn> you got it from debian, blame debian :)
[20:40] <Alkarex> Compn: Yes :-P
[20:40] <michaelni> nevcairiel, right, yes :)
[21:30] <ubitux> saste: hey :)
[21:30] <ubitux> saste: it seems i won't need any xml parser or so
[21:30] <ubitux> https://github.com/ubitux/FFmpeg/blob/424d5901126b0a9bcb258e52a0421fd7e18e2a32/libavformat/realtextdec.c
[21:31] <ubitux> it seems fairly simple that way
[21:31] <ubitux> maybe i'll reuse extract_next_chunk() for SAMI
[21:40] <saste> ubitux: allright, the simpler the better
[21:41] <ubitux> i'll submit the RT demuxer & decoder in a few hours hopefully
[21:41] <ubitux> and then i hope i'd be done with the insane formats
[21:41] <ubitux> only the simple ones should be left
[22:13] <CIA-119> ffmpeg: 03Paul B Mahol 07master * rc9e183b490 10ffmpeg/libavfilter/ (34 files): 
[22:13] <CIA-119> ffmpeg: lavfi: update some deprecated functions
[22:13] <CIA-119> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[22:17] <ubitux> thank you for this durandal_1707 :)
[23:22] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r9946a6aa55 10ffmpeg/libavcodec/x86/diracdsp_yasm.asm: 
[23:22] <CIA-119> ffmpeg: diracdsp: try to fix segfault
[23:22] <CIA-119> ffmpeg: This might fix Ticket1412
[23:22] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:25] <CIA-119> ffmpeg: 03Anton Khirnov 07master * ra5db8e4a1a 10ffmpeg/libavformat/utils.c: 
[23:25] <CIA-119> ffmpeg: lavf: don't segfault when a NULL filename is passed to avformat_open_input()
[23:25] <CIA-119> ffmpeg: This can easily happen when the caller is using a custom AVIOContext.
[23:25] <CIA-119> ffmpeg: Behave as if the filename was an empty string in this case.
[23:25] <CIA-119> ffmpeg: CC: libav-stable at libav.org
[23:25] <CIA-119> ffmpeg: 03Anton Khirnov 07master * rae702edf43 10ffmpeg/libavformat/flacdec.c: flacdec: read attached pictures.
[23:25] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r5c14b282d1 10ffmpeg/: 
[23:25] <CIA-119> ffmpeg: Merge remote-tracking branch 'qatar/master'
[23:25] <CIA-119> ffmpeg: * qatar/master:
[23:25] <CIA-119> ffmpeg:  flacdec: read attached pictures.
[23:26] <CIA-119> ffmpeg:  lavf: don't segfault when a NULL filename is passed to avformat_open_input()
[23:26] <CIA-119> ffmpeg: Merged-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Sun Jun 17 2012