[Ffmpeg-devel-irc] ffmpeg-devel.log.20120910

[burek]

[06:38] <ubitux> http://www.joelonsoftware.com/articles/fog0000000069.html
[10:06] <durandal_1707> michaelni: what are those snprintf changes about?
[10:29] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * re4fe4d01fc 10ffmpeg/ffprobe.c: 
[10:29] <CIA-56> ffmpeg: ffprobe: simplify default printer
[10:29] <CIA-56> ffmpeg: Remove pointless default_print_footer() and
[10:29] <CIA-56> ffmpeg: default_print_chapter_header() callbacks, also avoid to print an
[10:29] <CIA-56> ffmpeg: empty line when noprint_wrappers is selected.
[11:49] <burek> did you know for http://ideone.com/ ? it's like a pastebin.com but it runs the pasted code as well (around 40 diff compilers available) :)
[11:52] <j-b> nice
[11:52] <av500> paste ffmpeg.c into it
[14:24] <CIA-56> ffmpeg: 03Luca Barbato 07master * r28165fee1b 10ffmpeg/libavfilter/vf_yadif.c: yadif: cosmetics
[14:24] <CIA-56> ffmpeg: 03Samuel Pitoiset 07master * ra0845bae49 10ffmpeg/libavcodec/ (mjpeg.c mjpeg.h mjpeg2jpeg_bsf.c mjpegdec.c mjpegenc.c): 
[14:24] <CIA-56> ffmpeg: mjpeg: Rename some symbols to avpriv_* instead of ff_*
[14:24] <CIA-56> ffmpeg: These symbols will be used from the RTP/JPEG depacketizer.
[14:24] <CIA-56> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[14:24] <CIA-56> ffmpeg: 03Samuel Pitoiset 07master * r3c19815416 10ffmpeg/ (7 files in 3 dirs): 
[14:24] <CIA-56> ffmpeg: rtp: Depacketization of JPEG (RFC 2435)
[14:24] <CIA-56> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[14:24] <CIA-56> ffmpeg: 03Mans Rullgard 07master * rf181662db9 10ffmpeg/tests/fate/mp3.mak: 
[14:24] <CIA-56> ffmpeg: fate: mp3: drop redundant CMP setting
[14:24] <CIA-56> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[14:24] <CIA-56> ffmpeg: 03Mans Rullgard 07master * r8c5ed7a66a 10ffmpeg/libavutil/mem.h: 
[14:24] <CIA-56> ffmpeg: mem: fix typo in check for __ICC
[14:24] <CIA-56> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[14:24] <CIA-56> ffmpeg: 03Diego Biurrun 07master * ra84edbacaf 10ffmpeg/libavcodec/x86/Makefile: x86: dsputil: Only compile motion_est code when encoders are enabled
[14:24] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rbff2afb3e9 10ffmpeg/: (log message trimmed)
[14:24] <CIA-56> ffmpeg: Merge remote-tracking branch 'qatar/master'
[14:24] <CIA-56> ffmpeg: * qatar/master:
[14:24] <CIA-56> ffmpeg:  x86: dsputil: Only compile motion_est code when encoders are enabled
[14:25] <CIA-56> ffmpeg:  rtp: Depacketization of JPEG (RFC 2435)
[14:25] <CIA-56> ffmpeg: 03Martin Storsjö 07master * raefea4d0de 10ffmpeg/libavcodec/ (bitstream.c mjpegenc.c mpeg4videoenc.c put_bits.h): 
[14:25] <CIA-56> ffmpeg: Rename ff_put_string to avpriv_put_string
[14:25] <CIA-56> ffmpeg: This allows using it from libavformat as well. This will be used
[14:25] <CIA-56> ffmpeg: by the RTP/JPEG depacketizer.
[14:25] <CIA-56> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[15:46] <CIA-56> ffmpeg: 03Ben Jackson 07master * r616ef2e8bd 10ffmpeg/libavcodec/pngenc.c: 
[15:46] <CIA-56> ffmpeg: lavc/pngenc: Enable frame threading for PNG encodes
[15:46] <CIA-56> ffmpeg: Signed-off-by: Ben Jackson <ben at ben.com>
[15:46] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:31] <pcbb> what is the error code -5?
[16:44] <pcbb> av_read_frame were leaking memory on mjpeg/http stream ... works fine with rtsp
[16:45] <michaelni> pcbb, where does it leak?
[16:46] <michaelni> you can use vakgrind to find ...
[16:47] <michaelni> -5 might be -EIO (check errno.h and related headers)
[16:47] <pcbb> michaelni -EIO? what is it?
[16:48] <nevcairiel> IO error
[16:49] <pcbb> humm
[16:49] <nevcairiel> that can mean a lot of stuff, usually it gives more hints through its logging
[16:49] <pcbb> don't logging...
[16:50] <pcbb> I see
[16:50] <pcbb> the error now...
[16:50] <pcbb> ffmpeg lib accepts https or need modifications??
[16:50] <michaelni> btw nevcairiel, theres an alternative OTF in matroska patch ([PATCH] [matroska] add OTF support for attachments), i think you sent the other, can you comment? id like to apply one of them ...
[16:51] <pcbb> don't logging loging :(
[16:51] <nevcairiel> yeah i saw it, i really don't care if it gets a new codec id or not, the advantage of keeping the same id is that applications dont need changes in most cases, the disadvantage is that it may happen that some obscure font engine needs to know the difference
[16:52] <michaelni> dont we need a seperate one for stream copy too ?
[16:53] <pcbb> av_read_frame -> 'http://root:<password>@pcbb.dyndns.org:8080/axis-cgi/mjpg/video.cgi'
[16:53] <nevcairiel> I dont really know much about how transmuxing works
[16:53] <pcbb> it's correct I think
[16:53] <nevcairiel> i guess it helps the muxer to know the difference
[16:53] <michaelni> pcbb, https should be supported if you have the dependancies installed
[16:54] <michaelni> nevcairiel, ok then ill apply the one that adds a new codec id
[16:57] <pcbb> michaelni the url is fine ?
[16:57] <pcbb> I need call find stream information ?
[16:57] <pcbb> can be this
[16:57] <pcbb> ;P
[17:01] <pcbb> -.-
[17:01] <pcbb> I'll be crazy
[17:01] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rb422e2c58a 10ffmpeg/libavformat/utils.c: 
[17:01] <CIA-56> ffmpeg: compute_pkt_fields: fix harmless integer overflow
[17:01] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:01] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r41dba453f9 10ffmpeg/libavformat/utils.c: 
[17:01] <CIA-56> ffmpeg: utils/update_stream_timings: fix division by zero
[17:01] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:01] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r108957c661 10ffmpeg/libavformat/avidec.c: 
[17:01] <CIA-56> ffmpeg: avidec/guess_ni_flag: fix harmless integer overflow
[17:01] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:06] <michaelni> pcbb, just tried a http aith auth, and that works
[17:06] <michaelni> With
[17:07] <michaelni> so dunno, maybe submit a bug report with full output to trac
[17:07] <michaelni> https://ffmpeg.org/trac/ffmpeg
[17:08] <michaelni> also i would suggest you dont send your root password over http if thats what you are doing :)
[17:12] <pcbb> michaelni how set user and pass with ffmpeg lib ... without in url ?
[17:13] <pcbb> michaelni is strange because rtsp works fine!!!
[17:15] <CIA-56> ffmpeg: 03Sébastien Brochet 07master * re14725cecc 10ffmpeg/ (3 files in 2 dirs): 
[17:15] <CIA-56> ffmpeg: add OTF support for attachments
[17:15] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:29] <pcbb> michaelni??
[17:39] <michaelni> pcbb, if it doesnt with ffmpeg and for example works with wget then please submit a full bug report to https://ffmpeg.org/trac/ffmpeg
[17:51] <pcbb> cool
[17:51] <pcbb> I've a warning message now!!!
[17:52] <pcbb> Format loas detected only with low score of 1
[18:00] <ubitux> saste: where would the class_name be printed?
[18:00] <saste> ubitux: context
[18:00] <ubitux> i'm looking at your ffprobe macro thing where the class name becomes the same as the result from get_name callback
[18:01] <saste> in case of errors, e.g. parsing options errors
[18:01] <ubitux> mmh
[18:01] <ubitux> wouldn't av_log() use the get_name callback?
[18:01] <saste> i'm cheating because i'm renaming the context "IniContext" -> "ini"
[18:01] <saste> but the latter is saner
[18:02] <ubitux> (and not the AVClass class_name)
[18:02] <saste> ah that one
[18:02] <saste> i'm confused about that, just following cargo cult
[18:02] <ubitux> i see two names, the one from the callback, and the one for the AVClass.class_name
[18:02] <ubitux> and since you make them the same, i was wondering what was the purpose of class_name
[18:03] <saste> yes i never remember why that
[18:03] <ubitux> and if it actually couldn't be NULL when the callback is defined
[18:03] <saste> the answer is in log.[hc] I suppose
[18:04] <saste> I'm silly: 57be00bea7d66ea28caab6e32753bc167b7bccf1
[18:05] <saste> no i'm not, just noisy
[18:05] <ubitux> :)
[18:16] Action: saste is accumulating local patches which no one reviews
[18:22] Action: michaelni reviews 2 patches
[18:24] <saste> michaelni: thx :)
[18:30] <saste> ubitux: now that you're here, any opinion on the -show_entries patch?
[18:31] <saste> i think that i'll drop "/" as separator, and I'll simply add the necessary opt_show_* helpers
[18:36] <ubitux> i need to test it, i'll have a look in a moment
[18:55] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r206c34e17d 10ffmpeg/libavfilter/ (f_setpts.c version.h): 
[18:55] <CIA-56> ffmpeg: lavfi/setpts: set SAMPLE_RATE to NAN when input is no audio
[18:55] <CIA-56> ffmpeg: Should be more robust/consistent.
[18:55] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r722762f7e1 10ffmpeg/ (3 files in 2 dirs): 
[18:55] <CIA-56> ffmpeg: lavfi/setpts: add FRAME_RATE constant
[18:55] <CIA-56> ffmpeg: Useful for dealing with constant frame-rate video.
[19:35] <ubitux> saste: just tested your patch, and a simple ./ffprobe -show_streams -of json foobar.mkv is all broken :p
[19:35] <ubitux> i guess that's what you were talking about in your comment
[19:36] <ubitux> (it looks like this here: http://pastie.org/4697018)
[19:37] <saste> yes it is what I expect
[19:37] <saste> current patch is broken
[19:38] <ubitux> btw, still not motivated to add some ffprobe reg test? :)
[19:39] <saste> well in this case the problem is the interaction with options
[19:45] <ubitux> saste: something like if (do_show_streams && !i_haz_show_stream_streams_entries) opt_show_entries(..., "show_entries", "stream") wouldn't do the trick?
[19:55] <saste> ubitux: no basically we need to make the writers more aware of the structure of the tree to show
[19:56] <saste> for example how to ask to only print a tag with a certain name in streams?
[19:56] <saste> also tags are special because are considered a "section" by certain writers, and a "pseudo-section" by others (e.g. compact, default)
[19:57] <saste> I could keep the old patch, which is a striaghtforward generalization of -show_format_entry
[19:58] <saste> and improve the behavior of -show_entries later
[19:58] <saste> (even if that could break parsing later)
[19:59] <brimestone> Hello, what are the option for me to accelerate FFmpeg/FFmbc trancoding? is there a known hardware outhere that will do this?
[20:12] <Compn> hmm, not sure what hardware is recommended for that
[20:13] <Skyler_> core i7 :3
[20:16] <Compn> stuff like crystalhd or nvidia vdpau or intel vaapi ...
[20:16] <Compn> but most of those are used for playback, not transcoding 
[20:34] <brimestone> Skyler i have a MacPro with 32GB RAM and and 12Core machine
[20:34] <brimestone> i also have a 3 Disk RAID0 array.. so hardware wise.. its fully beef'ed up.. 
[20:35] <Skyler_> next, check your encoding settings
[20:35] <Skyler_> if you care about speed, make sure you've picked a fast preset.
[20:36] <brimestone> well i have a very specific setting
[20:37] <brimestone> http://pastebin.com/Ah1QEpPw
[20:37] <JEEB> oh, dnxhd
[20:37] <Skyler_> ahhh
[20:37] <JEEB> no idea how well equipped that or the video filter are for speed/multithreading
[20:37] <Skyler_> -mbd rd might make it rather slow.
[20:38] <JEEB> oh
[20:38] <Skyler_> I'm not sure whether that filter or encoder are threaded; if not you could run multiple instances.
[20:38] <Skyler_> to load all 12 cores
[20:39] <brimestone> so cut the source into 12 chunks and stich it all up?
[20:40] <gnafu> You could pipe one instance of ffmbc (or ffmpeg) to another, such as one to decode, one to apply the filter, and one to encode to DNxHD.
[20:40] <gnafu> Not sure how much that would help.
[20:41] <gnafu> What is your typical input file?  More DNxHD?
[20:50] <brimestone> its a ProRes media from an Arri Alexa Camera 
[20:50] <Skyler_> oh wow, an Arri
[20:53] <brimestone> so im writing a script so that i can add a adjust the color from LogC to Rec709 and at the same time convert it to DNxHD for editing
[20:56] <brimestone> so.. thats it huh.. no available hardware accelerator available
[20:57] <Skyler_> doubtful, especially for such an obscure format
[20:57] <brimestone> obscure.. hahahaha
[20:57] <Skyler_> but definitely start by not using -mbd rd
[20:57] <ohsix> you are already undermining your own case
[20:57] <brimestone> ill test it 
[20:58] <Skyler_> I'm not sure how optimized for speed the ffmpeg dnxhd is either.
[20:59] <brimestone> ohhh
[21:00] <Skyler_> but splitting it among cores will definitely help
[21:03] <cbsrobot-> brimestone: btw do you have a 12bit proves sample somewhere ?
[21:03] <cbsrobot-> *prores
[21:08] <brimestone> wow.. from 15 seconds to 4 second trancoding a 1 second footage
[21:08] <brimestone> i do.. 
[21:08] <cbsrobot-> can you share it ?
[21:08] <cbsrobot-> just a few frames if possible
[21:08] <brimestone> yes.. let me cut someething that wont get me in trouble
[21:09] <cbsrobot-> make sure it is 12 bit !
[21:09] <brimestone> i know.. wiat 
[21:10] <brimestone> i have to cut it cuz some footage are license and all.. let me get some generic test pattern stuff
[21:13] <brimestone> ok i got one 
[21:14] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rb3b27b63a0 10ffmpeg/libavformat/mpegtsenc.c: 
[21:14] <CIA-56> ffmpeg: mpegtsenc: add forgotten ()
[21:14] <CIA-56> ffmpeg: Fixes integer overflow
[21:14] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:14] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r5829b7d29e 10ffmpeg/ffmpeg.c: 
[21:14] <CIA-56> ffmpeg: ffmpeg: check timebase validity before considering it to be used for stream copy
[21:14] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:14] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r17b3251c6c 10ffmpeg/libavcodec/ra144enc.c: 
[21:14] <CIA-56> ffmpeg: ra144enc: avoid integer overflows.
[21:14] <CIA-56> ffmpeg: The values are all positive but signed variables where used,
[21:14] <CIA-56> ffmpeg: which overflowed.
[21:14] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:14] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r14de55784d 10ffmpeg/libswscale/output.c: 
[21:14] <CIA-56> ffmpeg: sws/output: use unsigned variables where harmless overflows are expected.
[21:14] <CIA-56> ffmpeg: signed overflow is undefined in C ...
[21:14] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:14] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rf0d5050090 10ffmpeg/ffmpeg.c: 
[21:14] <CIA-56> ffmpeg: ffmpeg: fix bytes written statistics.
[21:14] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:15] <brimestone> http://tedsimbajon.com/prores10bit.mov
[21:15] <brimestone> just download it 
[21:15] <JEEB> I think 10bit is already supported, 12bit is what needs implementation IIRC
[21:16] <brimestone> the source footage is only 10bit? 
[21:16] <brimestone> 10 bit 4:4:4:4
[21:19] <cbsrobot-> I was looking for 12 bit ...
[21:19] <brimestone> ohh sorry.. i coulda swear our footage was 12 bit
[21:19] <brimestone> i guess its just 10
[21:20] <JEEB> can't you have the encoder just encode it as 12bit?
[21:20] <cbsrobot-> nope
[21:20] <brimestone> why would you do that? source is just 10
[21:21] <brimestone> i think with Alexa, you can only do 12 bit with uncompress mode
[21:21] <JEEB> <brimestone> why would you do that? source is just 10 <- the idea is to get a sample of the 12bit encoding mode
[21:21] <JEEB> I don't think it really matters if the source is 12bit or not
[21:21] <brimestone> http://www.arri.com/camera/digital_cameras/cameras/camera_details.html?product=8&subsection=technical_data&cHash=177d18e5da
[21:23] <brimestone> ohh just to create a 12bit file? hmmm im not even sure if ffmpeg does 12 bit encode
[21:23] <cbsrobot-> hop it does not
[21:24] <cbsrobot-> thats why I'm asking if you have one
[21:25] <cbsrobot-> brimestone, btw
[21:25] <brimestone> yes?
[21:25] <cbsrobot-> I think dnxhd is intra only
[21:25] <brimestone> i believe so
[21:25] <cbsrobot-> you could try and add CODEC_CAP_FRAME_THREADS | CODEC_CAP_INTRA_ONLY to the codec capabilites
[21:26] <Skyler_> does it do frame threads yet?
[21:26] <cbsrobot-> not sure it works ...
[21:27] <cbsrobot-> it has slice_threads
[21:27] <Skyler_> oh, that might work then.
[21:27] <Skyler_> -threads 8 or -threads 12, see how that works?
[21:28] <cbsrobot-> and -vf mp=eq2 should also be available in ffmpeg
[21:29] <cbsrobot-> and timecode too
[21:30] <brimestone> No such filter: 'mp' thats what i get with ffmpeg
[21:31] <ohsix> cheeky bot, most people are warded off the -devel channel for those questions :p
[21:32] <cbsrobot-> I don't mind relocating  &.
[21:33] <ohsix> i don't mind not relocating, just stating what's normal :p do what you like
[21:33] <cbsrobot-> but I did not see Skyler_ on the other channel for a while &. sadly 
[21:45] <cbsrobot-> brimestone: if it works with CODEC_CAP&. - please send a patch.
[21:57] <pcbb> I'm using ffmpeg lib to capture Motion JPEG over RTSP. The received frame is not a JPEG, does not have the headers and not have the structure of a JPEG file too. How can I convert it to a JPEG?
[22:47] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rc5f0b6bf41 10ffmpeg/libavcodec/atrac3.c: 
[22:47] <CIA-56> ffmpeg: atrac3: replace powf(2,...) by exp2f()
[22:47] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:47] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r76fac07334 10ffmpeg/libavcodec/mpegaudiodec.c: 
[22:47] <CIA-56> ffmpeg: mpegaudiodec: replace pow(2.0,...) by exp2()
[22:47] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:13] <pcbb> I'm using ffmpeg lib to capture Motion JPEG over RTSP. The received frame is not a JPEG, does not have the headers and not have the structure of a JPEG file too. How can I convert it to a JPEG?
[23:13] <pcbb> michaelni?
[23:19] <gnafu> If it's not JPEG, what is it?
[23:21] <gnafu> It almost sounds like a riddle.
[23:24] <pcbb> the4 format was changed because the rtp protocol
[00:00] --- Tue Sep 11 2012