[Ffmpeg-devel-irc] ffmpeg-devel.log.20120601

Sat Jun 2 02:05:03 CEST 2012

[00:14] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r3c276ac0f8 10ffmpeg/libavformat/dv.c: 
[00:14] <CIA-119> ffmpeg: dv-demux: dont mess with codec values
[00:14] <CIA-119> ffmpeg: Fixes part of Ticket1369
[00:14] <CIA-119> ffmpeg: Found-by: ami_stuff
[00:14] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:14] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r7cb8663362 10ffmpeg/libavcodec/h264.c: 
[00:14] <CIA-119> ffmpeg: h264: Fail on DPC its not fully implemented
[00:14] <CIA-119> ffmpeg: Fixes part of Ticket1369
[00:14] <CIA-119> ffmpeg: Found-by: ami_stuff
[00:14] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:14] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * re47e23698b 10ffmpeg/libavcodec/libmp3lame.c: 
[00:14] <CIA-119> ffmpeg: libmp3lame: add missing layout terminator
[00:14] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:15] Action: Compn wonders how many h264 things are unsupported still
[00:15] <Compn> probably a lot
[00:17] <Daemon404> h264 has a buttload that nobody uses or has implemented or cares about
[00:27] <pasteeater> i wonder what ami_stuff is doing to find all of those oddball bugs.
[00:27] <pasteeater> probably submitted the most bugs on trac
[00:38] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r8a0efa9cc0 10ffmpeg/libavcodec/qdm2.c: 
[00:38] <CIA-119> ffmpeg: qdm2: fix incorrect error spam
[00:38] <CIA-119> ffmpeg: Fixes Ticket1375
[00:38] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:06] <kierank> Compn: the important things are supported
[01:06] <kierank> but there are decoder mismatches in places
[01:21] <michaelni> kierank, please report decoder mismatches on ffmpegs trac if they arent there
[01:21] <michaelni> that way we can keep trak of them and i can fix them when i find the time
[02:33] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r03ce421c13 10ffmpeg/libavcodec/8svx.c: 
[02:33] <CIA-119> ffmpeg: 8svx: fix crash
[02:33] <CIA-119> ffmpeg: Fixes Ticket1377
[02:33] <CIA-119> ffmpeg: Found-by: Piotr Bandurski <ami_stuff at o2.pl>
[02:33] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:33] <CIA-119> ffmpeg: 03Piotr Bandurski 07master * r59352cc219 10ffmpeg/libavcodec/sgienc.c: 
[02:33] <CIA-119> ffmpeg: sgienc: add a limit for maximum supported resolution
[02:33] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:21] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rabf5f6ea2a 10ffmpeg/tests/md5.sh: 
[04:21] <CIA-119> ffmpeg: fate: try to fix 3rd md5 implementation
[04:21] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[07:28] <ubitux> Compn: the file isn't independant
[07:29] <ubitux> you can't just merge easily gpl and lgpl code :p
[08:07] <CIA-119> ffmpeg: 03Clément BSsch 07master * r64df092b25 10ffmpeg/libswresample/swresample.c: swr: fix a few typo.
[09:48] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * raac41ec119 10ffmpeg/configure: 
[09:48] <CIA-119> ffmpeg: Let ffmpeg select the setpts filter.
[09:48] <CIA-119> ffmpeg: Do not disable ffmpeg with --disable-everything.
[12:20] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * re6866b1c67 10ffmpeg/tests/ (fate/acodec.mak ref/fate/acodec-ra144): 
[12:20] <CIA-119> ffmpeg: fate: only check stddev for acodec-ra144
[12:20] <CIA-119> ffmpeg: ra144 uses floats so bitexactness cannot be guranteed
[12:20] <CIA-119> ffmpeg: This should fix a long standing issue with icc
[12:20] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:20] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r422418b664 10ffmpeg/libavcodec/dnxhdenc.c: 
[12:20] <CIA-119> ffmpeg: dnxhd: replace asserts by av_asserts
[12:20] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[12:35] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r23b365f1c4 10ffmpeg/ffmpeg.c: 
[12:35] <CIA-119> ffmpeg: ffmpeg: fix memleak
[12:35] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:27] <ubitux> TimNich: ping
[14:54] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rf89aa6751c 10ffmpeg/libavformat/avio.c: 
[14:54] <CIA-119> ffmpeg: lavf: use av_freep() in ffurl_close()
[14:54] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:54] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r5f509fa2e1 10ffmpeg/libavutil/ (avutil.h log.h): 
[14:54] <CIA-119> ffmpeg: log: add category to AVClass
[14:54] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:54] <CIA-119> ffmpeg: 03Martin Storsjö 07master * r2dbb8cc6a3 10ffmpeg/libavformat/tcp.c: 
[14:54] <CIA-119> ffmpeg: tcp: Check the return values from bind and accept
[14:54] <CIA-119> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[14:54] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * redf96d631c 10ffmpeg/libavformat/ (avio.c url.h): 
[14:54] <CIA-119> ffmpeg: url: add ffurl_closep() which also sets the context pointer to NULL
[14:54] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:54] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rf218121a4d 10ffmpeg/libavformat/http.c: 
[14:54] <CIA-119> ffmpeg: http: use ffurl_closep()
[14:54] <CIA-119> ffmpeg: Fixes Ticket1383
[14:54] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:59] <Compn> ubitux : make it gpl de/encoder file then
[14:59] <Compn> :P
[15:01] <ubitux> huh
[15:01] <ubitux> why not but well&
[15:01] <ubitux> ideally we would need to keep a copy under lgpl
[15:01] <ubitux> and when gpl is activated, enable the "complete" version
[15:02] <ubitux> not sure it's that reasonnable
[15:02] <iive> what is the problem?
[15:03] <ubitux> http://ffmpeg.org/pipermail/ffmpeg-devel/2012-May/125284.html
[15:08] <ubitux> rhaa i don't get it, fate fails once, and then work again
[15:08] <ubitux> wtf is wrong.
[15:11] <Compn> ubitux : the trick is to commit gpl version and have companies pay for lgpl version :)
[15:11] <ubitux> :)
[15:19] <TimNich> ubitux: pong, (back from lunch...)
[15:20] <ubitux> TimNich: i've an experimental branch for you to test the mov timecode :p
[15:20] <ubitux> TimNich: https://github.com/SmartJog/ffmpeg/tree/timecode
[15:20] <ubitux> can you try this?
[15:21] <ubitux> also note the timecodes won't be transmuxed/transcoded from mov to mov yet
[15:21] <ubitux> since the timecode meta is in the tmcd track
[15:21] <TimNich> Great, I'm not sure if I'll have time today now, rebuilding and debugging an LTFS box
[15:21] <ubitux> and not in the video stream being converted, and not in the format either
[15:22] <TimNich> 'll take a look as soon as I can.
[15:22] <ubitux> ideally i should also insert the timecode meta in the related video streams in the mov demuxer
[15:22] <ubitux> ok :)
[15:31] <iive> to be honest, I haven't seen baptiste in ffmpeg for quite some time.
[15:32] <av500> maybe he is in #ffmbc
[16:01] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rb6fdf8dea7 10ffmpeg/libavcodec/yop.c: 
[16:01] <CIA-119> ffmpeg: yopdec: check that palette fits in the packet
[16:01] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:01] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rfebc013dc5 10ffmpeg/libavcodec/yop.c: 
[16:01] <CIA-119> ffmpeg: yopdec: check frame oddness to be within supported limits
[16:01] <CIA-119> ffmpeg: Fixes Ticket1365
[16:01] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:20] <ubitux> michaelni: i think there is an issue in how the output aformat is inserted
[16:20] <ubitux> it should honor the last filter output settings instead of the input
[16:21] <ubitux> and that's why it currently fails with the mapchannel test
[16:21] <ubitux> the mapchannel test is basically saying:
[16:21] <ubitux> this is a 5.1 input, pick only two (ch1 and ch0), and then downmix to mono (-ac 1)
[16:22] <ubitux> but the aresample is doing this: [aresample @ 0xa905260] chl:5.1 fmt:s16 r:22050Hz -> chl:mono fmt:s16 r:22050Hz
[16:22] <ubitux> even thought the pan filter (which should be inserted before) is outputing stereo
[16:27] <ubitux> michaelni: also, no way to fix the fps auto insert?
[16:27] <ubitux> (breaking too much things?)
[16:29] <michaelni> ubitux, why do you want to autio insert the fps filter ? is there some feature req or bug this fixes ?
[16:30] <ubitux> i think it's a good thing to move a lot of stuff to lavfi and lighten a bit ffmpeg.c
[16:31] <ubitux> why not doing it?
[16:32] <michaelni> vf fps is buggy ...
[16:33] <michaelni> and not easy to fix ...
[16:34] <michaelni> it will become easier once some patches from IIRC nicolas are in lavfi
[16:34] <ubitux> ah it's the poll & request stuff again
[16:37] <ubitux> where is the dup & drop done ATM when -r is specified?
[16:37] <ubitux> (it's directly related to -r, right?)
[16:39] <michaelni> dup /drop is done in ffmpeg search for "dup!"
[16:41] <ubitux> ah right, and indeed that code moved out
[16:41] <ubitux> is it broken in libav too?
[16:42] <ubitux> or it's broken only in ffmpeg because we need to unify a bit the apis?
[16:42] <michaelni> vf_fps ? yes its broken in libav too
[16:42] <ubitux> ok
[16:43] <michaelni> but note the patches ive meant where about frame durations not pull
[16:43] <michaelni> not saying there are no bugs with pull too ...
[16:45] <ubitux> what are the biggest issues? and how nicolas work will help?
[16:45] <TimNich> ubitux:  Had a quick look, in between other things. Timecode track duration looks good in QT player and it counts properly now...
[16:45] <michaelni> and iam not sure the patches even exist outside my imagination :)
[16:45] <ubitux> TimNich: ah? nice! :)
[16:47] <michaelni> ubitux, basically AVFilterBufferRef needs a duration field 
[16:48] <michaelni> someone should add that
[16:52] <matty3269> Hi, quick question... after I have a packet from av_frame_read I do a av_packet_dup, once it has been coped to a AVPacketList struct. Do I then need to free both packets with av_free_packet?
[16:52] <matty3269> I am getting a double free if do that
[16:52] <av500> no
[16:52] <av500> only free the one you dup'ped
[16:52] <av500> or I am wrong
[16:53] <TimNich> ubitux:  Looking with ffprobe I notice that ffmbc  rewrites the creation_time, but ffmpeg carries the original forward, which I guess is the correct  result of the other changes you have made...
[16:53] <matty3269> ah, so I don't need to free the packet from av_read_frame?
[16:53] <matty3269> and that won't leak memory either?
[16:54] <av500> I think you do have to
[16:54] <matty3269> hmm, strange, when I do that, glibc complains about a double free.
[16:55] <ubitux> < michaelni> ubitux, basically AVFilterBufferRef needs a duration field // ok :)
[16:55] <ubitux> TimNich: i'm not sure to understand; something is wrong with the creation_time?
[16:58] <TimNich> ubitux: http://pastebin.com/Etg3tWwH  compare the creation time, also the ffmpeg version has SAR/DAT twice....
[16:59] <matty3269> so, should this kind of code be okay? http://pastebin.com/VjHhDwG9
[17:00] <ubitux> TimNich: first probe is ffmbc?
[17:01] <ubitux> ah right, ffmbc.mov.
[17:01] <TimNich> "encoder FFmbc 0.7" line above Duration as well
[17:02] <ubitux> the creation_time stuff shouldn't be related to the timecode patchset
[17:02] <ubitux> or is it a regression?
[17:03] <TimNich> no but what about moving the general metadata handling changes
[17:03] <ubitux> the SAR/DAR output is indeed different, that doesn't sound like it is related to what i did either, more sth like the print utils in lavf or something
[17:04] <TimNich> not even the same figures, but why only printed twice for ffmeg?
[17:04] <TimNich> if its in ffprobe code
[17:04] <ubitux> ah wait indeed it's the same tool
[17:04] <ubitux> beh.
[17:05] <ubitux> anyway i can't make fate running properly now, i'll submit that patchset later
[17:05] <ubitux> < TimNich> no but what about moving the general metadata handling changes // sorry i don't understand what you mean
[17:07] <TimNich> \well you were proposing moving the timecode muxer stuff to the metadata handler, so I was wondering if something crept into the metadata handling code by mistake which has upset other metadata
[17:09] <ubitux> i may have broken something, i'll look into this
[17:09] <ubitux> thank you for your tests :)
[17:13] <TimNich> I have just run the same script on c0c2424 and got the same probe output  (creation time and double SAR/DAR) so thats not your code.
[17:29] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r0766b6e3ec 10ffmpeg/libavcodec/rv10.c: 
[17:29] <CIA-119> ffmpeg: rv20: fix lowres out of array read
[17:29] <CIA-119> ffmpeg: Fixes Ticket1239
[17:29] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:21] <TimNich> ubitux:  I have gone back as far as 2a976de and it still has the regression so I don't think its your problem, haven't managed to bisect good yet and I'm off for a few days (God bless  Her Majesty)
[18:37] <ubitux> heh ok :)
[18:37] <ubitux> great then, thanks for looking into it
[18:41] <michaelni> ubitux, i think ive fixed the mapchan test
[18:41] <ubitux> oh?
[18:41] <michaelni> just moved the filter insertion up :)
[18:41] <ubitux> heh ok :D
[18:42] <ubitux> before the aformat?
[18:44] <michaelni> yes
[18:44] <michaelni> ubitux, ok to commit or should i post a patch ?
[18:45] <ubitux> no no go ahead :p
[18:47] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rd43812e235 10ffmpeg/ffmpeg.c: 
[18:47] <CIA-119> ffmpeg: ffmpeg: move pan filter insertion up.
[18:47] <CIA-119> ffmpeg: This fixes one (disabled) fate test.
[18:47] <CIA-119> ffmpeg: Thanks-to: ubitux to pointing at the odd filter order and stereo/mono messup
[18:47] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:47] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r9758b1cb88 10ffmpeg/tests/fate/mapchan.mak: 
[18:47] <CIA-119> ffmpeg: fate: enable fate-mapchan-6ch-extract-2-downmix-mono
[18:47] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:47] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r461030a7ac 10ffmpeg/libavcodec/dct-test.c: 
[19:47] <CIA-119> ffmpeg: dct-test: perform emms before gettime().
[19:47] <CIA-119> ffmpeg: This fixes the hypothetical case of a gettime implementation using floats
[19:47] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:47] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r0d83edaba9 10ffmpeg/libavcodec/dct-test.c: 
[19:47] <CIA-119> ffmpeg: dct-test: give emms a memory clober.
[19:47] <CIA-119> ffmpeg: Without this, the compiler in theory could attempt to keep
[19:47] <CIA-119> ffmpeg: things in float registers across the emms which would cause
[19:47] <CIA-119> ffmpeg: bad things to happen.
[19:47] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:05] <CIA-119> ffmpeg: 03Oka Motofumi 07master * r5c742005fb 10ffmpeg/libavformat/avisynth.c: 
[20:05] <CIA-119> ffmpeg: avisynth: Make sure the filename passed to avisynth is in the right code page
[20:05] <CIA-119> ffmpeg: avisynth is a non-unicode application and cannot accept UTF-8
[20:05] <CIA-119> ffmpeg: characters. Therefore, the input filename should be converted to
[20:05] <CIA-119> ffmpeg: the correct code page that it expects.
[20:05] <CIA-119> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[20:05] <CIA-119> ffmpeg: 03Martin Storsjö 07master * r641f4a885f 10ffmpeg/libavformat/tcp.c: 
[20:05] <CIA-119> ffmpeg: tcp: Check the return values from bind and accept
[20:05] <CIA-119> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[20:05] <CIA-119> ffmpeg: 03Martin Storsjö 07master * r3cbcfa2dec 10ffmpeg/libavformat/http.c: 
[20:05] <CIA-119> ffmpeg: http: Clear the old URLContext pointer when closed
[20:05] <CIA-119> ffmpeg: This fixes issues with opening http urls that have authentication
[20:05] <CIA-119> ffmpeg: or redirects, introduced in commit e999b641.
[20:05] <CIA-119> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[20:05] <CIA-119> ffmpeg: 03Martin Storsjö 07master * r5952564185 10ffmpeg/libavformat/http.c: 
[20:05] <CIA-119> ffmpeg: http: Simplify code by removing a local variable
[20:11] <CIA-119> ffmpeg: 03Carl Eugen Hoyos 07master * r7980a260a9 10ffmpeg/configure: 
[20:11] <CIA-119> ffmpeg: Let ffmpeg select the null and anull filters.
[20:11] <CIA-119> ffmpeg: They are needed since 16e25ef.
[21:12] <CIA-119> ffmpeg: 03Clément BSsch 07master * r2248db946c 10ffmpeg/ffprobe.c: ffprobe: move packets_and_frames work to writer context.
[21:12] <CIA-119> ffmpeg: 03Clément BSsch 07master * rdb839544d6 10ffmpeg/ffprobe.c: ffprobe/xml: use writer multiple_sections instead of local context variable.
[21:12] <CIA-119> ffmpeg: 03Clément BSsch 07master * rfd0c83c666 10ffmpeg/ (doc/ffprobe.texi ffprobe.c): ffprobe: add flat output format.
[22:27] <silverrocker> how is half pixel interpolation perfomed when H.264 is processed at the border of the frames?
[23:25] <ubitux> mmh for the sad() i need to run, since i don't "need" to process the picture in mb, is it interesting (from a performance PoV) to process the image as is anyway?
[23:25] <ubitux> or a linear processing is ok?
[23:26] <ubitux> also, can i assume DSPContext.sad[0] will always process 8B aligned data?
[23:28] <kierank> there should be a list of assumptions in dsputil.c
[23:28] <burek> does ffmpeg support input from usb web cameras in h264 mode?
[23:28] <ubitux> (i mean sad[1] sorry)
[23:29] <burek> something like: ffmpeg -f v4l2 -vcodec h264 -i /dev/video0 ...
[23:29] <ubitux> kierank: i don't see much thing for sad[]
[23:29] <kierank> that's a shame
[23:29] <ubitux> basically all i see is c->sad[0]= pix_abs16_c and c->sad[1]= pix_abs8_c
[23:33] <Daemon404> burek, i dont think so
[23:33] <Daemon404> but i could be wrong
[23:36] <burek> Daemon404, I've compiled latest vlc today and it does work with my h264 web cam
[23:36] <Daemon404> vlc has a ton of it own stuff though
[23:36] <burek> so I just wanted to ask is it some kind of a new feature or I don't know how to ask ffmpeg to do the same
[23:37] <Daemon404> i remember v4l2 supporting mjpeg, but not h264
[23:37] <Compn> h264 webcam, isnt it usually streamed over rtmp or something ?
[23:37] <Compn> or rtsp
[23:38] <Compn> i dont think v4l2 has h264 either...
[23:38] <burek> no
[23:38] <burek> using v4l2, at least with vlc
[23:39] <burek> I run it like: vlc v4l2:///dev/video0:chroma=h264
[23:39] <burek> so I thought it maybe might be possible with ffmpeg, using: ffmpeg -f v4l2 -vcodec h264 -i /dev/video
[23:39] <burek> but it's not
[23:49] <Compn> http://forum.videolan.org/viewtopic.php?f=13&t=87677
[23:49] <Compn> ah, guess its in there
[23:49] <Compn> probably ffmpegs' v4l2 wrapper doesnt support it ?
[23:49] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * rf23a2418fb 10ffmpeg/libavcodec/wnv1.c: 
[23:49] <CIA-119> ffmpeg: wmv1: check that the input buffer is large enough
[23:49] <CIA-119> ffmpeg: Fixes null ptr deref
[23:49] <CIA-119> ffmpeg: Fixes Ticket1367
[23:49] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:49] <CIA-119> ffmpeg: 03Michael Niedermayer 07master * r0bae6661cd 10ffmpeg/libavcodec/fraps.c: 
[23:49] <CIA-119> ffmpeg: fraps: fix version 0/1 input data size check.
[23:49] <CIA-119> ffmpeg: Fixes array overread.
[23:49] <CIA-119> ffmpeg: Fixes Ticket1371
[23:49] <CIA-119> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Sat Jun  2 2012