[Ffmpeg-devel-irc] ffmpeg-devel.log.20111128

Tue Nov 29 02:05:04 CET 2011

[00:41] <CIA-41> ffmpeg: 03Vitor Sessak 07master * rca55606a51 10ffmpeg/libavutil/x86/x86inc.asm: 
[00:41] <CIA-41> ffmpeg: x86inc: Flag shufps as an floating-point instruction for the AVX emulation code.
[00:41] <CIA-41> ffmpeg: Without this, code like "shufps m0, m1, m2, 0xaa" would not work in CPUs
[00:41] <CIA-41> ffmpeg: not supporting SSE2.
[00:41] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:14] <CIA-41> ffmpeg: 03Kostya Shishkov 07master * r498605b4ad 10ffmpeg/libavcodec/vc1dec.c: 
[01:14] <CIA-41> ffmpeg: vc1: select interlaced scan table by FCM element
[01:14] <CIA-41> ffmpeg: Interlaced videos can contain progressive frames too and now wrong scantable
[01:14] <CIA-41> ffmpeg: is selected for them.
[01:14] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[01:14] <CIA-41> ffmpeg: 03Victor Vasiliev 07master * r12bc20502a 10ffmpeg/libavformat/ (Makefile avi.c avi.h avidec.c avienc.c riff.c riff.h wav.c): 
[01:14] <CIA-41> ffmpeg: Generalize RIFF INFO tag support; support reading INFO tag in wav
[01:14] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[01:14] <CIA-41> ffmpeg: 03Alex Converse 07master * rf11b0e9543 10ffmpeg/libavcodec/wmavoice.c: wmavoice: Make format string match variable type.
[01:14] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rd9ba767d61 10ffmpeg/libavformat/mpc.c: 
[01:14] <CIA-41> ffmpeg: musepack: fix signed shift overflow in mpc_read_packet()
[01:14] <CIA-41> ffmpeg: Using an unsigned variable avoids problems with overflows.
[01:14] <CIA-41> ffmpeg: There is further no need for a 64-bit intermediate here.
[01:14] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[01:14] <CIA-41> ffmpeg: 03Aneesh Dogra 07master * r97980db487 10ffmpeg/libavcodec/indeo3.c: 
[01:14] <CIA-41> ffmpeg: indeo3: error out if no motion vector is set.
[01:14] <CIA-41> ffmpeg: This fixes a crash on a corrupt bitstream (bugzilla #93).
[01:14] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[01:14] <CIA-41> ffmpeg: 03Vitor Sessak 07master * r6b6ee58249 10ffmpeg/libavutil/x86/x86inc.asm: 
[01:14] <CIA-41> ffmpeg: x86inc: Flag shufps as an floating-point instruction for the AVX emulation code.
[01:14] <CIA-41> ffmpeg: Without this, code like "shufps m0, m1, m2, 0xaa" would not work in CPUs
[01:14] <CIA-41> ffmpeg: not supporting SSE2.
[01:14] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[01:14] <CIA-41> ffmpeg: 03Justin Ruggles 07master * r0df5e869cb 10ffmpeg/libavcodec/ (mpegaudiodec.c mpegaudiodec_float.c): 
[01:15] <CIA-41> ffmpeg: * qatar/master:
[01:15] <CIA-41> ffmpeg:  vc1: use an enum for Frame Coding Mode
[01:15] <CIA-41> ffmpeg:  doc: cleanup filter section
[01:15] <CIA-41> ffmpeg:  indeo3: error out if no motion vector is set.
[01:15] <CIA-41> ffmpeg:  x86inc: Flag shufps as an floating-point instruction for the AVX emulation code.
[01:15] <CIA-41> ffmpeg: 03Alex Converse 07master * r028a2375e2 10ffmpeg/libavformat/mov.c: mov: Make format string match variable type.
[01:15] <CIA-41> ffmpeg: 03Cheng Sun 07master * r3f5aa7dfa6 10ffmpeg/libavcodec/pthread.c: 
[01:15] <CIA-41> ffmpeg: pthread: track thread existence in a separate variable.
[01:15] <CIA-41> ffmpeg: This fixes a compile error on mingw32 when using p->thread
[01:15] <CIA-41> ffmpeg: directly (as if it were a pointer) to track thread existence,
[01:15] <CIA-41> ffmpeg: because the type is opaque and may be a non-pointer.
[01:16] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[01:16] <CIA-41> ffmpeg: 03Luca Barbato 07master * r9270b8a3d1 10ffmpeg/doc/filters.texi: 
[01:16] <CIA-41> ffmpeg: doc: cleanup filter section
[01:16] <CIA-41> ffmpeg: Use the @command{} tag when needed and cleanup the examples.
[03:44] <shifter1> anyone alive?
[03:45] <kcm1700> everyone is alive.
[03:45] <shifter1> that's good to hear
[03:45] <shifter1> I am having problems with ffserver, I think I know where they are coming from
[03:46] <shifter1> am I in the right place?
[03:52] <Compn> well bug reports should be reported in ffmpeg-user list or the ffmpeg bug tracker , unless you have a patch, then it can be discussed on ffmpeg-devel list or here
[04:01] <shifter1> Compn: I don't have a .diff file but I believe libavformat/sdp.c:405 just below  case CODEC_ID_AAC should be 'if (fmt && fmt->oformat && fmt->oformat->priv_class'
[04:02] <shifter1> this is in function sdp_write_media_attributes
[04:02] <Compn> well thats close enough :)
[04:02] <Compn> yeah, if you stick around, someone should pick up on it
[04:03] <shifter1> ffserver segfaults because fmt->oformat is NULL, and it tries to access fmt->oformat->priv_class
[04:03] <Compn> michaelni is always good at reviewing code 
[04:04] Action: Compn goes afk
[04:10] <michaelni> shifter1, does it work with your suggested change ?
[04:10] <shifter1> yes, the audio works, but IM recompiling now since I had it running in -g3 and -O0
[04:11] <shifter1> it was pretty laggy, but it didn't crash
[04:14] <shifter1> michaelni: audio works, video doesn't seem to, but that might be a problem with my ffserver config
[04:14] <shifter1> but it no longer crashes
[04:18] <michaelni> shifter1, the change looks harmless, ill apply it, what should i use as author ?
[04:18] Action: shifter1 blushes
[04:18] <shifter1> uh, cmastrange3 at gatech.edu ?
[04:19] <michaelni> ok
[04:19] <shifter1> carl if you need a first name
[04:19] <shifter1> "Carl"
[04:19] <shifter1> :P
[04:19] <shifter1> I've never been asked
[04:20] <michaelni> iam always trying to credit the author of a change, though very rarely i forget ...
[04:20] <shifter1> Well thanks, glad to have helped.  
[04:21] <Compn> shifter1 : be proud, you will soon have code in ffmpeg :P
[04:30] <CIA-41> ffmpeg: 03Carl 07master * r2cf4bd7751 10ffmpeg/libavformat/sdp.c: 
[04:30] <CIA-41> ffmpeg: sdp: Fix null pointer dereference with aac and ffserver.
[04:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rd3fc335bac 10ffmpeg/libavcodec/huffyuv.c: 
[04:30] <CIA-41> ffmpeg: huffyuvenc: support alphaless rgb32
[04:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb14a2381f8 10ffmpeg/libavutil/ (pixdesc.c pixfmt.h): 
[04:30] <CIA-41> ffmpeg: pixfmt: Add 32bit rgb without alpha formats
[04:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r0af7d7082f 10ffmpeg/libavcodec/huffyuv.c: 
[04:30] <CIA-41> ffmpeg: huffyuvenc: store alpha for bgr32
[04:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb89ce2d6af 10ffmpeg/ (6 files in 2 dirs): 
[04:30] <CIA-41> ffmpeg: swscale: alpha less rgb32 support.
[09:35] <Tjoppen> bcoudurier: http://code.google.com/p/ffmbc/issues/detail?id=49 look like decent opatom samples, and seem to work with the patches. thoughts?
[09:51] <Tjoppen> audio ones work if I add their essence container ul. for some reason it thinks there's two audio streams though
[09:53] <bcoudurier> ah yes, I should integrate this
[09:54] <Tjoppen> video seems to demux without a problem though
[09:54] <Tjoppen> going to investigate why two streams are created for the audio atoms
[09:57] <Tjoppen> only A1 triggers it
[09:59] <Tjoppen> ah, it's using the same sourpackageid for A1 and V1
[10:39] <Tjoppen> there. fixable by checking that data_definition_ul is consistent between packages
[10:40] <bcoudurier> nice
[10:41] <bcoudurier> Im going to bed, have a good day :)
[10:41] <Tjoppen> ok. I'll post on the ml
[10:42] <Tjoppen> ah, he left
[14:06] <CIA-41> ffmpeg: 03Clément BSsch 07master * rb1ca5634fd 10ffmpeg/libavcodec/ (avcodec.h mpeg12.c mpeg12enc.c options.c version.h): mpeg12: raise timecode to codec context.
[14:06] <CIA-41> ffmpeg: 03Clément BSsch 07master * rfbe6e29646 10ffmpeg/ffprobe.c: ffprobe: print codec timecode if available.
[15:15] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rae5873f13b 10ffmpeg/libavcodec/huffyuv.c: 
[15:15] <CIA-41> ffmpeg: huffyuvenc: switch from alphaless rgb32 to rgb24
[15:15] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:23] <kcm1700> is it worth optimizing code around libavcodec/mpeg4videodec.c:468 ?
[15:24] <kcm1700> it's get_amv function
[15:34] <ubitux> kcm1700: according to the comment, yes :p
[15:35] <ubitux> i guess a LUT could be used
[15:38] <kcm1700> that'd be nice. thanks
[16:53] <iive> i don't see how it could be done with look up table.
[16:53] <iive> the values are quite big.
[18:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r4b03d96022 10ffmpeg/libavcodec/ffv1.c: 
[18:10] <CIA-41> ffmpeg: ffv1dec: use PIX_FMT_0RGB32 when there is no transparency plane
[18:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:10] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rf7b160e829 10ffmpeg/libavcodec/ffv1.c: 
[18:10] <CIA-41> ffmpeg: ffv1enc: Store transparency plane.
[18:10] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:11] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r2027d073ae 10ffmpeg/libavcodec/ffv1.c: 
[18:11] <CIA-41> ffmpeg: ffv1dec: transparency plane support.
[18:11] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:11] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r492aab8589 10ffmpeg/libavcodec/ffv1.c: 
[18:11] <CIA-41> ffmpeg: ffv1enc: PIX_FMT_0RGB32 support
[18:11] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:22] <michaelni> kcm1700, the naive way to optimize the 16x16 summing is to do it recursive
[18:22] <michaelni> that is when 2 v>>shift values match than all points between them must match too
[18:23] <michaelni> so if you test the 4 corners and they match then you are done already just take the value *256
[18:24] <michaelni> if they mismatch smaller squares could be used
[18:24] <michaelni> its sad that mpeg didnt just use the analytical value of the integral, that would have been nicer
[18:25] <michaelni> but maybe theres a trick iam missing by which it can be calculated in a more pretty way ...
[18:32] <michaelni> and note, when trying the recursice thing there is surely a point where its faster to stop spliting
[18:33] <michaelni> another obvious way would be instead of doing things horizontal & vertical rather along lines that are close to constant along the plane
[20:03] <burek> should I send these spell-checked documents from Google Code In to some mailing list :)
[20:03] <burek> or should we just ignore it or what? :)
[20:58] <michaelni> burek, general.html ?
[20:59] <michaelni> its generated from  doc/general.texi, we need a patch against that 
[21:01] <michaelni> if you diff the file the student will give you against the html you should see the changes
[21:01] <michaelni> dunno which is the easiest way to integrate in the texi 
[21:02] <michaelni> if its not much then doing it by hand would be just a few minutes work
[21:02] <michaelni> the student should have worked on the texi :) ... i think he didnt but actually i didnt look at all
[21:02] <michaelni> i just looked at the list of tasks and saw general.html spellcheck claimed
[21:04] <burek> general, faq, ffmpeg, ffprobe, ffplay..
[21:04] <burek> its not much
[21:04] <burek> its just a spell check :)
[21:05] <burek> i can do that, just dont know how :)
[21:05] <michaelni> you dont know texi ?
[21:05] <michaelni> you just open it in a text editor of your choice
[21:05] <michaelni> it should be obvious
[21:05] <burek> ok :)
[21:05] <burek> ill figure it out :)
[21:06] <burek> is there any link to an original document or is it in the git?
[21:06] <michaelni> git
[21:06] <burek> ok, ill find it :)
[21:07] <michaelni> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=doc/general.texi
[21:08] <burek> thanks :)
[21:09] <burek> i just gave them tasks as easier as i could
[21:09] <burek> so that complete it, without a lot of trouble
[21:09] <burek> anyway, there are not much changes, so it will be ok :)
[21:11] <michaelni> i would have been surprised if there was alot that was misspelled in there
[21:12] <michaelni> though someone might probably reword alot in our docs 
[21:12] <michaelni> some of it could be worded more clearly
[21:14] <gnafu> michaelni: http://4.bp.blogspot.com/_D_Z-D2tzi14/S8TRIo4br3I/AAAAAAAACv4/Zh7_GcMlRKo/s400/ALOT.png ?
[21:14] <gnafu> ;D
[21:15] <michaelni> :)
[21:18] <pasteeater> burek: a generalized method to make a simple patch: http://pastebin.com/jJbUh1tb
[21:19] <burek> thanks pasteeater :)
[21:21] <pasteeater> there is also: http://git.videolan.org/?p=ffmpeg.git;a=blob_plain;f=doc/git-howto.txt;hb=HEAD
[21:21] <pasteeater> ...but it's quite verbose.
[21:32] <burek> thanks again :)
[23:33] <Compn> wait a minute
[23:33] <Compn> when did they start selling .xxx domains ??
[23:33] <gnafu> Compn: A while ago (by Internet standards).  Where have you been?  ;-)
[23:33] <ohsix> quite a while ago, they only started running ads on tv for them about a week ago though :> and they are pretty funny with the guy in the suit
[23:34] <Compn> damn, i'm under a rock
[23:35] <Compn> this month tho, so thats good
[23:35] <Compn> not like an entire year or anything
[23:57] <Anssi> michaelni: a bit of a late response, but no, don't really have any ideas...
[00:00] --- Tue Nov 29 2011