[Ffmpeg-devel-irc] ffmpeg-devel.log.20120720

burek burek021 at gmail.com
Sat Jul 21 02:05:02 CEST 2012


[01:57] <durandal_1707> michaelni: why our caf demuxer reads tag in le order instead of be ?
[01:58] <Freakshow> guys, where are the structs for avlog?
[01:59] <durandal_1707> libavutil/log ?
[01:59] <Freakshow> ah, thanks... just looking there
[02:00] <durandal_1707> michaelni: i mean it is not consistent because other stuff use MKBETAG so it is broken
[02:06] <durandal_1707> hmm, ffplay does not support planar sample format?
[02:07] <Compn> its an anti-alac conspiracy!
[02:07] Action: Compn pokes durandal_1707
[02:07] <Compn> ehe
[02:08] <durandal_1707> Compn: as of recentlt mplayer does not play alac any more
[02:12] <durandal_1707> michaelni: i get segv with AV_SAMPLE_FMT_S32_to_AV_SAMPLE_FMT_FLT () 
[02:17] <durandal_1707> but crash happens with ffplay only ???
[02:22] <J_Darnley> Hello.  I thought I'd better let you all know that you've got a spelling error in your news item about SOCIS: "partecipate" should be "participate".
[02:25] <J_Darnley> That's all.  Bye.
[02:44] <Freakshow> anyone know where the structs for rtsp and rtmp are defined?
[02:44] <Freakshow> durandal_1707 possibly?
[02:44] <Freakshow> :D
[02:44] <durandal_1707> somewhere in libavformat?
[02:45] <Freakshow> basically looking into https://ffmpeg.org/trac/ffmpeg/ticket/1446
[02:45] <durandal_1707> i cant help you (i don't do protocols)
[02:46] <Freakshow> no worries... 
[02:47] <Compn> all the rtsp/rtmp is in libavformat yep
[03:35] <michaelni> durandal_1707, fixed ffplay locally, ill commit after a few more tests
[03:45] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rcdafec69ce 10ffmpeg/ffplay.c: 
[03:45] <CIA-41> ffmpeg: ffplay: fix planar audio
[03:45] <CIA-41> ffmpeg: This caused ALAC to fail since the last merge
[03:45] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:14] <durandal_1707> hmm i get wrong sound wiht ffmpeg -f oss and 5.1 sound
[04:17] <Compn> you sure run into a lot of bugs
[04:20] <durandal_1707> Compn: new stuff breaks other stuff (and i start running in circles) ...
[04:23] <durandal_1707> one bug less (this one is pretty old)
[04:54] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * re2c96fbf3e 10ffmpeg/ffplay.c: 
[04:54] <CIA-41> ffmpeg: ffplay: fix warning that i just mistakenly introduced
[04:54] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:54] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r5e99df019a 10ffmpeg/libavcodec/alac.c: 
[04:54] <CIA-41> ffmpeg: alacdec: several players have problems with planar audio still thus temporary put packed output back by default
[04:54] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:57] <michaelni> Compn, mplayer should work with alac again
[04:58] <Compn> ok
[08:16] <nevcairiel> imho if some player uses git ffmpeg it should also manage to adapt to changing requirements or accept breaking
[08:18] <ohsix> before there were releases that's all there was, and you can see how that argument is sillt :]
[08:18] <nevcairiel> trying to support incomplete players with compat things like this is just bad
[08:19] <nevcairiel> well there are frequent releases now
[08:20] <truexfan81> are we talking about a specific media player?
[08:20] <nevcairiel> mplayer i guess
[08:21] <truexfan81> i do git pulls from ffmpeg, and recompile mplayer once a week, i've never had a problem, doing right now actually
[08:23] <nevcairiel> the big problem with compat things like this is that its quite possible that developers will simply not notice that they need to support planar audio and you need to apply more and more of these compat hacks
[08:24] <ohsix> too bad they didn't notice huh
[08:24] <truexfan81> unless it applies them automatically during the make, i'm not aware of it using any hacks
[08:25] <truexfan81> rot ro, just encountered a problem lol
[08:29] <truexfan81> see if this fixes it, sometimes its easier to just delete the ffmpeg folder, and let the mplayer configure script git clone it again
[08:30] <ohsix> git reset heh
[08:30] <ohsix> save everyone some bandwidth
[08:30] <ohsix> mplayer has releases too ...
[08:30] <truexfan81> fail, same error
[08:37] <truexfan81> and its broke lol
[08:37] <truexfan81> and stupid me, i already did the sudo make uninstall lol
[08:38] <truexfan81> this is where it fails http://pastebin.com/CyUdS449
[08:40] <ohsix> are there features in the mplayer git version you can't do without?
[08:40] <truexfan81> huh? the mplayer configure script clones the latest, so i don't have a older version to go back to
[08:41] <ohsix> and your distro doesn't? i'm not sure why you even want to be building it
[08:42] <truexfan81> i noticed a playback performance increase when compiling it myself
[08:42] <ohsix> files you couldn't play before on that machine became playable?
[08:43] <truexfan81> 1 or 2 yes, but mostly everything plays more smoothly, also with the version in the distro i had video tearing issues that when away when i compiled it
[08:43] <truexfan81> went*
[08:45] <truexfan81> for kicks, i'm compiling ffmpeg without mplayer, just to see what happens
[08:47] <truexfan81> i'm tempted to try mplayer2 just to see if it does any better
[08:47] <ohsix> eh i wouldn't bother unless you find some objective measurement
[08:48] <truexfan81> or maybe see if i can get vlc to compile with vdpau support?
[09:28] <truexfan81> well for now i have a working media player, using the version of mplayer2 from the ubuntu repos
[10:36] <CIA-41> ffmpeg: 03Nicolas George 07master * rbf8ce5a5c3 10ffmpeg/ffmpeg.c: 
[10:36] <CIA-41> ffmpeg: ffmpeg: probe buffersinks once more after EOF.
[10:36] <CIA-41> ffmpeg: Reap buffers stored in the buffer sinks even when
[10:36] <CIA-41> ffmpeg: the avfilter_graph_request_oldest() loop only returns EOF.
[10:36] <CIA-41> ffmpeg: avfilter_graph_request_oldest() can cause frames to arrive
[10:36] <CIA-41> ffmpeg: to the sinks even when it returns EOF.
[11:29] <CIA-41> ffmpeg: 03Nicolas George 07master * re62fd6619f 10ffmpeg/ (6 files in 2 dirs): 
[11:29] <CIA-41> ffmpeg: oggdec: add support for Opus codec.
[11:29] <CIA-41> ffmpeg: This patch also introduces CODEC_ID_OPUS.
[11:29] <CIA-41> ffmpeg: 03Nicolas George 07master * ra6cf296bd9 10ffmpeg/ (6 files in 2 dirs): lavc: Opus decoder using libopus.
[11:29] <CIA-41> ffmpeg: 03Nicolas George 07master * r30549294ef 10ffmpeg/libavformat/matroska.c: 
[11:29] <CIA-41> ffmpeg: matroska: add support for Opus.
[11:29] <CIA-41> ffmpeg: Opus in Matroska is still a draft.
[11:45] <CIA-41> ffmpeg: 03Nicolas George 07master * rb34ce5a26b 10ffmpeg/doc/faq.texi: faq: explain dependency problems with static libs.
[12:27] <rohankasat> hi 
[12:28] <rohankasat> can we stream .pls or .m3u ( playlist files) directly from console using ffplay ?
[13:00] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r3bd3e46c91 10ffmpeg/doc/avtools-common-opts.texi: doc/avtools-common-opts: fix a few typos in the "Stream specifiers" section
[13:15] <burek> where did video encoders dissapear from html docs?
[13:16] <burek> there is just 13. Encoders
[13:16] <burek> in ffmpeg.html
[13:33] <CIA-41> ffmpeg: 03Nicolas George 07master * rb0629366a2 10ffmpeg/ (5 files in 3 dirs): 
[13:33] <CIA-41> ffmpeg: lavf: make cmdutils's check_stream_specifier public.
[13:33] <CIA-41> ffmpeg: The new function is called avformat_match_stream_specifier()
[13:33] <CIA-41> ffmpeg: and does not include logging the error message.
[13:33] <CIA-41> ffmpeg: 03Nicolas George 07master * r97f86680eb 10ffmpeg/libavfilter/ (buffer.c buffersrc.c internal.h): lavfi: make copy_buffer_ref common.
[13:33] <CIA-41> ffmpeg: 03Nicolas George 07master * r461f506f7b 10ffmpeg/libavcodec/utils.c: lavc: set best_effort_timestamp for audio too.
[13:33] <CIA-41> ffmpeg: 03Nicolas George 07master * rcb6a20fa56 10ffmpeg/libavfilter/internal.h: lavfi: add FF_*LINK_IDX convenience macros.
[13:38] <rohankasat>  can we stream .pls or .m3u ( playlist files) directly from console using ffplay ? or how can we parse it , and find the streaming url's or streaming link's inside the playlist file ?
[13:39] <Tjoppen> .. you can parse it using a small shell script
[14:18] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r7c6ebe2b97 10ffmpeg/libavcodec/alac.c: 
[14:18] <CIA-41> ffmpeg: alac: drop packed sample output support with the next major bump
[14:18] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:04] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * ra4f94f24b4 10ffmpeg/libavformat/udp.c: 
[15:04] <CIA-41> ffmpeg: udp: fix compilation when HAVE_PTHREAD_CANCEL isnt defined
[15:04] <CIA-41> ffmpeg: This happens with for example mplayer.
[15:04] <CIA-41> ffmpeg: Fixing it in ffmpeg allows new ffmpeg to be compiled with older mplayer
[15:04] <CIA-41> ffmpeg: which would not be possible if the fix was just in mplayer.
[15:04] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:26] <ubitux> http://ffmpeg.org/pipermail/ffmpeg-devel/2012-July/127981.htmlhttp://archive.birdhouse.org/etc/kibosig.txt :)
[15:29] <CIA-41> ffmpeg: 03Joseph Artsimovich 07master * r99c4e91dfa 10ffmpeg/ (3 files in 2 dirs): (log message trimmed)
[15:29] <CIA-41> ffmpeg: dnxhd: Fix 10-bit DNxHD quant matrices
[15:29] <CIA-41> ffmpeg: Convert them to zigzag order, as the rest of them are.
[15:29] <CIA-41> ffmpeg: When I was adding support for 10-bit DNxHD, I just copy-pasted the
[15:29] <CIA-41> ffmpeg: missing quant matrices from the spec. Now it turns out the existing
[15:29] <CIA-41> ffmpeg: matrices in dnxhddata.c were in zigzag order. This resulted in wrong
[15:29] <CIA-41> ffmpeg: quantization for 10-bit DNxHD. The attached patch fixes the problem by
[15:32] <nevcairiel> ah sweet, i had someone complain about 10-bit DNxHD decoding problems, but never got around to investigating, wonder if he is happy now
[17:17] <CIA-41> ffmpeg: 03chinshou 07master * r55de89e25d 10ffmpeg/cmdutils.c: 
[17:17] <CIA-41> ffmpeg: cmdutils/alloc_buffer: fix typo
[17:17] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:17] <juanmabc> hi, #ffmpeg seems of no help outside of command line usage, so since you develop the api itself, i guess you could know what are the keys to be data aligned apart from using av_maloc/av_free, like some DECLARE_ALIGNED or whatever needed, etc. reference code in src/mlMedia.c: http://code.google.com/p/openmedialibrary 
[18:22] <juanmabc> http://fpaste.org/c2s8/
[18:25] <juanmabc> the codec capabilities of your library is awesome btw, though api lack examples outside of ffplay SDL.
[19:02] <Compn> juanmabc : also feel free to ask on libav-users list, its the list for 3rd party ffmpeg/libav libraries
[19:02] <Compn> development that is
[19:02] <Compn> probably you wont get much c++ help from this channel tho
[19:02] <Compn> a number of devels dislike c++ :)
[19:03] <durandal_1707> it is just that lav* is written in C
[19:10] <juanmabc> Compn: it is plain c, the if _cplusplus is not meet, it was for windows compiler to get built sine the c one does not support std99
[19:11] <juanmabc> so i'm at the same language that ffmpeg itself, gcc (not g++)
[19:16] <Compn> ah
[19:16] <Compn> well most people are asleep then ;)
[19:22] <juanmabc> ok :D
[19:25] <gnafu> juanmabc: A fellow fpaste user.  Hello!
[19:26] Action: gnafu can't be of any help, since he's not a programmer, but wanted to give a greeting.
[19:45] <juanmabc> my bug should be beetwen lines 218-222
[19:46] <juanmabc> gonna check again ffplay.c meanwhile :D
[19:50] <juanmabc> and the bug rises in 408
[19:50] <juanmabc> 2 options: media->video.frm->data is not aligned or  media->video.convfrm->data is not, src or dest
[19:52] <nevcairiel> you can easily test if a pointer is aligned, just look at its address
[19:52] <nevcairiel> but av_malloc should always produce 32-byte alignment on recent builds
[19:52] <juanmabc> media->video.frm is = avcodec_alloc_frame(); nothing done directly to data, it fills that with avcodec_decode_video2
[19:52] <juanmabc> nevcairiel: how this test is done?
[19:53] <juanmabc> 0.10.4 is *recent build* ?
[19:55] <juanmabc> media->video.convfrm = avcodec_alloc_frame();  that does an avpicture_fill with an av_malloc buffer, more work in this so likely to be the one
[19:56] <nevcairiel> if you think a pointer is not aligned, just test if it is, print its address or something like that, easy enough to spot alignment
[19:56] <juanmabc> let me see
[19:57] <juanmabc> frm->data: 0x8dd99a0, convfrm->data: 0x8daa280
[19:58] <juanmabc> funny is i don't ever play with data directly, the most near by me is an avpicture_fill to convfrm
[19:58] <juanmabc> play = alloc or something
[19:58] <nevcairiel> more interesting would be the first member inside the data array
[19:59] <nevcairiel> the question is however, do you actually have any problems you're trying to solve here, you didnt specify
[20:00] <juanmabc> frm->data[0]: 0x8b63b20, convfrm->data[0]: 0x8b25a40 (i use &frm->data[0], &convfrm->data[0])
[20:00] <juanmabc> yes this, [swscaler @ 0x8b0fdc0] Warning: data is not aligned! This can lead to a speedloss
[20:01] <nevcairiel> those addresses are at least 16-byte aligned
[20:01] <nevcairiel> not sure what swscale checks for
[20:02] <nevcairiel> it also checks the image stride for alignment
[20:02] <nevcairiel> depending on what you specify there
[20:03] <nevcairiel> check that you set the target linesize to something at least mod16
[20:05] <juanmabc> funny to mention it, i don set convfrm->linesize nowhere
[20:06] <nevcairiel> it should be set to something
[20:07] <nevcairiel> usually a mod16 or mod32 aligned version of the width
[20:10] <juanmabc> convfrm->linesize[0] is 1536 any attempt to modify that does not show more than black
[20:11] <_Nike_> please anyone here? I'm going crazy now.
[20:11] <_Nike_> :S
[20:11] <juanmabc> nevcairiel: wich is 16 * 16 * 4
[20:11] <nevcairiel> sounds fine
[20:11] <juanmabc> er, *6 not *4
[20:12] <juanmabc> the address it reports is near convfrm so i'm gonna guess is it
[20:13] <juanmabc> and the only place i don't know what it does is in avpicture_fill
[20:13] <juanmabc> line 221
[20:13] <juanmabc> http://fpaste.org/c2s8/
[20:14] <CIA-41> ffmpeg: 03Paul B Mahol 07master * ra2dab7512e 10ffmpeg/libavcodec/exr.c: 
[20:14] <CIA-41> ffmpeg: exr: compression attribute is mandatory
[20:14] <CIA-41> ffmpeg: Do not continue decoding if one is missing.
[20:14] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[20:14] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r01f76a779c 10ffmpeg/libavcodec/exr.c: 
[20:14] <CIA-41> ffmpeg: exr: merge common code
[20:14] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[20:14] <CIA-41> ffmpeg: 03Paul B Mahol 07master * rb040ffc84c 10ffmpeg/libavcodec/exr.c: 
[20:14] <CIA-41> ffmpeg: exr: display warning if multiple compression attributes are found
[20:14] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[20:14] <juanmabc> mmm, does the quantity to av_malloc need to be mod16?
[20:15] <_Nike_> My program makes a TCP connection with the RTP, and I need to send the buffer that I get to ffmpeg to parse. Parse the H264 stream bytes. Because they are fragmented. I need to separate the NALS. How can I do this without FFMPEG connect in RTSP? If my program connects and receives packets.
[20:20] <mulicheng> Short of hacking ffmpeg sources, has nobody ever requested an standard API method that could return the error_count while decoding video?
[20:20] <mulicheng> int e=((MpegEncContext*)c->priv_data)->error_count
[20:20] <mulicheng> Would be handy to get feedback during decode if errors have happened.
[21:12] <juanmabc> so my final problem is frm->data[1]%16 and frm->data[2]%16 = 8
[21:15] <CIA-41> ffmpeg: 03Nicolas George 07master * rfd58678b86 10ffmpeg/libavdevice/v4l2.c: 
[21:15] <CIA-41> ffmpeg: v4l2: remove call to v4l2_fd_open.
[21:15] <CIA-41> ffmpeg: v4l2_fd_open is already called at the end of v4l2_open;
[21:15] <CIA-41> ffmpeg: calling a second time leaks structures inside libv4l2.
[21:15] <CIA-41> ffmpeg: Fix trac ticket #1560.
[21:36] <juanmabc> so all this leads to the avcodec_decode_video2 Note: "You might have to align the input buffer avpkt->data. The alignment requirements depend on the CPU: on some CPUs it isn't necessary at all, on others it won't work at all if not aligned and on others it will work but it will have an impact on performance."
[21:37] <nevcairiel> the input alignment has nothing to do with the output alignment
[21:38] <nevcairiel> the real question is why the data pointers are not aligned
[21:40] <juanmabc> i checked sws_scale test that logs that warning, my final problem is frm->data[1]%16 and frm->data[2]%16 = 8, that is direct from avcodec_decode_video2
[21:41] <juanmabc> all the internal stuff is just media->video.frm = avcodec_alloc_frame();
[21:41] <juanmabc> i let it be filled by decode, and sws_scale to convfrm
[21:43] <juanmabc> i got to go, i'll check later, thanks
[21:46] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r85761efa95 10ffmpeg/libavcodec/tiertexseqv.c: 
[21:46] <CIA-41> ffmpeg: tiertexseq: remove redundant string from log message
[21:46] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[23:37] <juanmabc> hi, unaligned guy again! (because i feel that way ;P)
[23:37] <juanmabc> so after that code: avcodec_decode_video2(media->video.cc, media->video.frm, &media->video.rdy, &media->video.pktprc);
[23:38] <juanmabc> media.video.frm->data[1] and [2] %16 are 8, so 8 aligned i guess and not 16 aligned
[23:38] <kierank> can you just not use the image alloc functions
[23:38] <juanmabc> i guess also that is G and B
[23:38] <juanmabc> me?
[23:38] <kierank> yes
[23:39] <juanmabc> alloc_frame?
[23:39] <juanmabc> care to explain? :D
[23:43] <kierank> there's an image alloc function
[23:43] <kierank> and you should use it
[23:43] <kierank> but that doesn't look like what you're trying to do anyway
[23:52] <juanmabc> ffmpeg 0.10.4 here, av_image_alloc is still not defined
[23:52] <juanmabc> what i'm trying to solve is : [swscaler @ 0x9f7ffa0] Warning: data is not aligned! This can lead to a speedlos
[23:54] <juanmabc> http://fpaste.org/c2s8/ following "media->video.frm" is not that hard, control+f it and see, and i know for sure data[1] and data[2] are the ones, since i did the sws_scale check code from trunk in place
[00:00] --- Sat Jul 21 2012


More information about the Ffmpeg-devel-irc mailing list