[Ffmpeg-devel-irc] ffmpeg-devel.log.20120117

burek burek021 at gmail.com
Wed Jan 18 02:05:03 CET 2012

[00:06] <CIA-31> ffmpeg: 03Michael Niedermayer 07release/0.9 * r63945e2226 10ffmpeg/libavcodec/indeo4.c: 
[00:06] <CIA-31> ffmpeg: indeo4: fix LE reader define for backport
[00:06] <CIA-31> ffmpeg: Fixes Ticket920
[00:06] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:17] <CIA-31> ffmpeg: 03Clément BSsch 07master * r72ad228ada 10ffmpeg/tests/Makefile: fate: add GEN tag for mapchan sw tests.
[00:42] <CIA-31> ffmpeg: 03Carl Eugen Hoyos 07master * rdc2f652ae4 10ffmpeg/libavcodec/imgconvert.c: 
[00:42] <CIA-31> ffmpeg: Fix pix_fmt_info for yuva420.
[00:42] <CIA-31> ffmpeg: This allows correct output pix_fmt auto-selection
[00:42] <CIA-31> ffmpeg: if the input pix_fmt is yuva420.
[00:42] <CIA-31> ffmpeg: Fixes ticket #290.
[00:42] <CIA-31> ffmpeg: 03Carl Eugen Hoyos 07master * r328e79329a 10ffmpeg/libavcodec/ (pnm.c pnmdec.c): Support gray8a pam decoding.
[00:42] <CIA-31> ffmpeg: 03Carl Eugen Hoyos 07master * r00430075bd 10ffmpeg/libavcodec/ (pnm.c pnmdec.c): Support 64bit pam decoding.
[00:55] <durandal_1707> unfortunately it looks like i would need to write new adpcm decoder, because THP adpcm variant (obviously match NGC_DSP in vgmstream) need something which brstm does not have
[00:57] <durandal_1707> michaelni: i have dumb -af patch for ffmpeg but i couldn't prove it is useful at all
[00:57] <ubitux> there is an existing -af patch in some lost branch
[00:57] <ubitux> you should ask stefano; i don't remember why it's not upstream
[00:58] <ubitux> (except there was some issues with it)
[01:02] <ubitux> it seems the commit isn't available anymore in saste's tree
[01:02] <ubitux> you may want to ask him about it
[01:10] <michaelni> durandal_1707, i agree with ubitux, stefano is the one to ask about -af
[01:41] <CIA-31> ffmpeg: 03Michael Niedermayer 07master * r905c4dc2b0 10ffmpeg/ (libavcodec/dpxenc.c tests/ref/lavf/dpx): 
[01:41] <CIA-31> ffmpeg: dpxenc: Dont store lavc indent when bitexact is set.
[01:41] <CIA-31> ffmpeg: This avoids fate breakage from bumping version.
[01:41] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[02:34] <slipperz> I've just started taking a look at the source. If I wanted to get familiar with the libavformat codebase where should I start?
[02:47] <durandal_1707> does FFmpeg have splay tree implementation?
[02:51] <CIA-31> ffmpeg: 03Christophe GISQUET 07master * r3eeb755763 10ffmpeg/libavcodec/rv34.c: 
[02:51] <CIA-31> ffmpeg: rv34: Inter/intra MB code split
[02:51] <CIA-31> ffmpeg: Split inter/intra macroblock handling code. This will allow further
[02:51] <CIA-31> ffmpeg: optimizations such as performing inverse transform and block reconstruction
[02:51] <CIA-31> ffmpeg: in a single pass as well as specialize code.
[02:51] <CIA-31> ffmpeg: Signed-off-by: Janne Grunau <janne-libav at jannau.net>
[02:51] <CIA-31> ffmpeg: 03Justin Ruggles 07master * rb2c75b6e63 10ffmpeg/libavcodec/ (avcodec.h internal.h pcm.c utils.c version.h): 
[02:51] <CIA-31> ffmpeg: avcodec: Add avcodec_encode_audio2() as replacement for avcodec_encode_audio()
[02:51] <CIA-31> ffmpeg: This allows audio encoders to optionally take an AVFrame as input and write
[02:51] <CIA-31> ffmpeg: encoded output to an AVPacket.
[02:51] <CIA-31> ffmpeg: This also adds AVCodec.encode2() which will also be usable by video and
[02:51] <CIA-31> ffmpeg: subtitle encoders once support is implemented in the public functions.
[02:51] <CIA-31> ffmpeg: 03Justin Ruggles 07master * r5ee5fa021f 10ffmpeg/libavcodec/ (avcodec.h utils.c): 
[02:51] <CIA-31> ffmpeg: avcodec: add a public function, avcodec_fill_audio_frame().
[02:51] <CIA-31> ffmpeg: This is a convenience function for the user to fill audio AVFrame information.
[02:52] <CIA-31> ffmpeg: This is a preparation for adding SMJPEG muxer.
[02:52] <CIA-31> ffmpeg: Signed-off-by: Kostya Shishkov <kostya.shishkov at gmail.com>
[02:52] <CIA-31> ffmpeg: 03Paul B Mahol 07master * rffa0923eee 10ffmpeg/ (6 files in 3 dirs): 
[02:52] <CIA-31> ffmpeg: add SMJPEG muxer
[02:52] <CIA-31> ffmpeg: Signed-off-by: Kostya Shishkov <kostya.shishkov at gmail.com>
[02:52] <CIA-31> ffmpeg: 03Michael Niedermayer 07master * r67f5650a78 10ffmpeg/: (log message trimmed)
[02:52] <CIA-31> ffmpeg: Merge remote-tracking branch 'qatar/master'
[02:52] <michaelni> slipperz, start with the avformat.h header
[02:53] <CIA-31> (12 lines omitted)
[02:53] <slipperz> ty
[02:54] <michaelni> durandal_1707, no but ive thought about adding support for that to tree.c/h a long time ago
[02:55] <michaelni> though you make me a bit curious, what do you want to do with splay trees ?
[02:55] <durandal_1707> xpm
[02:55] <durandal_1707> you have better idea?
[02:57] <michaelni> iam not sure i understand what you want to do with a splay tree in xpm (x pixmap ?)
[02:58] <durandal_1707> it is for picking color, imagemagick do that
[03:00] <durandal_1707> for small icons there would be no difference but for images with huge number of pixels&colors it would become really slow
[03:03] <michaelni> i suspect a hashtable might be faster than a splay tree
[03:05] <michaelni> but of course i might be wrong and this depends on the data
[03:15] <slipperz> In avformat.h, I see some documentation but then there are some titles like, ""Reading from an opened file" and "Seeking" that are empty. Are these documented somewhere else? Or will I just have to figure it out ?
[03:26] <michaelni> slipperz, our docs are not as good as they should be sadly, but improvments are very welcome!
[03:27] <michaelni> so when you figure something out that wasnt documented, please document it and submit a patch 
[03:27] <slipperz> michaelni, will do :)
[03:31] <michaelni> thx
[03:36] <durandal_1707> what about floating point pixels?
[03:40] <michaelni> what are floating point pixels ? i get no match for "float" in xpm.pdf
[03:41] <michaelni> but generically, if R/G/B color values are floats nothing should change
[03:41] <michaelni> if the "index" is a float, well a float is just a fancy 32bit value
[03:41] <michaelni> one can just read it as integer, calculate a hash and use a hashtable
[03:43] <slipperz> michaelni, quick question, what does pts stand for?
[03:43] <michaelni> presentation time stamp
[03:44] <durandal_1707> michaelni: reading about RGBE image format, imagemagick cant show file
[03:45] <slipperz> ty
[03:48] <durandal_1707> heh this reminds me of OpenEXR patch
[04:23] <Daemon404> orite
[04:23] <Daemon404> openexr could use fun stuff
[04:23] <Daemon404> like 16-bit floats for values
[04:23] <Daemon404> "fun"
[05:23] <Compn> hmm wheres pross
[05:23] <Compn> did sigmadesigns ever release their 'xvid' source ?
[05:54] <slipperz> anyone around?
[07:21] <ubitux> youhou valgrind fate in the green \o/
[08:42] <CIA-31> ffmpeg: 03Reimar Döffinger 07master * r7e62315c91 10ffmpeg/libavcodec/x86/rv34dsp.asm: 
[08:42] <CIA-31> ffmpeg: Use correct register size.
[08:42] <CIA-31> ffmpeg: Fixes compilation with NASM.
[08:42] <CIA-31> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[09:59] <mateo`> j-b: hi, i am currently working on extracting the starting timecode from MXF files and Tim Nicholson proposed me to upload some sample. Is there an offcial ftp server to upload those samples and is it possible to have access ? thx
[10:30] <Ti-CoeuR> salut
[11:12] <funman> mateo`: ftp://streams.videolan.org/incoming at least
[11:13] <funman> though the server is crashy
[12:02] <CIA-31> ffmpeg: 03Michael Bradshaw 07master * ra02694c926 10ffmpeg/libavcodec/libopenjpegenc.c: libopenjpegenc: Simplify gray encoding.
[12:02] <CIA-31> ffmpeg: 03Michael Bradshaw 07master * r0abe25aa7b 10ffmpeg/libavcodec/libopenjpegenc.c: libopenjpegenc: Add error messages.
[12:02] <CIA-31> ffmpeg: 03Michael Bradshaw 07master * ree33eb40fc 10ffmpeg/libavcodec/libopenjpegenc.c: libopenjpegenc: Rename four conversion functions.
[12:02] <CIA-31> ffmpeg: 03Michael Bradshaw 07master * r56af084cb5 10ffmpeg/libavcodec/libopenjpegenc.c: libopenjpegenc: Remove two redundant function parameters.
[12:03] <mateo`> funman: thx
[12:11] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * r9f7144b44c 10ffmpeg/libavcodec/mpegvideo_enc.c: 
[12:11] <CIA-31> ffmpeg: lavc/mpegvideo_enc: improve log message in case of too large GOP
[12:11] <CIA-31> ffmpeg: Change log level from ERROR to WARNING and provide more specific
[12:11] <CIA-31> ffmpeg: information in the message itself.
[12:11] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * ra798c20a76 10ffmpeg/ (3 files in 2 dirs): lavfi/testsrc: add "decimals" option to the testsrc filter
[12:11] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * r999495734b 10ffmpeg/ (4 files in 3 dirs): 
[12:11] <CIA-31> ffmpeg: lavu/eval: add if() and ifnot() eval functions
[12:11] <CIA-31> ffmpeg: They allow to implement the if/then/else logic, which cannot be
[12:11] <CIA-31> ffmpeg: implemented otherwise.
[12:11] <CIA-31> ffmpeg: For example the expression:
[12:11] <CIA-31> ffmpeg: A*B + not(A)*C
[12:11] <CIA-31> ffmpeg: always evaluates to NaN if B is NaN, even in the case where A is 0.
[12:11] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * r24860a913c 10ffmpeg/ (libavutil/Makefile tools/ffeval.c): 
[12:11] <CIA-31> ffmpeg: tools: add ffeval tool
[12:11] <CIA-31> ffmpeg: The tool is useful for testing the internal arithmetic evaluation engine
[12:11] <CIA-31> ffmpeg: (indeed I plan to use it in FATE), and provides a handy calculator when
[12:12] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * re4e291061c 10ffmpeg/doc/eval.texi: doc/eval.texi: add precisation about the assumption for "trueness" in conditional expressions
[12:57] <ubitux> saste: durandal_1707 was talking about a -af patch for ffmpeg; i looked at your gitorious branches, but it seems not available anymore
[12:57] <ubitux> did you drop it?
[12:58] <saste> ubitux: given the number of reworking done in the audio engine I preferred to delay its proper integration
[12:58] <ubitux> ok
[12:58] <saste> so I stopped to update it in my audio-filters branches
[12:59] <saste> it should still be available if you look in the oldest branches, but it obviously doesn't apply anymore
[12:59] <ubitux> it is available in the oldest branch, but the commit seems not readable
[13:00] <saste> which one?
[13:01] <ubitux> https://gitorious.org/~saste/ffmpeg/sastes-ffmpeg/commits/audio-filters-20110217  last one here
[13:02] <ubitux> it's not really important, i was just curious about it
[13:03] <saste> uhm... weird
[13:05] <saste> i still have it in my local branches, i'll try to revive it in a new audio-filters branch
[13:05] <saste> give me a few days, or i can send you the patches
[13:06] <ubitux> i just wanted to point it to durandal_1707
[13:06] <ubitux> after "< durandal_1707> michaelni: i have dumb -af patch for ffmpeg but i couldn't prove it is useful at all"
[13:07] <saste> -af patch for ffmpeg was also changing the fate output, since it uses by default the lavfi path when lavfi is enabled
[13:07] <saste> so it needs thorough testing
[13:45] <ubitux> bcoudurier: since you are not satisfied with the timecode api, if you want to comment on the new one, you should do it now, it will be harder to change it later
[13:45] <ubitux> ffmpeg-devel / "Rework and move timecode API to lavutil" patchset
[15:23] <durandal_1707> is there function in ffmpeg to get number of unique colors in picture?
[15:24] <av500> no
[15:24] <durandal_1707> something like fast bitcount....
[15:24] <av500> fact_colorcount?
[15:25] <av500> and with unique you mean unique 24bit RGB Values?
[15:25] <av500> or whatever colorspace
[15:27] <saste> durandal_1707: write a filter for that!
[15:28] <durandal_1707> i googled little
[15:28] Action: durandal_1707 going to see what ImageMagick does
[15:29] <ubitux> saste: speaking of this, i wonder if vf showinfo could have stuff like http://silicontrip.net/~mark/lavtools/#yuvvalues
[15:29] <ubitux> it will likely make it more useful over ffprobe -show_frames for instance 
[15:31] <ubitux> the other tools might be interesting as well
[15:34] <durandal_1707> for counting number of unique colors in 24bit image we just need 2^24 bits
[15:43] <saste> we may design common functions for that
[15:44] <saste> but i'm not sure i want to make ffprobe depend on lavfi
[15:44] <saste> (well maybe conditionally...)
[15:46] <ubitux> saste: i meant having vf showinfo being more useful that ffprobe -show_frames by providing more insight/analyzed informations
[15:46] <ubitux> ’ different purpose
[15:47] <saste> ubitux: or maybe extend ffprobe as well, although i'm not sure it is a good idea provided what ffprobe is from the philosophical pov (just a thin wrapper around what lavc/lavf shows about packets/streams/frames)
[15:47] <ubitux> -show_frames: used to grab global/generic frames informations in easily exploitable formats, and vf showinfo more information
[15:48] <saste> while extending showinfo is fine of course
[15:48] <ubitux> indeed, i don't think ffprobe -show_frames should be extented to print more information
[15:48] <ubitux> but afaik, showinfo is somehow a limited ffprobe -show_frames 
[15:48] <ubitux> so the link was some proposition on how to improve it ;)
[15:49] <saste> ubitux: patches are welcome ;-)
[15:50] <ubitux> heh i know ;)
[15:50] <saste> i'm fighting with timestamps at the moment
[17:22] <ubitux> saste: you should decide weither to put the number or the string in first :p
[17:22] <ubitux> @item passthrough, 0 [...] @item 1, cfr
[17:22] <saste> yes, i noticed one second after sending the patch
[17:23] <saste> that's why i send patches, for getting a post-send-button enlightenment
[17:23] <ubitux> :D
[19:58] <durandal_1707> drop darwin
[20:06] <Daemon404> ?
[20:07] <durandal_1707> drop darwin support
[20:10] <CIA-31> ffmpeg: 03Michael Niedermayer 07master * ra129622390 10ffmpeg/libavcodec/utils.c: 
[20:10] <CIA-31> ffmpeg: lavc: Check CODEC_CAP_VARIABLE_FRAME_SIZE && !frame
[20:10] <CIA-31> ffmpeg: This combination is quite odd and almost certainly a bug if
[20:10] <CIA-31> ffmpeg: it happens.
[20:10] <CIA-31> ffmpeg: Reviewed-by: Justin Ruggles <justin.ruggles at gmail.com>
[20:10] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:10] <CIA-31> ffmpeg: 03Michael Niedermayer 07master * raf5004b557 10ffmpeg/libavcodec/utils.c: 
[20:10] <CIA-31> ffmpeg: lavc: use av_freep() for freeing extended_data
[20:10] <CIA-31> ffmpeg: This is safer
[20:10] <CIA-31> ffmpeg: Reviewed-by: Justin Ruggles <justin.ruggles at gmail.com>
[20:10] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[20:13] <berdario> Hello, I have a problem with ffmpeg... I downloaded the sources and I'm trying to dig up the issue by debugging it... I'm trying to supply an .sdp to ffplay, and the file is not recognized...  after some calls to av_probe_input_format2 and 3 it fails by returning  AVERROR_INVALIDDATA
[20:14] <berdario> can anyone help me?
[20:15] <berdario> I stopped here since av_probe_input_format3 is completely devoid of comments
[20:16] <michaelni> berdario, can we see the sdp file ?
[20:16] <berdario> michaelni, http://paste.pocoo.org/show/536262/
[20:18] <CIA-31> ffmpeg: 03Reimar Döffinger 07master * r3b55429d56 10ffmpeg/libavcodec/ (asv1.c avcodec.h eatqi.c mjpegdec.c utils.c version.h): (log message trimmed)
[20:18] <CIA-31> ffmpeg: Add and use av_fast_padded_malloc.
[20:18] <CIA-31> ffmpeg: The same as av_fast_malloc but uses av_mallocz and keeps extra
[20:18] <CIA-31> ffmpeg: always-0 padding.
[20:18] <CIA-31> ffmpeg: This does not mean the memory will be 0-initialized after each call,
[20:18] <CIA-31> ffmpeg: but actually only after each growth of the buffer.
[20:18] <CIA-31> ffmpeg: However this makes sure that
[20:18] <CIA-31> ffmpeg: 03Reimar Döffinger 07master * r71a2c9b265 10ffmpeg/ffmpeg.c: (log message trimmed)
[20:18] <CIA-31> ffmpeg: ffmpeg: check tcgetattr result.
[20:18] <CIA-31> ffmpeg: This fixes parallel FATE (make fate -j4) failing under valgrind with:
[20:18] <CIA-31> ffmpeg:  Syscall param ioctl(TCSET{S,SW,SF}) points to uninitialised byte(s)
[20:18] <CIA-31> ffmpeg:  at 0x5D98B23: tcsetattr (tcsetattr.c:88)
[20:18] <CIA-31> ffmpeg:  by 0x43D66C: term_init (ffmpeg.c:734)
[20:18] <CIA-31> ffmpeg:  by 0x43CD8D: main (ffmpeg.c:5071)
[20:21] <michaelni> berdario, look at libavformat/rtsp.c:sdp_probe() but note iam not the SDP expert  ...
[20:23] <berdario> uhhhhh, thanks... do you know who I should ask further questions to?
[20:24] <michaelni> luca abeni
[20:24] <michaelni> but he isnt on IRC AFAIK
[20:25] <berdario> michaelni, "c" is an optional field... I'll try to modify my .sdp to get it working, but maybe this should be considered a bug and get fixed in rtsp.c? (or is it known bug? a design choice?)
[20:26] <berdario> ok, mail is fine for him, I guess?
[20:26] <michaelni> i dont think its a known bug
[20:26] <michaelni> yes
[20:26] <michaelni> also if the field is optional then its probably a bug and should be reportet on ffmpeg.org/trac
[20:27] <berdario> ok, thanks :)
[20:27] <michaelni> np
[20:48] <berdario> fyi, "c=" isn't really optional: other programs (like vlc) take the connection information from the origin "o=", but a "c=" should be specified either in the session or in each media description (thus, it's not a problem if it's missing from the session description like in my sdp, but since there's no media description the sdp is not actually correct)
[20:49] <j-b> mateo`: yes, please use ftp://streams.videolan.org/incoming
[20:50] <berdario> I wonder if it could be possible to add an av_log() line inside sdp_probe() to help other people who may have this issue
[20:52] <berdario> (mhn, I'll write the patch later)
[21:18] <CIA-31> ffmpeg: 03Paul B Mahol 07master * raff9cb8377 10ffmpeg/libavcodec/ (pamenc.c png.c pnmdec.c pnmenc.c): 
[21:18] <CIA-31> ffmpeg: lavc: remove unused headers
[21:18] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:19] <CIA-31> ffmpeg: 03Tomas Härdin 07master * r1de5154dfb 10ffmpeg/libavformat/mxfdec.c: 
[21:19] <CIA-31> ffmpeg: mxfdec: Consider OPAtom files that don't have exactly one EC to be OP1a
[21:19] <CIA-31> ffmpeg: This fixes demuxing of 2011_DCPTEST_24FPS.V.mxf.
[21:19] <CIA-31> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:16] <CIA-31> ffmpeg: 03Reimar Döffinger 07master * r178f8ea76e 10ffmpeg/libavcodec/utils.c: 
[22:16] <CIA-31> ffmpeg: Use ff_packet_free_side_data instead of duplicating it.
[22:16] <CIA-31> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[22:16] <CIA-31> ffmpeg: 03Reimar Döffinger 07master * rd393f41a33 10ffmpeg/ffmpeg.c: 
[22:16] <CIA-31> ffmpeg: Fix memleak after switch to encode_audio2 API.
[22:16] <CIA-31> ffmpeg: Obviously the packet allocated by the encoder must be freed somewhere.
[22:16] <CIA-31> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[22:52] <durandal_1707> cinepack regression https://trac.videolan.org/vlc/ticket/5872
[23:39] <cbsrobot> durandal_1707: it works wth ffplay ....
[23:41] <michaelni> cbsrobot, i can confirm i see nothing odd with ffplay, how do i reproduce this ?
[23:53] <CIA-31> ffmpeg: 03Stefano Sabatini 07master * r6c06868e61 10ffmpeg/doc/ffmpeg.texi: doc/ffmpeg.texi: specify symbolic names for the vsync parameter
[00:00] --- Wed Jan 18 2012

More information about the Ffmpeg-devel-irc mailing list