[Ffmpeg-devel-irc] ffmpeg-devel.log.20120730

[burek]

[00:01] <Shimmy> Anyone please help: https://ffmpeg.org/trac/ffmpeg/ticket/747
[00:10] <durandal_1707> how to handle encoder that outputs stuff only every n-th frame?
[00:26] <durandal_1707> i get floating exception from muxer because of this
[00:27] <iive> ?
[00:28] <durandal_1707> encoder outputs data after it gets enough input
[00:29] <iive> sure, if you don't have output, don't send it to muxer.
[00:29] <iive> codecs are supposed to work as zlib.
[00:30] <durandal_1707> i set got_packet_ptr to 1 only if i get something
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * rad2022ac83 10ffmpeg/libavfilter/vsrc_testsrc.c: lavfi/testsrc: add comment in options
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * ra74dcb7dc0 10ffmpeg/libavfilter/vsrc_color.c: 
[00:38] <CIA-41> ffmpeg: lavfi/color: cache and reuse colored picture in context
[00:38] <CIA-41> ffmpeg: Avoid to fill the same picture again and again with the same content.
[00:38] <CIA-41> ffmpeg: Optimize computation, and provides an example for the use of the
[00:38] <CIA-41> ffmpeg: AV_PERM_REUSE permission flag.
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * re5d617aa6d 10ffmpeg/ (3 files in 2 dirs): lavfi/color: drop support for old deprecated syntax
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r47b6b7a2b9 10ffmpeg/libavfilter/vsrc_color.c: lavfi/color: reindent after the previous commit
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r9cf9214d0a 10ffmpeg/libavfilter/vsrc_color.c: 
[00:38] <CIA-41> ffmpeg: lavfi/color: free internal context options in uninit()
[00:38] <CIA-41> ffmpeg: Keeping the set options in the context when the component is still used
[00:38] <CIA-41> ffmpeg: can be useful for debugging purposes.
[00:38] <CIA-41> ffmpeg: 03Stefano Sabatini 07master * r316f6df9cc 10ffmpeg/libavfilter/avcodec.c: 
[00:38] <CIA-41> ffmpeg: lavfi/avcodec: rename picref->samplesref variable in avfilter_get_audio_buffer_ref_from_frame()
[00:39] <CIA-41> ffmpeg: The new name is more meaningful in that context.
[00:42] <michaelni> durandal_1707, you need to set CODEC_CAP_DELAY for the encoder not sure if this is the issue though
[00:44] <durandal_1707> that is for every frame? not just last one?
[00:44] <durandal_1707> issue is i get floating point in mp3 muxer
[00:44] <durandal_1707> nut works but output is unplayable
[00:45] <durandal_1707> what i get segv with libmp3lame when muxing into nut
[00:45] <Shimmy> Anyone please take a look: https://ffmpeg.org/trac/ffmpeg/ticket/747
[00:45] <durandal_1707> previously at least it would abort instead of crashing
[01:01] <CIA-41> ffmpeg: 03Anton Khirnov 07master * rf66eeff1c8 10ffmpeg/libavformat/utils.c: lavf: round estimated average fps to a "standard" fps.
[01:01] <CIA-41> ffmpeg: 03Anton Khirnov 07master * raba232cfa9 10ffmpeg/ (21 files in 3 dirs): (log message trimmed)
[01:01] <CIA-41> ffmpeg: lavf: deprecate r_frame_rate.
[01:01] <CIA-41> ffmpeg: According to its description, it is supposed to be the LCM of all the
[01:01] <CIA-41> ffmpeg: frame durations. The usability of such a thing is vanishingly small,
[01:01] <CIA-41> ffmpeg: especially since we cannot determine it with any amount of reliability.
[01:01] <CIA-41> ffmpeg: Therefore get rid of it after the next bump.
[01:01] <CIA-41> ffmpeg: Replace it with the average framerate where it makes sense.
[01:01] <CIA-41> ffmpeg: 03Luca Barbato 07master * r41f43202cf 10ffmpeg/libavformat/flvdec.c: 
[01:01] <CIA-41> ffmpeg: flvdec: remove spurious use of stream id
[01:01] <CIA-41> ffmpeg: We match streams by codec id now.
[01:01] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r7d1918affe 10ffmpeg/: (log message trimmed)
[01:01] <CIA-41> ffmpeg: Merge remote-tracking branch 'qatar/master'
[01:01] <CIA-41> ffmpeg: * qatar/master:
[01:01] <CIA-41> ffmpeg:  flvdec: remove spurious use of stream id
[01:01] <CIA-41> ffmpeg:  lavf: deprecate r_frame_rate.
[01:01] <CIA-41> ffmpeg:  lavf: round estimated average fps to a "standard" fps.
[02:27] <durandal_1707> michaelni: i get segv when using lame encoder
[02:32] <michaelni> durandal_1707, how can i reproduce it ?
[02:33] <durandal_1707> try encode to mp3 container, can you reproduce it?
[02:33] <durandal_1707> Daemon404: what happened with you work on nut?
[02:33] <Daemon404> what work?
[02:34] <Daemon404> im using nut at work
[02:34] <Daemon404> trhough teh public api
[02:34] <Daemon404> im not workong ON nut
[02:34] <michaelni> ffmpeg -i matrixbench_mpeg2.mpg test.mp3  <--- works fine
[02:43] <durandal_1707> michaelni: strange segv only happens if i compile with libshine support
[04:00] <CIA-41> ffmpeg: 03jamal 07master * r938e4470ae 10ffmpeg/ (libavcodec/options.c libavformat/options.c): 
[04:00] <CIA-41> ffmpeg: options: Fix warning about incompatible pointer type
[04:00] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:01] <CIA-41> ffmpeg: 03jamal 07master * r753a1b0988 10ffmpeg/libavutil/samplefmt.c: 
[04:01] <CIA-41> ffmpeg: samplefmt: Fix warning about discarded qualifier
[04:01] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:13] <oneal> hello, for XVBA acceleration, Is there a implementation in ffmpeg source code. 
[04:14] <oneal> or Is there a plan to implement this acceleration
[04:15] <durandal_1707> no and no
[04:16] <oneal> Thanks
[04:17] <oneal> I'm focusing on implement this, but I don't know whether community accpet this code or not
[04:18] <durandal_1707> how something can be accepted if code is not reviewed?
[04:21] <oneal> Okay ,I know your mean that I should submit my code at firstly, but I just want to know whether community is interesting this feature or not.
[04:22] <ohsix> the community is interested in every feature!
[04:22] <oneal> sounds good. Thanks
[04:46] <CIA-41> ffmpeg: 03Paul B Mahol 07master * r8a6d673eff 10ffmpeg/libavformat/flacdec.c: 
[04:46] <CIA-41> ffmpeg: flacdec: fix memleaks
[04:46] <CIA-41> ffmpeg: While here refactor code.
[04:46] <CIA-41> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[04:51] <Compn> oneal : yes, xvba and all hwaccel api will be accepted
[04:51] <Compn> :)
[05:09] <oneal> Thanks , @Compn
[05:12] <Compn> oneal : its best to ask here or on the mailing list to make sure you design it so that it can be accepted
[05:13] <Compn> i mean, which way to design it , which api to use, etc
[05:24] <oneal> Okay
[05:25] <oneal> I will give a detail design for dicussing befor implementing it.
[05:26] <oneal> make sure we are the same page and it can be accepted by community.
[11:43] <ubitux> oneal: i see you're talking about deshake on one of the libav channel; you know ffmpeg has one?
[11:43] <ubitux> with an opened related ticket for improvment btw (https://ffmpeg.org/trac/ffmpeg/ticket/1430)
[11:49] <maker> Hello. I have a few questions about code conventions I saw on your repo :) Following Accelerated C++, I got used to the habit of putting '!=' operator inside loops; while ffmpeg seems to use '<'. Apart from debugging reasons, have you considered that replacing the operand could make comparisons faster? At least, that's what I read there.
[11:50] <maker> follows the second one :D
[11:51] <maker> AVFilterLink->cur_buf->data has planes placed one immediately after the other? I mean, in https://github.com/mmaker/FFmpeg/commit/0b99e4e4d2ffeac7b833161ebc5562a284114f31#L2R35
[11:51] <maker> I could replace that expression with just a sum, without reassignment?
[11:52] <maker> s/I could/could I/
[12:02] <rjiejie> hi, i'm a newbie, someone knows about that which file is the "mp4" demuxer in ffmpeg source code ? thanks
[12:03] <saste> maker: no, data is usually padded, that is linesize != pixel_step * width
[12:03] <CIA-41> ffmpeg: 03Nicolas George 07master * r33449b1776 10ffmpeg/libavutil/ (audioconvert.c audioconvert.h): audioconvert: add av_get_channel_description().
[12:03] <CIA-41> ffmpeg: 03Nicolas George 07master * rf893904469 10ffmpeg/ (4 files in 2 dirs): 
[12:03] <CIA-41> ffmpeg: audioconvert: add av_get_standard_channel_layout().
[12:03] <CIA-41> ffmpeg: Also bump minor version and add APIchanges entry.
[12:03] <CIA-41> ffmpeg: 03Nicolas George 07master * rcc55470472 10ffmpeg/libavutil/audioconvert.c: audioconvert: include the description of channels in the library.
[12:03] <saste> maker: also consider that we have already a negate filter
[12:04] <saste> regarding the != vs. < , i'm fine with != but I wonder how much difference it does
[12:05] <saste> i also suppose the difference will be platform dependent
[12:10] <saste> rjiejie: libavformat, grep for mp4
[12:14] <rjiejie> saste: thanks, i have found the file is related "isom.*" :)
[13:35] <maker> re
[13:58] <maker> saste: jsut testing. Uhm,concering the difference between '<' and '!=', I think that the compiler transforms '<' in a cmp instruction, which in practice is a subtraction; instead '!=' should correspond to a test instruction, alias logic and. But it's just a detail.
[14:35] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rdfd8f2ff19 10ffmpeg/libavfilter/avfilter.h: 
[14:35] <CIA-41> ffmpeg: avfilter: fix minor typo
[14:35] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:35] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rcbb2c95e47 10ffmpeg/ffmpeg.c: 
[14:35] <CIA-41> ffmpeg: ffmpeg: init simple filter graph before trying to extract framerate from its output.
[14:35] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:35] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r3161958455 10ffmpeg/ (libavfilter/vf_yadif.c tests/ref/fate/filter-yadif-mode1): 
[14:35] <CIA-41> ffmpeg: yadif: update frame rate
[14:35] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[14:39] <durandal_1707> michaelni: why pal8 -> pal8 requires that palette is always copied to frame->data[1]? pal8->rgb does not have such limitation
[14:46] <michaelni> durandal_1707, swscale doesnt copy the palette autiomatically or do i miss-guess what you are trying to do ?
[14:47] <durandal_1707> decoder does write palette into frame->data[1] if it does not change, only pal8->pal8 seems to be affected by this
[14:48] <durandal_1707> *decoder does not
[14:49] <michaelni> for reget_buffer() the decoder should not need to re-write, for get_buffer() the decoder always has to write as it could be a new buffer
[14:57] <durandal_1707> so using reget_buffer should better than having pal buffer in decoder itself and doing memcpy?
[15:43] <michaelni> reget_buffer() can be much slower than get_buffer()+memcpy of data[1] because reget will copy data[0] too in some cases
[15:44] <michaelni> this depends on if the user application can handle reusing buffers (this is often not possible)
[17:22] <Freakshow> in the event that I have some changes that may be beneficial upstream in trunk... how can I get these submitted for review? should I send an email out to ffmpeg-devel for review?
[17:24] <nevcairiel> yeah, send a patch to the mailing list
[17:24] <durandal_1707> Freakshow: use git format-patch, and send-email
[17:25] <Freakshow> great... thanks guys.
[17:25] <Freakshow> is there any particular subject keyword that should used?
[17:26] <nevcairiel> just make it "[PATCH] <first line of commit message here>"
[17:37] <Daemon404> git send-email does that on its own anyway.
[17:55] <durandal_1707> michaelni: between performance and simplicity i choose later
[17:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r64a2c51f7b 10ffmpeg/tests/ref/fate/zmbv-8bit: 
[17:55] <CIA-41> ffmpeg: fate: fix forgotten hunk for framerate order change in ffmpeg.c
[17:55] <CIA-41> ffmpeg: broken by: cbb2c95e47cd83cb65b2ce77cbb0296e85fa506b
[17:55] <CIA-41> ffmpeg: i had fixed that commit but pushed from an old branch
[17:55] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:55] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r64f07754fe 10ffmpeg/ (common.mak configure): 
[17:55] <CIA-41> ffmpeg: buildsys: fix c++ compilation
[17:55] <CIA-41> ffmpeg: Based on code by: carl
[17:56] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[17:58] <michaelni> durandal_1707, its your decoder (PAF) its your decission if you use reget or get+data1 copy
[17:58] <durandal_1707> the later case needs more lines in code
[18:00] <michaelni> yes but it will be slower with most video players
[18:00] <michaelni> thats because the have a que/fifo and cant reuse a buffer before it has been displayed
[18:01] <michaelni> so they will copy data[0] + [1]
[18:01] <durandal_1707> for game codecs with so small resolution that is really not noticeable
[18:01] <michaelni> true
[18:14] <michaelni> hi bcoudurier 
[18:14] <michaelni> any news about fate.ffmpeg.org ?
[18:14] <bcoudurier> called them this morning
[18:15] <michaelni> what did they say ?
[18:17] <CIA-41> ffmpeg: 03Nicolas George 07master * ra85b4a5696 10ffmpeg/libavfilter/video.c: 
[18:17] <CIA-41> ffmpeg: lavfi: use correct PTS for link age.
[18:17] <CIA-41> ffmpeg: When start_frame has returned, the buffer reference's PTS
[18:17] <CIA-41> ffmpeg: may have been modified by another filter.
[19:48] <Daemon404> j-b, ping.
[23:32] <CIA-41> ffmpeg: 03Mans Rullgard 07master * rf3eb008343 10ffmpeg/libavcodec/ (7 files): (log message trimmed)
[23:32] <CIA-41> ffmpeg: eamad/eatgq/eatqi: call special EA IDCT directly
[23:32] <CIA-41> ffmpeg: These decoders use a special non-MPEG2 IDCT. Call it directly
[23:32] <CIA-41> ffmpeg: instead of going through dsputil. There is never any reason
[23:32] <CIA-41> ffmpeg: to use a regular IDCT with these decoders or to use the EA IDCT
[23:32] <CIA-41> ffmpeg: with other codecs.
[23:32] <CIA-41> ffmpeg: This also fixes the bizarre situation of eamad and eatqi decoding
[23:32] <CIA-41> ffmpeg: 03Justin Ruggles 07master * rbe51e589cd 10ffmpeg/libavfilter/af_asyncts.c: af_asyncts: avoid overflow in out_size with large delta values
[23:32] <CIA-41> ffmpeg: 03Mans Rullgard 07master * r87cf481aa8 10ffmpeg/libavcodec/ (mpegvideo.c mpegvideo_common.h): 
[23:32] <CIA-41> ffmpeg: mpegvideo: remove unnecessary inclusions of faandct.h
[23:32] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[23:32] <CIA-41> ffmpeg: 03Mans Rullgard 07master * r591766a3a9 10ffmpeg/libavcodec/eamad.c: 
[23:32] <CIA-41> ffmpeg: eamad: remove use of MpegEncContext
[23:32] <CIA-41> ffmpeg: There is no sense in pulling in this monster struct just for
[23:32] <CIA-41> ffmpeg: a handful of fields. The code does not call any functions
[23:32] <CIA-41> ffmpeg: expecting an MpegEncContext.
[23:32] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[23:32] <CIA-41> ffmpeg: 03Mans Rullgard 07master * r70c8753d33 10ffmpeg/tests/fate/filter.mak: 
[23:32] <CIA-41> ffmpeg: fate: amix: specify f32 sample format for comparison
[23:32] <CIA-41> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[23:32] <CIA-41> ffmpeg: 03Justin Ruggles 07master * r70d71b5801 10ffmpeg/ (doc/filters.texi libavfilter/af_asyncts.c): 
[23:32] <CIA-41> ffmpeg: af_asyncts: add first_pts option
[23:32] <CIA-41> ffmpeg: This allows for padding/trimming at the start of stream. By default, no
[23:32] <CIA-41> ffmpeg: assumption is made about the first frame's expected pts, so no padding or
[23:32] <CIA-41> ffmpeg: trimming is done.
[23:33] <CIA-41> ffmpeg: 03Anton Khirnov 07master * r29d70274ec 10ffmpeg/libavformat/wv.c: wv: K&R formatting cosmetics
[00:00] --- Tue Jul 31 2012