[Ffmpeg-devel-irc] ffmpeg-devel.log.20130221

burek burek021 at gmail.com
Fri Feb 22 02:05:02 CET 2013


[00:12] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:47e92153b61b: mjpegdec: check for packed rgb mjpeg.
[00:28] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:551d8b58ebc0: brender_pix: use a grayscale palette as fallback instead of a uninitialized one.
[01:01] <michaelni> someone wants to test(&review) "v4l2: bug #1570 and possible solution" ?
[01:03] <ubitux> saste: 
[01:03] <ubitux> libavfilter/vf_overlay.c:201:19: warning: overlay_formats may be used uninitialized in this function [-Wmaybe-uninitialized]
[01:03] <ubitux> libavfilter/vf_overlay.c:202:19: warning: main_formats may be used uninitialized in this function [-Wmaybe-uninitialized]
[01:04] <ubitux> michaelni: isn't it specific to some devices?
[01:04] <ubitux> iirc originally i wasn't able to reproduce
[01:04] <ubitux> but the issue might have evolved
[01:05] <michaelni> could be specific yes but some testing is still better than none
[01:06] <ubitux> http://fate.ffmpeg.org/report.cgi?time=20130220225403&slot=x86_64-archlinux-gcc-valgrindundef  erm it seems we have some more in vf idet too
[01:06] <ubitux> (which looks unrelated to the others)
[01:09] <cone-561> ffmpeg.git 03Clément BSsch 07master:580e2285577c: lavf/swfdec: remove 8-bit audio FIXME.
[01:09] <cone-561> ffmpeg.git 03Clément BSsch 07master:7346354b5597: lavf/microdvd: fix muxing.
[01:09] <cone-561> ffmpeg.git 03Clément BSsch 07master:9ed6af4be2c6: fate/subtitles: add MicroDVD remux test.
[01:59] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:dc9edb06b011: vf_idet: Fix macro arguments sideeffect
[01:59] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:cdb7a1ac6345: vf_idet: adapt default scores due to bugfix
[02:01] <ubitux> thx :)
[02:30] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.8:2b6f3be08250: avcodec_align_dimensions2: Ensure cinepak has large enough buffers.
[02:30] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.8:dfeef3a209ec: cook: check js_subband_start for validity
[02:36] <pajamaboy> quick question, does ffmpeg support authentication against an rtmp target, it appears to in a recent patch, but doesn't seam to be working for me
[03:26] <cone-561> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n0.8.14': unknown revision or path not in the working tree.
[03:26] <cone-561> Use '--' to separate paths from revisions
[03:26] <cone-561> refs/tags/n0.8.14:HEAD: vf_idet: adapt default scores due to bugfix
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:1fd86f9a2136: vqavideo: fix return type
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:8aedb7515674: tiff: check bppcount
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:5af2fd317d84: wma: check byte_offset_bits
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:17966ae5bb43: qdm2: increase noise_table size
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:0a57df38f468: Update for 0.8.14
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:2b6f3be08250: avcodec_align_dimensions2: Ensure cinepak has large enough buffers.
[03:54] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:dfeef3a209ec: cook: check js_subband_start for validity
[03:55] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:a92cdd896345: Merge branch 'release/0.8' into release/0.7
[03:55] <cone-561> ffmpeg.git 03Michael Niedermayer 07release/0.7:3be6a8e1eb97: update for 0.7.15
[04:04] <Compn> ubitux : i have an interesting question. do any containers have subtitle codepage metadata ?
[04:05] <Compn> like does mkv have stream 3 srt rus cp1251 ?
[04:06] <Compn> or utf8 maybe
[04:18] <cone-561> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n0.7.15': unknown revision or path not in the working tree.
[04:18] <cone-561> Use '--' to separate paths from revisions
[04:18] <cone-561> refs/tags/n0.7.15:HEAD: vf_idet: adapt default scores due to bugfix
[08:10] <Belinda> hi
[08:11] <Belinda> I am trying to make a patch to add S16 support to aacdec.c but i am failing continuously. Any ideas?
[09:12] <cone-190> ffmpeg.git 03Carl Eugen Hoyos 07master:d714576037e3: Use SMPTE standard notation when showing mxf Universal Labels.
[09:12] <cone-190> ffmpeg.git 03Carl Eugen Hoyos 07release/1.0:4c2b6ce26cf9: Fix bits_per_coded_sample when encoding png with frame-level multithreading.
[09:12] <cone-190> ffmpeg.git 03Carl Eugen Hoyos 07release/1.1:4be63111d1a2: Fix bits_per_coded_sample when encoding png with frame-level multithreading.
[10:10] <highgod> Hi, I want to ask a question. I want to modify the configure file to make it detect cl/cl.h and libOpenCL.a and make the ffmpeg link libOpenCL.a,the code I add is "enabled opencl    && require  opencl CL/cl.h OPENCL_init -lOpenCL", but the output is "ERROR: opencl not found", I am not familiar with modifing configure file. I used --extra-cflags=-I/opt/include --extra-ldflags=-L/opt/lib/x86 
[10:10] <highgod> --extra-libs="-lOpenCL" to compile the opencl code
[10:31] <Belinda> I am trying to make a patch to add S16 support to aacdec.c but i am failing continuously. Any ideas?
[10:31] <durandal_1707> why?
[10:32] <durandal_1707> aacdec is floating point decoder
[10:34] <nevcairiel> decoders intentionally only output their native format, if you need another format, use swresample to convert it
[10:49] <Belinda> actually i am using it on android and android only recognizes s16, so i wanted to add support for "requested_name".
[10:55] <durandal_1707> the aac decoder is float one, but there are fixed decoders on web
[10:55] <av500> one can also convert float to fixed
[10:56] <av500> clamp( lrintf( *(float*)src * (1 << 15)))
[10:57] <durandal_1707> does that give good results?
[10:58] <av500> define good
[10:59] <divVerent> it sure is good enough :P
[10:59] <divVerent> actually, shouldn't one use ((1 << 15) - 1)? nitpicking :P
[10:59] <Belinda> i tried float_to_int16_interlave_c  but it didn't work
[10:59] <av500> why not?
[10:59] <nevcairiel> like i said, use swresample to convert float to int16
[11:00] <nevcairiel> thats what its there for
[11:00] <nevcairiel> among other things
[11:00] <durandal_1707> av500: because floats from aacdec are between -1 and 1
[11:00] <nevcairiel> a patch to allow s16 output would not really be accepted
[11:01] <Belinda> :/
[11:01] <nevcairiel> it used to be possible, but that was removed from all sorts of decoders in favor of native output
[11:02] <nevcairiel> so instead of converting inside the decoder, just convert after the decoder
[11:02] <av500> yes, but one can easily find the old code
[11:02] <nevcairiel> same code required, jus
[11:02] <nevcairiel> just in your app
[11:02] <av500> yes
[11:02] <nevcairiel> or just use the library which has functions for this =p
[11:03] <Belinda> actually i found the old code but there are many structural changes, it is easy to ruin
[11:04] <Belinda> since i am doing this on the fly, i try to reduce cpu and ram usage
[11:04] <av500> its only audio
[11:04] <av500> doing that conversion wont kill you
[11:04] <av500> and you need to do it anyway
[11:05] <nevcairiel> the decoder works in float anyway, so it needs conversion at some point, the cpu required is the same no matter where its done
[11:06] <av500> yep
[11:07] <durandal_1707> so on slow arch with float it is slow already
[11:07] <av500> aac takes like 30mhz on an ARM
[11:08] <av500> find me an android device that you can even clock that low
[11:08] <av500> its a non issue
[11:10] <durandal_1707> av500: on arm? maybe
[11:10] <av500> yes
[11:10] <av500> on an armv7
[11:11] <av500> basically you have more cpu power due to leakage than due to that bit of audio decoding
[11:12] <durandal_1707> so fixed point decoder would not be faster?
[11:13] <av500> probably
[11:13] <av500> well
[11:14] <av500> since armv7 with neon etc has float I am not even sure fixed is faster
[11:27] <Belinda> AV_SAMPLE_FMT_S16 is not supported by swr_convert?
[11:28] <durandal_1707> it is supported
[11:28] <durandal_1707> read documentation
[11:28] <Belinda> ok sorry, i've made a mistake
[11:37] <ubitux> Compn: some formats contain the info yes
[11:38] <ubitux> (microdvd for instance)
[11:54] <durandal_1707> michaelni: this ^ spammer is really retarded
[11:55] <durandal_1707> delete him and png (that links to another site)
[11:56] <durandal_1707> also can permission for WikiStart be more restrictive?
[11:56] <durandal_1707> is there better bug tracker which have better way of handling fine grained permissions?
[11:57] <durandal_1707> if yes, switch to it
[12:38] <cone-212> ffmpeg.git 03Nicolas George 07master:01649c79c87c: lavfi/buffersrc: set channel layout if it is known.
[12:43] <cone-212> ffmpeg.git 03Nicolas George 07master:9fa5a47197c6: ffmpeg: set resample_pix_fmt for sub2video.
[12:51] <highgod> Hi,I add check opencl code to configure file, and set the check code to enabled opencl    && require opencl CL/cl.h clEnqueueNDRangeKernel -lOpenCL,clEnqueueNDRangeKernel is a function of libopencl.a, but it failed,Then I review the libopencl.a and found that the function is clEnqueueNDRangeKernel at 36, but if I set the code to enabled opencl    && require opencl CL/cl.h clEnqueueNDRangeKernel at 36 
[12:51] <highgod> -lOpenCL, it si still wrong. does anyone konws how to fix it? Thanks
[13:01] <highgod> require_cpp?
[13:03] <cone-212> ffmpeg.git 03Vicente Jimenez Aguilar 07master:b5ad422bf4e6: doc: filters: Correct BNF FILTER description
[13:03] <cone-212> ffmpeg.git 03Diego Biurrun 07master:3e85b46ecf3c: arm: vp8: Add missing #includes for header to compile standalone
[13:03] <cone-212> ffmpeg.git 03Diego Biurrun 07master:a0a1ac0f091d: dxva2: Add missing #define to make header compile standalone
[13:03] <cone-212> ffmpeg.git 03Michael Niedermayer 07master:6b8f21190d5a: Merge remote-tracking branch 'qatar/master'
[13:05] <cone-212> ffmpeg.git 03Paul B Mahol 07master:a3d890527ed3: exr: reset compression in decode_frame()
[13:05] <cone-212> ffmpeg.git 03Paul B Mahol 07master:59015c1118ac: exr: constify s in rle_uncompress()
[13:05] <highgod> sill don't work
[13:16] <highgod> require2 works
[13:21] <Compn> durandal_1707 / michaelni : also, it would help to give devels more access on trac to ban spammers.
[13:27] <michaelni> theres a bayesian spam filter only problem is we have too few spamers so it hasnt enabled itself yet
[13:29] <michaelni> also trac can use various external spam detectors like akismet but IIUC we need a api key for each of these
[13:34] <michaelni> and about banning, that slavon888 guy used a different IP each time IIRC
[13:35] <iive> might be good idea to block known exit nodes of tor network and vpn/proxy tunnels.
[13:35] <iive> probably some spam lists have them all.
[13:37] <michaelni> trac already uses multipe ip blacklists, i dunno whats on them exactly though
[13:46] <michaelni> Compn, if you want to ban IPs or content see for example trac.edgewall.org/wiki/BadContent and trac.openpsa2.org/wiki/BadIP
[13:50] <michaelni> also in case i worded it unclearly, some volunteer needs to get us api keys for akismet and the others listed at http://trac.edgewall.org/wiki/SpamFilter with exception of where they key is just for training the external service) 
[13:50] <michaelni> also i dont know if these keys are free 
[13:51] <michaelni> that is if someone wants trac to use these external spam detectors ...
[14:04] <durandal_1707> i gonna push rest of exr patches
[14:04] <durandal_1707> than gonna try slice threading
[15:04] <blackmage> Does anyone know if pushing to an rtmp destination that requires authentication is supported
[15:04] <blackmage> it looks like it was added in a patch, but i'm not certain, based upon some of the docs i've seen
[15:06] <durandal_1707> what docs?
[15:06] <durandal_1707> didn't you said in another channel that your buid is missing ssl support?
[15:06] <cone-212> ffmpeg.git 03Michael Niedermayer 07master:bdeb61ccc679: h264: check that luma and chroma depth match
[15:09] <blackmage> i'm seeing mentions of it, but here is the one that says you need ssl
[15:09] <blackmage> http://ffmpeg.zeranoe.com/forum/viewtopic.php?f=7&t=723#p2581
[15:10] <durandal_1707> ssl is needed
[15:12] <durandal_1707> or not
[15:32] <durandal_1707> could fate survive 8 mb of new samples?
[15:52] <durandal_1707> it looks like it needs user[:pass]@hostname
[15:52] <durandal_1707> (reading source code)
[15:53] <blackmage> okay, i'll try that, thanks durandal
[16:29] <wm4> so it seems some libavformat demuxer actually expect a special AVIOContext and use pb->opaque without checking
[16:30] <wm4> e.g. hls.c accesses this, and if the caller passes a custom AVIOContext, it will crash
[16:45] <cone-212> ffmpeg.git 03Michael Niedermayer 07master:669d50b44596: ec: check that reference pictures have matching parameters before using them.
[17:06] <Compn> wm4 : btw , write down any info you are learning, so the next person who wants to do what you do, doesnt have to guess and figure out ... :)
[17:06] <Compn> aka please update the docs or make source code comments, or wiki entry , anything
[17:08] <wm4> uh what, like what
[17:52] <Compn> wm4 : i thought you were adding a new format, not investigating a bug, nm
[18:07] <durandal_1707> finally! ffmpeg is recognised by google+: https://plus.google.com/+ffmpeg
[18:08] <Compn> finally, they can track ffmpeg people on google servers, yay...
[18:17] <durandal_1707> https://docs.google.com/presentation/d/1awg1CHM1w128iOBp_JOxE2DgHfywBeyjDe2bkx1vfVQ/edit?pli=1#slide=id.g96dbcc0f_116
[18:20] <ubitux> :)
[18:23] <ubitux> i like slide 6 too
[18:31] <durandal_1707> you can even chat with other readers
[18:52] <durandal_1707> http://revolution.byu.edu/programmingconcepts/potatoPseudocode.png
[19:31] <durandal_1707> anyone want to review my patches?
[20:04] <ubitux> durandal_1707: can i help?
[20:05] <saste> ubitux, durandal_1707: a second look at blend may help, especially with regards to interface/syntax
[20:08] <ubitux> blend looked good to me when i had a look at it
[20:08] <saste> ubitux, also boxblur and overlay patch
[20:08] <ubitux> named options?
[20:09] <ubitux> saste: btw, did you see the warnings in overlay?
[20:09] <saste> yes, and we're still missing a few filters if you want to help with low hanging fruits
[20:09] <saste> ubitux, nope
[20:25] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:8036ba299d33: h264: move next/last picture init to decode_slice_header
[20:27] <durandal_1707> sase left
[20:46] <ubitux> erm max_samples isn't honored.
[20:58] <ubitux> looks like a mistake in 82541d83
[20:58] Action: ubitux mailing Nicolas
[20:59] <durandal_1707> "Failed to inject frame into filter network" <--- what that means
[21:01] Action: ubitux reviewing boxblur
[21:18] <cone-48> ffmpeg.git 03Paul B Mahol 07master:47619e41c4a1: blend filter
[21:19] <ubitux> yay :)
[21:19] <durandal_1707> now fix bugs in it
[21:19] <ubitux> :))
[21:20] <ubitux> durandal_1707: how i imagine you currently: http://z0r.de/5024
[21:21] <durandal_1707> ubitux: whatever it is i do not have flash/java or some other plugin
[21:22] <ubitux> i think it works with ffplay
[21:22] <ubitux> but it won't loop
[21:22] <ubitux> too bad :)
[21:22] <ubitux> indeed it works
[21:22] <ubitux> you need to imagine the audio and video loop now
[21:23] <ubitux> (they are not in sync)
[21:23] <durandal_1707> ubitux: ???
[21:23] <ubitux> durandal_1707: wget 'http://z0r.de/L/z0r-de_5024.swf'
[21:24] <ubitux> the basic idea of the loop (at least most of them) is to have a/v not really in sync
[21:24] <ubitux> for instances 5 frames looping, and a longer background audio, looping too
[21:24] <durandal_1707> wget: Command not found.
[21:24] <ubitux> (the a/v sync doesn't really matter)
[21:24] <ubitux> mmh
[21:24] <ubitux> fetch?
[21:25] <ubitux> curl ... > out.swf?
[21:25] <durandal_1707> hmm why clip is in rawvideo/audio, feared of patents?
[21:27] <ubitux> no idea
[21:27] <ubitux> happens often
[21:27] <ubitux> (in swf)
[21:48] <durandal_1707> saste: gonna review [PATCH] Handling special characters in a URL ? it take really to loong
[21:49] <saste> durandal_1707, too long? he posted a reply tomorrow
[21:49] <durandal_1707> ping reply
[22:00] <ubitux> actually the min/max samples works really well
[22:00] <ubitux> awesome Nicolas is awesome
[22:00] <ubitux> saste: you should use that feature too ;)
[22:01] <ubitux> i wonder if asetnsamples couldn't use it too
[22:02] <wm4> why isn't there libswblend :(
[22:02] <saste> ubitux, you should thank anton to be fair
[22:02] <saste> wm4, you know the answer
[22:02] <wm4> "patches welcome"?
[22:02] <saste> yes
[22:02] <wm4> "no thank you"
[22:03] <ubitux> saste: no
[22:03] <ubitux> it's a feature from Nicolas
[22:03] <ubitux> not present in libav
[22:03] <saste> is it?
[22:03] <ubitux> yes :)
[22:03] <ubitux> see 0689d5e17ac0a3269c604b4df2c01140be328647
[22:04] <durandal_1707> wm4: why libswblend?
[22:04] <ubitux> durandal_1707: to blend subtitles \o/
[22:04] <ubitux> (fast!)
[22:04] <wm4> durandal_1707: it would gather useless functionality into a library, instead of shattering it across libavfilter where it's needed for various things
[22:04] <durandal_1707> there are actually many libs that do blend but they all sucks afaik
[22:05] <llogan> saste: worthless fact for the day: there are more than 1500 -user subscribers.
[22:05] <ubitux> how many -devel?
[22:05] <llogan> also i heard back from mike. he'll be able to deal with account stuff.
[22:05] <durandal_1707> day before were 1800 (so it is 0.3 % drop)
[22:06] <llogan> ubitux: 1327
[22:06] <ubitux> nice
[22:06] <ubitux> 10 left
[22:06] <durandal_1707> that all trackers (gmane,google, etc)
[22:06] <llogan> digest members are included but digests are disabled and they do not recieve mail
[22:08] <llogan> saste: so i'll try to get the gsoc ideas page up later today (unless someone gets to it first), but i've been in a timestuck editing videos for the last few days...
[22:08] <saste> llogan, no hurry
[22:09] <saste> i'll be able to work actively on gsoc not before the first week of march
[22:09] <durandal_1707> llogan: so is number of users/devels stagnating or they are all spammers?
[22:10] <llogan> i don't actively watch the subscriber numbers and there are no spammers with accounts AFAIK (not including hijacked [yahoo mostly] accounts)
[22:14] <durandal_1707> there is no log or something like graph?
[22:17] <llogan> durandal_1707: not in the admin web panel, AFAIK.
[22:20] <llogan> but a graph would be interesting
[22:20] <wm4> durandal_1707: have you found a sample for vf_stereo3d yet?
[22:21] <saste> anyone wants to review the partial overlay patch?
[22:21] <saste> it's the missing one in my overlay patchset
[22:21] <durandal_1707> wm4: not yet there, why?
[22:21] <wm4> just wondering
[22:22] <durandal_1707> it is not that mandatory for porting
[22:22] <wm4> its output looks like modern art for me
[22:22] <durandal_1707> anything with bunch of different colors can be used
[22:23] <durandal_1707> wm4: get glasses
[22:23] <wm4> and I wonder what stuff like "anaglyph_green_magenta_half_color" actually means
[22:24] <durandal_1707> nothing relevant for anime scene
[22:29] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:8a6449167a6d: avcodec_decode_audio4: check got_frame_ptr before handling initial skip
[22:29] <cone-48> ffmpeg.git 03Andrea3000 07master:8d8c59480e1b: matroska: fix missing ,
[22:31] <durandal_1707> wm4: looks like that filter is only way to experience 3d at home for free
[22:43] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:04220b473e9d: h264: ensure that get_format() is called when changing format but not otherwise.
[22:43] <cone-48> ffmpeg.git 03Matt Wolenetz 07master:82a4a4e7caa9: Fix Win64 AVX h264_deblock by not using redzone on Win64
[23:06] <cone-48> ffmpeg.git 03James Almer 07master:67b3fcf65578: configure: Use libiconv on systems that lack built-in iconv
[23:06] <cone-48> ffmpeg.git 03Mirjana Vulin 07master:8d2eb5fe58ae: mips: optimization for float aac decoder (sbr module)
[23:06] <cone-48> ffmpeg.git 03Mirjana Vulin 07master:46d52a0b098f: mips: optimization for float aac decoder (ps module)
[23:10] <ubitux> it almost makes me wanting a mips device
[23:11] <michaelni> ubitux, ask the @mips.com people, maybe they would give you one for free
[23:16] <ubitux> not sure if i'll really make use of it, it's just a whim :)
[23:16] <ubitux> but it's nice to see all these optims
[23:18] <wm4> durandal_1707: it appears matroska even has hints for this 3D crap...
[23:20] <JEEB> H.264 can have hints as well, but yeah -- matroska had some crap too
[23:22] <durandal_1707> there is unsupported H.264 extension IIRC
[23:23] <JEEB> MVC you mean? that's separate, and pure H.264 decoders just don't have to care as it will get one of the views that way just fine
[23:23] <JEEB> I mostly meant the SEI (or VUI was it?) stuff
[23:23] <JEEB> where you can note what kind of multi-view stream it is
[23:23] <JEEB> x264 has support for those
[23:24] <Skyler_> SEI
[23:41] <ubitux> huh
[23:41] <ubitux> seems i have the audio normalization working
[00:00] --- Fri Feb 22 2013


More information about the Ffmpeg-devel-irc mailing list