[Ffmpeg-devel-irc] ffmpeg-devel.log.20141022

[burek]

[00:16] <cone-665> ffmpeg.git 03Vittorio Giovara 07master:b69183f65d0f: nutenc: check for negative index rather than assert
[00:16] <cone-665> ffmpeg.git 03Michael Niedermayer 07master:dd3f1563d766: Merge commit 'b69183f65d0f6f4cd7a4e6c0deb57d59b0185aba'
[00:42] <amalia> ramiro ping
[00:53] Last message repeated 1 time(s).
[00:53] <amalia> michaelni I am running Scientific Linux 6.4 right now on my machine. Is it eligible for FATE testing ?
[00:59] <rcombs> amalia: sure
[00:59] <rcombs> run `make fate` with the relevant sample configuration
[01:02] <michaelni> amalia, is that running inside virtualbox/qemu or the host OS of the machine
[01:03] <cone-665> ffmpeg.git 03Vittorio Giovara 07master:af7ca6ea124b: nutdec: check av_new_packet return value
[01:04] <cone-665> ffmpeg.git 03Michael Niedermayer 07master:42316907b874: Merge commit 'af7ca6ea124b82b337a6b96e10963e88eba57ebe'
[01:06] <amalia> michaelni The host OS of the machine is Scientific Linux 6.4
[01:11] <michaelni> ahh, ok, did ramiro give any instructions what you should do for the qualification task ? i dont want to create confusion by suggesting something different
[01:11] <michaelni> either way we dont have a Scientific Linux fate client yet
[01:13] <amalia> Great. So setting up FATE on this SL6 client can be a qualification task right ? ramiro Gave me no qualification task 
[01:13] <kierank> no
[01:14] <kierank> setting up fate doesn't involve programming
[01:14] <michaelni> kierank, opw doesnt require programming
[01:14] <kierank> really?
[01:14] <kierank> oh
[01:14] <kierank> my bad
[01:14] <michaelni> opw != gsoc
[01:14] <amalia> :)
[01:15] <amalia> I've got a Desktop machine which runs SL6 and Fedora upon which I can install virtual box
[01:15] <amalia> Hope that helps
[01:16] <michaelni> a virtualbox SL6 fate client which submits results to our server should work as part of the qualifiation yes
[01:17] <michaelni> but non virtual would be a bit difficult, as we want to run the fate client regularly and long term
[01:17] <michaelni> and i think neither we nor you have a dedicated box for every client ;)
[01:35] <michaelni> amalia, so, i would suggest you install virtualbox and then create a new virtualbox client with 10gb virtual disk and install SL6 in that virtual client, then install gcc, rsync, ssh, ccache, git in it, git clone ffmpeg and setup a script which runs/submits fate results to the server on it
[01:37] <michaelni> ramiro should still have an old script that does the fate running in a loop, that can be used probably
[01:37] <cone-665> ffmpeg.git 03Carl Eugen Hoyos 07master:b6e8f10bcbf2: lavf/cinedec: Fix a typo in the file header.
[01:38] <amalia> Okay. he told me about the script yesterday. I'll ask him for it 
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:3998071039b2: avcodec/qpeg: fix off by 1 error in MV bounds check
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:1ac4ae2a32d1: avcodec/smc: fix off by 1 error
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:603f4ecd14a1: avcodec/svq3: Dont memcpy AVFrame
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:fdc8f4e5b4f1: avcodec/vorbisdec: Fix off by 1 error in ptns_to_read
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:cbfd6b1fa4b1: avformat/mpegts: Check desc_len / get8() return code
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:5926bea980ad: avformat/mpegts: use a padded buffer in read_sl_header()
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:279003eb9ede: avcodec/h264: Check mode before considering mixed mode intra prediction
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:e2e0c5b7f64d: avcodec/hevc: fix chroma transform_add size
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:5f7e48a11391: avcodec/on2avc: Check number of channels
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:1ebd7d2ccfe1: swresample/swresample: fix sample drop loop end condition
[02:26] <cone-665> ffmpeg.git 03Christophe Gisquet 07release/2.3:f2fba0749486: utvideoenc: properly set slice height/last line
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:d887a1214588: postproc/postprocess: fix quant store for fq mode
[02:26] <cone-665> ffmpeg.git 03Michael Niedermayer 07release/2.3:0f4c149730f1: postproc: fix qp count
[02:26] <cone-665> ffmpeg.git 03Lukasz Marek 07release/2.3:4fbdac00e94e: lavd: export all symbols with av_ prefix
[02:47] <cone-665> ffmpeg.git 03Michael Niedermayer 07master:c5ffd7aee5a6: avcodec/mjpegdec: use FF_CEIL_RSHIFT() for width
[03:14] <cone-665> ffmpeg.git 03James Almer 07master:e397edbfbab0: tools/crypto_bench: add CAST5 support
[08:25] <rcombs>  <Bucket> FFFT is Fuckin' Fast Fourier Transform
[10:04] <wm4> nevcairiel: how good is the ffmpeg_dxva2.c code actually? I assume it's perfectly sane, but lacks hacks for old hw or drivers?
[10:04] <nevcairiel> probably something like that
[10:06] <wm4> still sounds good enough
[10:58] <Daemon404> interesting
[10:59] <Daemon404> the image3 (format_pipe formats) causes a slight regression in that it no longer reports nb_frames=1
[11:00] <wm4> someone is using nb_frames?
[11:02] <Daemon404> i dunno this falls into a 'weird' category
[11:02] <Daemon404> the image2/image2pipe/image3 stuff is weird
[11:02] <Daemon404> lavf for images.... is ... special
[11:03] <wm4> I think I'd like a flag whether something is an image
[11:03] <nevcairiel> image2/3 can also be several images
[11:04] <nevcairiel> behave like a video even
[11:05] <Daemon404> wm4, sometimes you can check attached_Pic
[11:05] <Daemon404> sometimes.
[11:05] <wm4> lol
[11:05] <Daemon404> im currently using a hack to see if it's a standalone image
[11:05] <wm4> sometimes I forget how weird the API is
[11:06] <Daemon404> stockholm?
[11:06] <wm4> possible
[11:07] <Daemon404> but like nevcairiel said there's also image sets
[11:07] <wm4> but without timestamps, I suppose
[11:07] <nevcairiel> of course, its just empty images
[11:08] <nevcairiel> you can specifc a framerate to image2 i think, and it makes up timestamps
[11:08] <nevcairiel> specify*
[11:08] <wm4> empty?
[11:09] <wm4> also, is there no flag for formats with no timestamps? I thought there was one
[11:09] <nevcairiel> i think there is
[11:10] <wm4> also, seems like img2 sets 25 fps as default framerate
[11:10] <nevcairiel> quite possible
[11:11] <wm4> but doesn't set any specific flags, just AVFMT_GENERIC_INDEX
[12:37] <ubitux> it seems disabling threading doesn't help for that fifo crap
[12:37] <ubitux> fuck, that's really annoying, i don't have a workaround
[12:37] <ubitux> any idea what could have gone wrong here? (http://trac.ffmpeg.org/ticket/4051)
[12:37] <ubitux> mux.c + pts/dts 
[12:38] <ubitux> sounds hellish :(
[12:38] <nevcairiel> why does dts turn negative after so many frames
[12:50] <rcombs> wraparound?
[12:50] <rcombs> (guess without looking)
[12:50] <nevcairiel> its only 50, not 5 million
[12:51] <rcombs> large spacing between timestamps?
[12:52] <rcombs> (i.e. either very low frame rate or very high timebase?)
[12:54] <the-erm> I don't know if this is the best place to ask this question, and sorry if it isn't ... are there any plans to allow ffmpeg to play .dss or .ds2 files?
[12:55] <the-erm> I run linux ... if that makes a difference.
[12:57] <the-erm> I guess I should be more specific ... (DSS) Digital Speech Standard
[13:10] <rcombs> the-erm: you're welcome to write a decoder!
[13:10] <the-erm> That sounds like a wonderful idea ... if I could find the spec on it.
[13:11] <the-erm> oh .. and knew c or c++.
[13:11] <rcombs> yeah, both of those things tend to help :P
[13:49] <J_Darnley> the-erm: all I can really suggest is making a feature request on trac (that will be promptly ignored)
[13:49] <the-erm> :)
[13:50] <J_Darnley> You can then also post a couple of samples for people to use and any other information about the format you might have.
[13:51] <rcombs> and then people can ignore those too!
[14:03] <ubitux> nevcairiel: it doesn't turn negative, it is from the beginning
[14:04] <ubitux> like, it's the first audio packet that enters in this loop, and it has dts=-57
[14:08] <ubitux> first video* packet, sorry
[14:09] <ubitux> it basically pulls 57750 audio packets from /dev/zero before it writes the first video packets
[14:38] <cone-84> ffmpeg.git 03Michael Niedermayer 07master:a39201818fea: avformat/matroskadec: Fix cluster parsing loop which gathers seek information
[15:16] <cone-84> ffmpeg.git 03Neil Birkbeck 07master:ad5c43bb36b0: avfilter/vf_idet: Fixes issue with idet not flushing last frame.
[15:44] <ubitux> michaelni: i think i'll need some little guidance for #4051
[15:45] <ubitux> afaict, after 2 video frames are pushed, ffmpeg is going to pull infinitely from the audio source
[15:46] <ubitux> then pushing 40 more video frames (while audio is still grabbed at "full speed") causes the derp
[15:46] <ubitux> i wonder why ffmpeg is not blocking the audio input somehow
[15:47] <ubitux> (even with -threads 1 everywhere)
[16:17] <michaelni> ubitux, i have a fix for the assert but didnt yet look at why the audio runs crazily ahead
[16:20] <ubitux> oh? cool thank you
[16:29] <Daemon404> grepping source and reading git history to figure out how the hell to use ffmpeg.c options...
[16:29] <Daemon404> classic.
[16:33] <ubitux> michaelni: it's actually related to threading somehow
[16:34] <ubitux> i can't control it with the command line, but with a build with --disable-pthreads it seems to help
[16:34] <cone-84> ffmpeg.git 03Michael Niedermayer 07master:ec6a5fc6cca2: avformat/mux: Fix assertion failure due to max_interleave_delta and poor input
[16:34] <ubitux> at least the infinite pulling of audio
[16:37] <ubitux> thank you for this!
[17:14] <michaelni> ubitux, patch posted to fix the other issue in 4051
[17:16] <reynaldo> ubitux: what ever happened with Eejya Singh's (akira4) OPW quali? did she vanished?
[17:16] <reynaldo> vanish/
[17:18] <ubitux> reynaldo: she's here :)
[17:18] <ubitux> reynaldo: i'm going to push her patch today
[17:18] <ubitux> i'm just waiting for 24 hours after the stl sample update
[17:18] <reynaldo> \o/
[17:18] <reynaldo> congrats
[17:18] <ubitux> but the patch is ok since about 48h
[17:19] <reynaldo> ubitux: maybe if you could reserve some time to update her info on https://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12-Qualis
[17:19] <reynaldo> date given / status / first submited and stuff
[17:19] <reynaldo> along the quali details that is
[17:20] <reynaldo> should help out ranking her latter on
[17:20] <ubitux> right
[17:20] <ubitux> i'll do it today
[17:20] <reynaldo> cool
[17:20] <reynaldo> thanks a lot clement
[17:20] <ubitux> when i'm home
[17:45] <rcombs> wm4: heh, I can't tell if the slow seek issue is happening without looking at it in a debugger because my SSD's too fast
[17:45] <wm4> yeah, it just requires parsing the file up until the seek target
[17:46] <wm4> even on a hdd this can be pretty fast
[17:48] <Daemon404> strace master race
[22:31] <cone-84> ffmpeg.git 03Michael Niedermayer 07master:9bdb7a6c1a8b: ffmpeg: avoid endlessly reading from one input while another is unavailable
[22:31] <cone-84> ffmpeg.git 03Christophe Gisquet 07master:0e8bfd3c9347: avcodec/tiffenc: properly compute packet size
[22:45] <cone-84> ffmpeg.git 03Lukasz Marek 07master:f524d2e47c05: ffserver: move configuration code to separate file
[22:45] <cone-84> ffmpeg.git 03Lukasz Marek 07master:d2f8ca4959a1: ffserver_config: clean up variable initialization
[22:45] <cone-84> ffmpeg.git 03Lukasz Marek 07master:27c61c32cee4: ffserver_conf: factorize parse function per config tag
[23:15] <ubitux> what's the "date given" column on the opw page?
[23:16] <ubitux> the date the task was given to the student?
[23:17] <cone-84> ffmpeg.git 03Eejya Singh 07master:6dc99fdf0e43: Added STL demuxer and decoder
[23:19] <ubitux> what's the score thing?
[23:21] <ubitux> reynaldo: anyway, wiki updated
[23:52] <ubitux> i like that libavfilter email.
[23:54] <J_Darnley> Yeah, it is pretty thorough.  One thing I think it lacks is replacing the expressions and escaping on the command line.
[23:54] <ubitux> escaping is solved with the filtergraphs in a file
[23:55] <ubitux> replacing the expressions, well, dunno, i don't think it's the most troublesome problem
[23:56] <J_Darnley> Are you still certain that you won't have more than 1 consecutive '\' if using a file?
[23:57] <ubitux> i don't know
[23:57] <ubitux> i actually rarely use \ even in the cmd line
[23:57] <ubitux> with the 2 levels of quotes, only drawtext caused me issues
[23:57] <J_Darnley> Well this "drawtext=text=this is a \\\'string\\\'\\: may contain one\, or more\, special characters" is in the docs
[23:57] <ubitux> it's indeed a problem with the windows shell when you're trying to specify C:\foo\bar typically, but in this case the file form should be enough
[23:58] <J_Darnley> with an even worse one juts below for the command line
[23:58] <ubitux> you can put the text in a file as well
[23:58] <J_Darnley> Yeah, I get that but it still looks like that would require 3 consecutive /
[23:59] <J_Darnley> uh \
[23:59] Action: J_Darnley realizes he is bike shedding
[00:00] --- Thu Oct 23 2014