[Ffmpeg-devel-irc] ffmpeg-devel.log.20170206

burek burek021 at gmail.com
Tue Feb 7 03:05:03 EET 2017


[01:02:24 CET] <cone-363> ffmpeg 03Michael Niedermayer 07master:1e5cfad57e88: avcodec/pthread_frame: Check av_packet_ref() for failure
[01:02:25 CET] <cone-363> ffmpeg 03Michael Niedermayer 07master:dfc6e30cd4b9: ffmpeg: Remove redundant null check
[03:58:03 CET] <Compn> this may also be an easy thing to try for a new developer, https://trac.ffmpeg.org/ticket/3720
[04:54:32 CET] <Compn> carl : can this bug be closed ? https://trac.ffmpeg.org/ticket/2749
[05:03:35 CET] <Compn> https://trac.ffmpeg.org/ticket/2377 may also be a good new demuxer for beginner developer
[06:00:27 CET] <Compn> possibly another good new jpg format for beginner dev https://trac.ffmpeg.org/ticket/4728
[06:15:45 CET] <sware> is there something I can reference on how ffmpeg picks a codec internally when it's initialized from chrome?
[06:15:50 CET] <sware> er, picks a decoder*
[06:16:04 CET] <Compn> say what
[06:16:40 CET] <Compn> chrome iirc only enables a few codecs from ffmpeg
[06:16:43 CET] <sware> trying to figure out how ffmpeg picks the decoder used when it's initialized from the chrome glue code
[06:16:58 CET] <sware> yes but some codecs have multiple decoders including hardware accelerated decoding
[06:17:36 CET] <Compn> i figure it would be up to chrome
[06:17:44 CET] <Compn> not part of ffmpeg ?
[06:17:49 CET] <sware> doesn't seem to be.
[06:17:58 CET] <Compn> what file are you reading ?
[06:18:11 CET] <BlueCrypt> Is the multi threading support/abstraction handled by each (en|de)coder?
[06:18:12 CET] <sware> the codec is, but not the decoder
[06:18:13 CET] <Compn> did you ask in #chromium ?
[06:18:32 CET] <sware> the chromium channels are completely dead. 500 people and no responses
[06:19:08 CET] <Compn> ffmpeg does not choose multithreading , nor hw decoding by itself
[06:19:14 CET] <Compn> the user (or in this case, chrome) does it
[06:19:47 CET] <BlueCrypt> Where does the actual multihreading code reside? I'm looking in mpeg4videoenc.c but that doesn't look right
[06:20:39 CET] <Compn> thread.h ?
[06:20:50 CET] <Compn> er
[06:21:35 CET] <BlueCrypt> so ff_thread is the abstraction that handles pthread/w32thread?
[06:22:03 CET] <Compn> you read doc/multithreading.txt ?
[06:22:09 CET] <Compn> and pthread.c
[06:22:58 CET] <sware> ah I think I found it actually. 
[06:22:59 CET] <Compn> i'm not exactly sure :)
[06:23:30 CET] <sware> avcodec_find_decoder. Down the rabbit hole we go
[06:23:40 CET] <Compn> sware : dont do it ;)
[06:23:51 CET] <sware> I'm already in it too deep :P
[06:24:18 CET] <sware> the last time i messed with ffmpeg was when I ported it to the 360. Been years. I wonder how much has changed.
[06:24:48 CET] <Compn> BlueCrypt : afaik some codecs need specific hacks to properly be threaded. i think this is due to the number of b-frames and other special out of order frames and predictions
[06:25:45 CET] <Compn> sware : i didnt even realize libavcodec had more than 1000 files...  :)
[06:26:12 CET] <sware> yeah most are very similar
[06:28:28 CET] <Compn> what are you working on now?
[06:30:28 CET] <sware> nothing special, just a video player
[06:30:36 CET] <sware> oh well thanks for rubber duckying. See ya
[06:31:16 CET] <Compn> BlueCrypt : what are you working on ?
[06:32:42 CET] <BlueCrypt> Compn: Trying to get decklink to give up it's dependency on pthread
[06:33:00 CET] <Compn> ah
[06:33:37 CET] <Compn> its an external device right? 
[06:33:43 CET] <Compn> pthread might be required... ?
[06:34:22 CET] Action: Compn should stop talking about stuff he hasnt a clue about
[06:34:22 CET] <Compn> ehe
[06:35:30 CET] <BlueCrypt> It's possible, I wont know till I dig deeper
[06:39:48 CET] <BlueCrypt> It looks like the Linux SDK has include <pthread.h> but the Windows one does not
[09:34:24 CET] <cone-326> ffmpeg 03Chris Cunningham 07master:ac25840ee328: lavf/matroskadec: fix is_keyframe for early Blocks
[10:35:42 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:aa20863f4400: avcodec/mjpegdec: Check for for the bitstream end in mjpeg_decode_scan_progressive_ac()
[10:35:43 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:7323a8ab2970: avcodec/dca_lbr: Fix off by 1 error in freq check
[10:35:44 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:d399f25bd1e6: avcodec/interplayvideo: Move parameter change check up
[10:35:45 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:7e1d9d25fe04: avcodec/pngdec: Check trns more completely
[10:35:46 CET] <cone-326> ffmpeg 03James Almer 07release/3.2:87a47c67a697: configure: bump year
[10:37:17 CET] <wm4> 3.2 could probably also use that matroska fix
[10:53:26 CET] <JEEB> nice, subtitle overlay issue seems to have been fixed already. it properly re-inits at a specific point now. that makes things simple :)
[10:53:38 CET] Action: JEEB finally got to dumping a sample
[10:58:49 CET] <cone-326> ffmpeg 03Michael Niedermayer 07master:4fcdc9f35972: avfilter/vf_midequalizer: Remove duplicate include
[11:30:03 CET] <wm4> sigh, looking at something that looks like fate failures due to rounding differences in subtitle encoding
[11:35:54 CET] <wm4> wow avformat_transfer_internal_stream_timing_info exists
[11:36:17 CET] <wm4> quite literally a shitty hack to make ffmpeg.c happy
[11:37:00 CET] <JEEB> lul
[12:10:23 CET] <cone-326> ffmpeg 03Michael Niedermayer 07master:e248522d1b0d: avcodec/movtextdec: Fix decode_styl() cleanup
[12:21:29 CET] <wm4> emu edge and lowres shit again...
[12:25:04 CET] <cone-326> ffmpeg 03Chris Cunningham 07release/3.2:d88493c02b4a: lavf/matroskadec: fix is_keyframe for early Blocks
[12:25:05 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:44ce16b7f90e: avcodec/movtextdec: Fix decode_styl() cleanup
[13:05:42 CET] <cone-326> ffmpeg 03Michael Niedermayer 07release/3.2:68ed682710f1: Update for 3.2.3
[13:32:11 CET] <wm4> how can I debug ffmpeg again?
[13:32:18 CET] <wm4> you need several magic configure options or so
[13:32:27 CET] <wm4> to disable optimization while not making likning fail
[13:35:04 CET] <BtbN> -O0 -fdce?
[13:36:43 CET] <wm4> gcc: error: '-O0': No such file or directory
[13:36:51 CET] <wm4> I'm not sure what it's trying to tell me
[13:37:15 CET] <BtbN> did you put it after -o?
[13:37:35 CET] <wm4> I used --extra-cflags
[13:38:17 CET] <BtbN> Or -Og -fdce
[13:38:56 CET] <BtbN> But it seems like something else is messed up if it tries to read/write that option as a file
[13:39:58 CET] <wm4> it adds those as first arguments to the gcc command
[13:39:58 CET] <jkqxz> I use "--enable-debug --disable-optimizations"; no cflags.
[13:40:05 CET] <wm4> and then fails the compiler test
[13:42:03 CET] <wm4> that worked better
[13:42:40 CET] <BtbN> are you sure you are using gcc?
[13:42:55 CET] <wm4> yes?
[13:43:09 CET] <BtbN> I don't see how on it would ever end up interpreting -O0 as a file
[14:11:07 CET] <wm4> any tips how to debug these opaque avformat.mak tests?
[14:41:54 CET] <wm4> man these filter data flow hacks in ffmpeg.c
[14:43:53 CET] <wm4> ubitux: I'm working on porting 3e265ca and a3a0230, but I expect it'll take a few days until it's work (if I can make it work at all)
[15:09:37 CET] <ubitux> wm4: oh cool, thanks
[15:33:18 CET] <ryantheseer> Anyone know how to compile FFmpeg as a single shared library (.dll) instead of one .dll for each library (avcodec, avformat, avfilter, etc.)?
[15:35:29 CET] <wm4> that's not possible with the build scripts provided by us
[15:35:57 CET] <ryantheseer> Hmm, okay.
[15:38:10 CET] <ryantheseer> I'm trying to integrate a patch to FFmpeg with OpenCV, but I'm having trouble compiling FFmpeg from the source in a way that the OpenCV scripts use.
[15:38:36 CET] <wm4> and they expect a single DLL?
[15:40:38 CET] <ryantheseer> They have checked in to GitHub a single file called opencv_ffmpeg.dll (with a 32-bit and 64-bit version), which is much smaller in size than the .dlls that are produced when I run make from the FFmpeg source code.
[15:41:12 CET] <wm4> lol, such shitty practices
[15:41:22 CET] <ryantheseer> I know, right!? :S
[15:41:35 CET] <wm4> they probably have some build scripts hidden somewhere
[15:41:40 CET] <wm4> (I don't know)
[15:41:53 CET] <wm4> possible that they disabled tons of features to shrink it
[15:42:05 CET] <ryantheseer> That's probably what they're doing.
[15:42:41 CET] <ryantheseer> Anyway, I'll try digging a bit more. Thanks!
[18:39:15 CET] <nevcairiel> wm4: I got that thing almost merge some month back until I hit a wall with lavfi data flow, a lot of fun in there 
[18:41:47 CET] <wm4> yeah
[18:52:14 CET] <cone-352> ffmpeg 03Michael Niedermayer 07n3.2.3:HEAD: avcodec/movtextdec: Fix decode_styl() cleanup
[19:50:23 CET] <cone-352> ffmpeg 03Hendrik Leppkes 07master:a6cee50fa206: configure: use dashes instead of slashes in lib.exe invocation
[20:07:27 CET] <llogan> why Cinepak?
[20:07:50 CET] <llogan> what's next? Indeo?
[20:08:25 CET] <llogan> I remember an early demonstration of Indeo where they projected the video onto a dummy. "3D".
[20:09:54 CET] <tmm1> anyone know of something like wireshark but for h264 streams
[20:13:27 CET] <kierank> h264dum
[20:13:28 CET] <kierank> p
[20:15:00 CET] <jkqxz> The reference decoder (JM) with the bitstream trace mode enabled.
[20:20:23 CET] <tmm1> thanks!
[21:24:04 CET] <rcombs> kierank: why does this thing depend on libmxf
[21:24:13 CET] <kierank> it's a standalone app afaik
[21:24:18 CET] <kierank> that's used as a debug tool in libmxf
[21:55:05 CET] <TD-Linux> llogan, what, are you against supporting Sega CD video in proprietary enterprise deployments?
[21:55:54 CET] <JEEB> lol
[21:56:07 CET] <JEEB> this one game I got does kind of make me want to poke at bink2
[21:56:53 CET] <sware> anyone here know what would cause video to jump forward and backwards constantly but keep playing? like it plays a chunk, then plays it again, then plays the next chunk and plays it again then the next. Like it's constantly repeating playback? Maybe a buffer issue?
[22:01:15 CET] <llogan> wrong channel. ffmpeg-devel is for discussions related to the development of FFmpeg and for making fun of Sega CD video in proprietary enterprise deployments
[22:04:06 CET] <sware> well it kind of involves the development if I have to fix the decoder? 
[22:37:45 CET] <Compn> sware : i've seen that happen a lot to people who dont know what they are doing with libavcodec :D
[22:44:07 CET] <sware> compn: in terms of encoding?
[22:44:52 CET] <Compn> encoding? no, i mean decoding
[22:45:53 CET] <Compn> if you have example command line that shows error... go for it
[22:46:13 CET] <sware> there are no errors
[22:47:58 CET] <sware> that's why I'm trying to ask experienced people, so I can identify the issue and then fix it myself
[23:10:51 CET] <sware> Compn: If you've seen this before and it's because of a bad configuration, can you explain what it is? 
[23:33:10 CET] <BBB> sware: how many frames does it jump back/forward? and what codec?
[23:33:43 CET] <BBB> also, that cinepak discussion just went from frowning to absurdist
[23:40:31 CET] <ubitux> sorry, maybe i shouldn't have participated in it
[23:47:14 CET] <BBB> ubitux: dont be sorry, nothing bad happened
[23:47:43 CET] <BBB> ubitux: the absurdist isnt you or wm4, its that dude continuously insisting that his patch is the most delicious thing that we dont wanna miss out on
[23:48:12 CET] <BBB> imagine just for a second that a library thats pretty much at system level would start responding to getenv()& its absurd
[23:48:30 CET] <BBB> (start responding to env variables using getenv(), I guess)
[23:48:49 CET] <ubitux> like str functions in libc responding to LANG=...? :-°
[23:49:49 CET] <BBB> yeah, like that :)
[23:51:49 CET] <jkqxz> What is all this API rubbish for?  It just gets in the way.  The decoder should put the decoded frame back into the environment using setenv() so that it can be used by programs which don't implement the API properly and can't be changed.
[23:54:37 CET] <jkqxz> ... using limited-range RGB to avoid zeroes, since the environment contains strings.
[23:56:44 CET] <llogan> why does he even give a shit about cinepak in the first place?
[00:00:00 CET] --- Tue Feb  7 2017


More information about the Ffmpeg-devel-irc mailing list