[Ffmpeg-devel-irc] ffmpeg-devel.log.20161023

[burek]

[00:01:13 CEST] <cone-176> ffmpeg 03Philip Langdale 07master:ee7d6738ca69: avcodec/cuvid: Allow reinitialization of decoder
[00:12:34 CEST] <cone-176> ffmpeg 03Carlos Fernandez 07master:728ccae8a2c9: lavf/mpegts: add missed fixes to scte35 section callback
[00:23:44 CEST] <BodecsB> Hi! May I ask about avf_concat?
[00:38:22 CEST] <Compn> BodecsB : sure, but this channel is for development, if you have a bug report ask in #ffmpeg
[00:40:20 CEST] <BodecsB> I am on writing a new feture into it, namely handling a command.
[00:41:11 CEST] <BodecsB> I would like to ask about flush_segment() function.
[00:41:22 CEST] <Chloe> Just ask
[00:41:59 CEST] <BodecsB> I see it handles the cases when audio streams are shorter than video stream by send_silence() funtion.
[00:43:29 CEST] <BodecsB> I tested its behaviour I in cases when video stream(s) are shorter it repeats the last video frame. But I do not see the handling of this case. Which part handles this?
[00:44:33 CEST] <BodecsB> Or it is hanled by the last video_buffer sink in filter chain?
[00:54:44 CEST] <BodecsB> I see I misstyped some character. So my real question is: What happens when video stream is shorter than audio stream? 
[03:24:05 CEST] <cone-054> ffmpeg 03Zhou Xiaoyong 07master:b9cd9226609b: avutil/mips: loongson add mmi utils header file
[03:24:05 CEST] <cone-054> ffmpeg 03Zhou Xiaoyong 07master:89ec4adad6cb: avcodec/mips: loongson optimize mmi load and store operators
[11:58:13 CEST] <cone-279> ffmpeg 03Andreas Cadhalpun 07master:2506a7cc09ca: faq: use relative links to own documentation
[13:07:17 CEST] <cone-279> ffmpeg 03Michael Niedermayer 07master:051517648b00: avutil/x86/emms: Document the emms_c() vs alloc/free relation.
[17:45:45 CEST] <ubitux> in the end, i'm adding the subtitles frame to lavfi because it simplifies everything
[17:45:56 CEST] <ubitux> it's just a shitton of glue code everywhere
[19:39:05 CEST] <RiCON> BtbN: any idea what's causing this: https://i.fsbn.eu/s5zr.txt ?
[19:39:39 CEST] <BtbN> RiCON, did you rebase the set yourself?
[19:39:44 CEST] <BtbN> Or apply the one from the ml?
[19:39:45 CEST] <RiCON> yes
[19:39:55 CEST] <RiCON> git am'd the bundle with the set
[19:39:57 CEST] <BtbN> there was a new patch for cuvid in between
[19:40:04 CEST] <BtbN> that uses a cuvid function
[19:40:11 CEST] <BtbN> I'll rebase the set, one moment
[19:42:12 CEST] <RiCON> oh, philipl's commit, right
[19:43:05 CEST] <BtbN> https://github.com/BtbN/FFmpeg/
[20:07:29 CEST] <RiCON> BtbN: yes, that fixed it
[20:19:42 CEST] <cone-539> ffmpeg 03Clément BSsch 07master:58672347cb20: lavfi: remove 2 unused lavc includes
[20:38:19 CEST] <BtbN> RiCON, for all the warnings... I have no idea how to make MSVC happy there.
[20:38:45 CEST] <BtbN> Or rather, gcc on windows.
[20:38:59 CEST] <BtbN> There is no typeof() in C.
[20:40:47 CEST] <RiCON> http://stackoverflow.com/a/27632640 ?
[20:41:35 CEST] <BtbN> GetProcAddress returns a FARPROC
[20:41:52 CEST] <BtbN> And it aparently can't be implicitly casted to a function pointer without a warning.
[20:42:10 CEST] <BtbN> So I'd have to add a cast to the correct type of the function to each single one.
[20:42:20 CEST] <BtbN> Which would massively blow up the source
[21:09:55 CEST] <jamrial_> BtbN: compat/w32pthreads.h just casts GetProcAddress to void*
[21:10:10 CEST] <jamrial_> and generates no warnings
[21:10:19 CEST] <BtbN> hm, I think I tried that already?
[21:10:23 CEST] <BtbN> I'll try that
[21:33:40 CEST] <BtbN> yes, seems to do the trick. Guess I messed something up the last time.
[22:29:55 CEST] <atomnuker> michaelni: got any advice on how to check if a forward float DCT is correct?
[22:30:21 CEST] <atomnuker> I'm writing an opus encoder and I've modified the imdct15 to apparently do a forward transform
[22:31:07 CEST] <atomnuker> but checking the output is difficult since I don't have a full encoder, the output isn't normalized and I don't really know an easy way to play fwritten() float samples back
[22:38:09 CEST] <michaelni> atomnuker, i think theres some mdct test in tests/fft.c, maybe that can be used
[00:00:00 CEST] --- Mon Oct 24 2016