[Ffmpeg-devel-irc] ffmpeg-devel.log.20190524

[burek]

[00:00:00 CEST] --- Fri May 24 2019
[05:49:29 CEST] <christo_m> im wondering if its possible to use force_original_aspect_ratio inside the -filter_complex option, but i can't seem to figure it out
[05:49:33 CEST] <christo_m> -vf "scale=1280:1280:force_original_aspect_ratio=decrease,pad=1280:1280:(ow-iw)/2:(oh-ih)/2,setsar=1" im trying to replicate this inside -filter_complex, because im unable to use both vf and filter_complex without errors.
[05:50:36 CEST] <cone-223> ffmpeg 03Xuewei Meng 07master:023ea5e360cb: libavfilter/dnn_native: Add support of dilated convolution in dnn_native.
[08:12:38 CEST] <cone-223> ffmpeg 03Ruiling Song 07master:1d74150a7dac: lavfi/opencl: add nlmeans_opencl filter
[12:56:36 CEST] <vtorri99978666> hello
[12:57:11 CEST] <vtorri99978666> is it normal that, when compiling ffmpeg 4.1.3 on Windows (MSYS2 + mingw-w64), there are a lot of warning ? 
[12:57:21 CEST] <vtorri99978666> especially deprecated functions which used
[12:59:11 CEST] <vtorri99978666> which are* used
[13:00:38 CEST] <iive> warnings are normal
[13:00:53 CEST] <vtorri99978666> using deprecated functions is not, imho
[13:01:24 CEST] <vtorri99978666> also some printf formats are not correct on Windows
[13:01:42 CEST] <vtorri99978666> some comparison like an unsigned int < 0
[13:02:04 CEST] <nevcairiel> feel  free to send patches
[13:02:51 CEST] <vtorri99978666> i'm waiting for 4.2 first
[13:19:43 CEST] <BtbN> I'm really not sure about this nvenc patches. Imo that bunch of code should be in a filter, not in an encoder.
[13:21:16 CEST] <mkver> The format errors that get emitted during compilation of hlsenc could be easily solved by using #ifdef.
[13:27:40 CEST] <mkver> And I don't see a reason to wait for 4.2 for patches.
[18:40:41 CEST] <cone-756> ffmpeg 03Paul B Mahol 07master:a9b5af0d1932: avfilter/vf_mix: use time_base from framesync
[18:40:41 CEST] <cone-756> ffmpeg 03Paul B Mahol 07master:a1992637ace2: avfilter/vf_mix: set sar to outlink
[18:40:41 CEST] <cone-756> ffmpeg 03Paul B Mahol 07master:5073adcc42d7: avfilter/vf_xmedian: set sar to outlink
[19:24:21 CEST] <cone-756> ffmpeg 03Josh Allmann 07master:4cfc34d9a8bf: avfilter/vf_scale_cuda: reset frame size after acquiring from hwframe
[19:24:22 CEST] <cone-756> ffmpeg 03Timo Rothenpieler 07master:a6818d5bd009: compat/cuda/ptx2c: don't drop final newline
[19:29:20 CEST] <BtbN> I wonder how this never was an issue before. Both of these.
[19:30:07 CEST] <philipl> that's odd.
[20:15:26 CEST] <mkver> The definition of AV_INPUT_BUFFER_PADDING_SIZE contains "Note: If the first 23 bits of the additional bytes are not 0, then damaged MPEG bitstreams could cause overread and segfault."
[20:15:56 CEST] <mkver> Does this mean that every the padding (or at least the first 23 bits of it) always needs to be zeroed?
[20:16:10 CEST] <nevcairiel> Practically, yes, it should be
[20:17:28 CEST] <mkver> Ok. The Matroska demuxer currently violates this: Of all the packets in a lace, only the last packet gets zeroed padding.
[20:17:51 CEST] <mkver> The padding of the other packets is the start of the next packets.
[20:18:10 CEST] <mkver> Will have to fix this (after my current patches have been dealt with).
[20:21:00 CEST] <jamrial> lacing is used mainly with audio, right? i don't think mkvtoolnix adds more than one frame per block
[20:21:57 CEST] <nevcairiel> lacing is generally rather rare
[20:22:43 CEST] <nevcairiel> i wonder if it uses  it with truehd and its 0.833ms short frames
[20:22:49 CEST] <nevcairiel> to save on some overhead
[20:23:22 CEST] <mkver> mkvmerge uses lacing for all audio tracks except TrueHD (if I am not mistaken).
[20:23:36 CEST] <mkver> And it uses lacing only for audio frames.
[20:23:38 CEST] <nevcairiel> why exclude that one specifically
[20:23:46 CEST] <nevcairiel> it has like the smallest audio frames ever
[20:24:37 CEST] <mkver> mkvmerge does not declare every TrueHD frame to be a keyframe; only the sync frames get declared that way.
[20:24:50 CEST] <mkver> That's what makes it special.
[20:25:34 CEST] <mkver> In order not to mix TrueHD syncframes with non-syncframes in one lace, lacing gets disabled.
[20:25:54 CEST] <mkver> This has gigantic overhead, of course.
[20:26:02 CEST] <nevcairiel> how unfortunate, could actually save some space there
[20:27:18 CEST] <mkver> (Btw: Our Matroska demuxer often emits "File is broken, keyframes not correctly marked!" errors on such files of the first audio frame it encounters is not flagged as keyframe.)
[20:27:36 CEST] <mkver> if not of
[20:28:59 CEST] <Lynne> AC-4 also has P frames, but it doesn't seem it'll be popular on the net anytime
[20:34:11 CEST] <JEEB> the only samples of AC-4 I see outside of broadcast are the DASH-IF sample files
[20:52:28 CEST] <cone-756> ffmpeg 03Gyan Doshi 07master:50789e356d65: avformat/cache - delete cache file after closing handle
[22:30:33 CEST] <kierank> Lynne: maybe for object based audio
[22:30:50 CEST] <kierank> people will use it online
[22:43:59 CEST] <cone-756> ffmpeg 03Marton Balint 07master:f9271d015812: avformat/utils: fix stream ordering for program ID stream specifiers
[22:44:00 CEST] <cone-756> ffmpeg 03Marton Balint 07master:1b4f74ffe235: avcodec/scpr: fix checking ret value of decode_run_i
[22:44:01 CEST] <cone-756> ffmpeg 03Marton Balint 07master:f811ac52284a: avcodec/scpr3: fix checking ret value of decode_run_i
[22:44:02 CEST] <cone-756> ffmpeg 03Marton Balint 07master:87e0247a6c0f: avfilter/f_loop: fix video loop issues with 0 size or when size is bigger than input
[22:44:03 CEST] <cone-756> ffmpeg 03Marton Balint 07master:1eb54323c82f: avfilter/f_loop: warn about unset loop sizes
[00:00:00 CEST] --- Sat May 25 2019