[Ffmpeg-devel-irc] ffmpeg-devel.log.20181018

[burek]

[00:04:39 CEST] <philipl> BtbN: I'm going to rework your change a bit to reduce the number of #ifdefs. I'm guessing you don't care if I do.
[00:04:54 CEST] <BtbN> I already reduced them quite a bit
[00:06:11 CEST] <philipl> eg: We can have one get_external_foo method.
[00:16:35 CEST] <philipl> https://github.com/philipl/mpv/commit/ad4e136af4aeed92e828f1c3b99b11754b5e7446
[00:16:43 CEST] Last message repeated 1 time(s).
[00:16:43 CEST] <philipl> https://github.com/philipl/mpv/commit/9d0e73adcf795b90b80aaef4e9791a423af7c297
[00:16:46 CEST] <philipl> rather
[00:17:16 CEST] <philipl> BtbN: My modified versions of your changes.
[00:19:35 CEST] <BtbN> the memory_fd extension can totally exist on Windows btw.
[00:19:41 CEST] <BtbN> ifdefing it out seems wrong
[00:19:59 CEST] <philipl> It can exist, it won't be used.
[00:20:04 CEST] <philipl> So the symmetry seems desirable.
[00:27:50 CEST] <BtbN> If that doesn't matter, sure
[00:29:56 CEST] <cone-421> ffmpeg 03Carl Eugen Hoyos 07master:e90d4e92729c: lavc/mjpegdec: Support 2:3 subsampling.
[00:43:24 CEST] <BtbN> The LunarG Vulkan SDK actually comes with a 100MB .lib file for shaderc
[00:43:29 CEST] <BtbN> but it's for MSVC builds only
[00:43:40 CEST] <BtbN> trying to link gives me several 100MB of missing references on link
[00:43:59 CEST] <BtbN> anyway, managed to build one myself now with the posix enabled toolchain
[00:44:08 CEST] <JEEB> congrats
[00:46:28 CEST] <BtbN> philipl, if I do "./mpv.exe --gpu-api=vulkan --vo=gpu --hwdec=nvdec ..." and I get "VO: [gpu] 1920x1080 cuda[nv12]" it means it works, right?
[00:46:47 CEST] <JEEB> yes, it means you're getting nv12 from hte "cuda" decoder
[00:47:05 CEST] <JEEB> lavc doesn't return you nv12 so it's definitely some sort of hwdec
[00:47:07 CEST] <BtbN> More interested if it actually render them via vulkan, or does sw download and re-upload or something
[00:47:19 CEST] <JEEB> --log-file=mpv_sucks.log
[00:47:27 CEST] <JEEB> higher verbosity and timestamps etc
[00:48:06 CEST] <JEEB> and if you just want the terminal stuff, then --msg-level=module=verbose where module is the module you care about (vd maybe?)
[00:48:38 CEST] <JEEB> (or debug instead of verbose, which is what log-file also uses)
[00:49:11 CEST] <BtbN> "[   0.758][d][vo/gpu/cuda-nvdec] vk_external_handle[1][0]: 0000000000000994 1036800 2073600" yeah, it's doing what it should. And it looks reasonable
[00:49:34 CEST] <JEEB> cool
[00:50:58 CEST] <philipl> BtbN: yeah, you succeeded!
[02:34:09 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:4e19cfcfa394: avformat/utils: Fix integer overflow in discontinuity check
[02:34:10 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:079d1a7175c4: avformat/utils: Never store negative values in last_IP_duration
[02:34:11 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:c6282141cba2: avcodec/ra144: Fix integer overflow in add_wav()
[02:34:12 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:25eb10d941f3: avcodec/ilbcdec: Limit indexes in create_augmented_vector()
[02:34:13 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:d01788665767: avcodec/ilbcdec: Fix multiple invalid left shifts
[02:34:14 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:d7dbad12f820: avcodec/ilbcdec: Fix multiple integer overflows
[02:34:15 CEST] <cone-421> ffmpeg 03Michael Niedermayer 07master:f72b9904fefa: avcodec/h264_cavlc: Check mb_skip_run
[02:37:25 CEST] <tmm1> i'm trying to figure out where in the code the h264 parser splits two fields into separate packets
[04:18:01 CEST] <tmm1> found http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/175525 which could be adapted to a bsf
[07:33:11 CEST] <cone-330> ffmpeg 03Karthick Jeyapal 07master:444cdb422d3c: avformat/dashenc: Dont ignore the codec tag from codec parameters
[10:08:20 CEST] <durandal_1707> what about adding (a)replace filters? they would replace (samples/)frames from one stream with another stream for some duration in random position in stream
[10:24:31 CEST] <cone-330> ffmpeg 03Paul B Mahol 07master:bf5bdbdb9267: avfilter/af_adelay: do not crash if allocation of chandelay failed
[11:01:53 CEST] <ubitux> durandal_1707: ablend?
[11:01:56 CEST] <ubitux> (hello)
[11:02:54 CEST] <ubitux> (with mode=overlay for replacing or sth)
[11:03:46 CEST] <ubitux> because you may want to mix it and not just replace it completely
[11:03:59 CEST] <ubitux> ablend is a bad name though
[11:08:10 CEST] <durandal_1707> ubitux: you are back!
[11:08:22 CEST] <ubitux> yeah i should, pretty soon
[11:08:32 CEST] <ubitux> i'm sorry for leaving again, i thought i had time, but i didn't
[11:08:49 CEST] <ubitux> but i finally moved home after a few months, fucking pain
[11:20:53 CEST] <JEEB> I moved two years ago and now they're cutting all the trees :<
[11:29:34 CEST] <atomnuker> I haven't moved yet and they've probably already ruined wherever I would move to
[11:38:56 CEST] <cone-330> ffmpeg 03Paul B Mahol 07master:0ce7005d34e2: avfilter/vf_lensfun: use av_freep
[11:38:57 CEST] <cone-330> ffmpeg 03Paul B Mahol 07master:2d9c290a2ffc: avfilter/vf_lensfun: use av_malloc_array()
[12:14:56 CEST] <durandal_1707> NEW FILTER in nearest filter factory! chromahold
[14:13:21 CEST] <ubitux> durandal_1707: http://www.ipol.im/pub/art/2018/23/
[15:13:02 CEST] <victorqueiroz> What should I do when I receive "missing picture in access unit with size 24" error?
[18:41:54 CEST] <cone-330> ffmpeg 03Aman Gupta 07release/4.0:380bc8585c75: Revert "avcodec/cbs_h264: silence errors about end_of_seq nalus"
[19:02:31 CEST] <thebombzen> I'm trying to figure out how to make "-r" work the same way for streamcopy as it does for decoding but I cant' make heads or tails of ffmpeg.c's process_input_packet
[19:02:38 CEST] <thebombzen> since I've never worked on ffmpeg.c before
[19:03:08 CEST] <thebombzen> where does it check for the value of "-r" as an input option when decoding?
[19:05:23 CEST] <BtbN> isn't -r just some magic that makes the demuxer in question return EAGAIN until the timestamp isn't too early anymore?
[19:18:44 CEST] <JEEB> not that I think...
[19:20:38 CEST] <JEEB> thebombzen: fftools/ffmpeg_opt.c:        "set frame rate (Hz value, fraction or abbreviation)", "rate" },
[19:20:48 CEST] <JEEB> see which struct member that goes into
[19:21:00 CEST] <thebombzen> aight thx
[19:21:38 CEST] <JEEB> BtbN: IIRC -r with encoding just uses ffmpeg.c's vsync to duplicate/drop decoded frames
[19:21:58 CEST] <BtbN> oh, -r, not -re, nvm
[19:22:08 CEST] <JEEB> yea, -re sleeps
[19:22:29 CEST] <JEEB> also -re is very fun when you find someone trying to use it with MPEG-TS or so
[19:22:35 CEST] <JEEB> suddenly wild discontinuity
[19:22:41 CEST] <JEEB> and suddenly the thing is waiting for N hours :P
[19:23:15 CEST] <kierank> lol
[19:52:22 CEST] <jamrial> durandal_1707: be consistent with using float constants and standard float c functions in your sinc source patch
[19:53:06 CEST] <jamrial> i see powf() but then sin(), plus lots of constants missing the f suffix while others don't
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:d7731ea9d43c: avcodec/proresenc_aw : use scan table from prores_data file
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:1a45b12cdc41: avcodec/proresenc_aw : use AVframe primaries, transfert, colorspace for frame header instead of default (unknown, unknown, Rec601)
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:cea5e90bde4a: avcodec/proresenc_aw : use qp close to the official encoder
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:98637be753df: avcodec/prores_enc : not calculate dct a each quantif search step
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:c51b36ec8f70: avcodec/proresenc_aw : add support for prores 444
[21:43:14 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:e832d769f465: avcodec/proresenc_aw : use for frame flag in the header the same value than the official encoder
[21:44:52 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:658bbc006002: swscale/x86/rgb2rgb.asm : add Ivo Van Poorten name to the top of the file
[21:44:53 CEST] <cone-571> ffmpeg 03Martin Vignali 07master:db4771af8196: swscale : add YA16 LE/BE output
[22:18:01 CEST] <j-b> 'morning
[22:19:11 CEST] <durandal_1707> sweet dreams
[22:19:36 CEST] <atomnuker> how commercial is demuxed on a scale of the fsf to apple?
[22:22:30 CEST] <j-b> atomnuker: 6/10
[22:22:42 CEST] <j-b> 10/10 being apple
[22:23:37 CEST] <durandal_1707> VDD is 3/10 ?
[22:23:47 CEST] <atomnuker> ibc is 8/10?
[22:24:03 CEST] <atomnuker> would nbc then be 10/10?
[22:24:26 CEST] <j-b> durandal_1707: sounds about right
[22:24:33 CEST] <j-b> fsf being 1/10
[22:24:40 CEST] <j-b> NAB would be 9/10
[22:26:40 CEST] <atomnuker> wonder what would 10/10 be
[22:26:49 CEST] <atomnuker> oh it has to be that apple dev event
[22:27:06 CEST] <j-b> WWDC
[22:27:10 CEST] <j-b> or Google I/O
[22:27:55 CEST] <atomnuker> +microsoft's thing
[22:28:23 CEST] <j-b> Microsoft thing is less bad than WWDC
[22:28:27 CEST] <j-b> it's way more technical.
[22:30:11 CEST] <wbs> the actual sessions at WWDC are rather technical
[22:31:57 CEST] <j-b> the ambiant around them are not.
[22:33:02 CEST] <wbs> especially the keynote
[22:37:46 CEST] <j-b> and the groupies...
[00:00:00 CEST] --- Fri Oct 19 2018