[Ffmpeg-devel-irc] ffmpeg-devel.log.20190906

[burek]

[00:00:35 CEST] <rcombs> well yes, the kernel closes all of a process's fds when it exits
[00:00:43 CEST] <rcombs> (and then the P-state returns to normal)
[00:01:09 CEST] <BtbN> So it's only an issue if you have a long running process that sometimes uses nvenc/cuda?
[00:01:12 CEST] <rcombs> but I want to allow the GPU to return to a low-power state while my app's still running, after I've finished using it for something
[00:01:14 CEST] <rcombs> right
[00:01:35 CEST] <rcombs> which is a bit edgy, I suppose, so I'm not especially surprised it's buggy
[00:01:47 CEST] <BtbN> For nvidia contacts, just search ffmpeg-devel for the last mails froms @nvidia.com and send it to the last 2 people or so.
[00:02:17 CEST] <rcombs> I'll give that a go, thanks
[00:02:41 CEST] <BtbN> ygupta at nvidia.com used to be our usual contact, but I haven't heard from him in a while.
[03:15:03 CEST] <tmm1> what is the correct way to handle H264_NAL_PREFIX in cbs_h264_read_nal_unit
[04:26:46 CEST] <cone-896> ffmpeg 03Steven Liu 07master:6c593fee6335: avformat/http: add ff_http_get_shutdown_status api for check the status of shutdown
[04:26:46 CEST] <cone-896> ffmpeg 03Steven Liu 07master:f267a2ac9c74: avformat/hlsenc: avformat/hlsenc: reopen new http session for http_persistent
[06:30:33 CEST] <cone-896> ffmpeg 03Steven Liu 07master:07b4bf5f11e4: avformat/hlsenc: move the warning message from every segment upload to init part
[06:30:34 CEST] <cone-896> ffmpeg 03Steven Liu 07master:fe77cad3a412: avformat/hlsenc: fix code style
[06:47:42 CEST] <cone-896> ffmpeg 03Steven Liu 07master:db92a3e4630f: avformat/hlsenc: fix compiling error of hlsenc
[09:10:29 CEST] <VenomFK> Hi, I am planning a to add a new hardware based filter in ffmpeg. Is there any documentation which can guide me things to take care while adding a new hardware based filter?
[09:12:57 CEST] <JEEB> thre's a few of them now for various hwaccel-related things
[09:13:06 CEST] <JEEB> not sure if they're standardized between themselves
[09:14:40 CEST] <VenomFK> I have some propriety tool which performs some post processing on the raw data, which I want to integrate in ffmpeg for ease of my usage
[09:18:42 CEST] <JEEB> anyways, see stuff like cuda or opencl or vulkan etc under libavfilter
[09:19:39 CEST] <VenomFK> Thanks
[09:49:57 CEST] <Radiator> Hello erveryone. I have a question on the libav and mpv. For my work I have to decode a video stream from a camera as fast as possible (as close as I can to >50ms). To do so I had an example from my superior from mpv : mpv --no-cache --untimed --no-demuxer-thread --vd-lavc-threads=1 udp://224.0.1.1:10001/ 
[09:51:25 CEST] <Radiator> So I browsed the source code of mpv which is quite a maze and couldn't find anything as like which library call is used etc...
[09:52:17 CEST] <Radiator> Do you have by any chance an idea on how to recreate the behavior of mpv with no cache, untimed, and no demuxer thread ?
[09:52:50 CEST] <JEEB> I think it's all based on libavformat protocol and format options so you should be capable of
[09:52:56 CEST] <JEEB> it's not magic, after all
[09:53:11 CEST] <JEEB> by default libavformat favours some amount of buffering for stability
[09:53:56 CEST] <Radiator> Great I'll give it a shot !
[09:54:50 CEST] <Radiator> Another thing (might be wrong channel tho) mpv is using libav or not ? It says they use ffmpeg but ffmpeg is libav from what I understand
[09:55:26 CEST] <JEEB> it uses the libav* libraries whichever you built against :P if you built mpv against FFmpeg, then FFmpeg's libav* libraries
[09:56:38 CEST] <Radiator> Ooooh I see, in the end it only depends on the building process.
[09:56:55 CEST] <Radiator> Well, thank you JEEB
[09:57:36 CEST] <JEEB> also this stuff probably is more relevant towards the user channel, since development is development within the FFmpeg project, not utilizing it :)
[09:58:23 CEST] <Radiator> Oh my bad, I thought it was dev in general
[10:40:43 CEST] <cone-306> ffmpeg 03Sourabh Sharma 07master:3424600e0b19: doc/filters: correct range for contrast in eq filter
[11:46:54 CEST] <JEEB> VenomFK: can I politely request you to not privately message me unless absolutely required?
[11:47:13 CEST] <VenomFK> Sure, apologies
[14:07:55 CEST] <durandal_1707> fate have drastically reduced in number of machines
[14:11:46 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:28dfa1e37169: avfilter/vf_v360: rewrite storing of remap positions and interpolations
[14:11:47 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:32dd429bc735: avfilter/vf_v360: stop using floats in interpolation
[14:11:48 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:f0d8005ec5be: avfilter/vf_v360: add asserts to guard against invalid conditions
[14:11:49 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:058bbf48c6ba: avfilter/vf_v360: x86 SIMD for interpolations
[14:11:50 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:d13154144500: avfilter/vf_v360: reduce allocations by reusing duplicated remaps
[14:11:51 CEST] <cone-306> ffmpeg 03Paul B Mahol 07master:cb8c353032ab: avfilter/vf_v360: add option to pick rotation order
[17:48:02 CEST] <cone-306> ffmpeg 03James Almer 07master:2200cf1acaf0: x86/vf_v360: make remap{1,2}_8bit_line_avx2 work on x86_32
[17:48:03 CEST] <cone-306> ffmpeg 03James Almer 07master:4857688732e2: x86/vf_v360: use a faster horizontal add in remap4_8bit_line_avx2
[21:20:24 CEST] <jamrial> cehoyos: do you intend to push "lavf/rawenc: Only accept the appropriate stream type for raw muxers"?
[21:20:43 CEST] <cehoyos> I'll do it tonight, I was away for two days.
[21:21:10 CEST] <jamrial> ok
[21:21:12 CEST] <cehoyos> In the future, please feel free to push patches you believe should be applied. (I have done it in the past)
[21:21:31 CEST] <JEEB> oh, so sticking random stuff through rawvideo etc wasn't a meant feature? :)
[21:22:01 CEST] <jamrial> it should make it to 4.2.1, so ok, i'll push and backport it in a minute
[21:23:17 CEST] <JEEB> I recently used it to dump subtitle packets into separate files with the segment muxer :) but I guess that just means that it should be "rawdata" or so for unknown stream types instead of "rawvideo" or so :)
[21:23:50 CEST] <cone-928> ffmpeg 03Carl Eugen Hoyos 07master:aef24efb0c1e: lavf/rawenc: Only accept the appropriate stream type for raw muxers.
[21:26:00 CEST] <cehoyos> Sorry: Do you understand that I feel pissed now?
[21:26:08 CEST] <cone-928> ffmpeg 03Carl Eugen Hoyos 07release/4.2:1dec90d45656: lavf/rawenc: Only accept the appropriate stream type for raw muxers.
[21:27:28 CEST] <JEEB> I didn't mean to make you pissed :< sorry. because I think your change is correct. I was just misusing a tool that I noticed being there.
[21:27:28 CEST] <jamrial> cehoyos: i'm sorry?
[21:27:57 CEST] <cehoyos> You asked me if I will commit the patch, I answered that I will do it and started doing it and you committed the patch.
[21:28:16 CEST] <cehoyos> I wonder now if you did that on purpose only to make my life more difficult
[21:28:41 CEST] <jamrial> you said you'd do it tonight, then told me to feel free to push patches that should be applied
[21:28:50 CEST] <cehoyos> Yes, "in the future"
[21:29:07 CEST] <jamrial> since michaelni_ apparently plans to tag 4.2.1 before tonight, i went and pushed it so it may make it to the release in question
[21:29:08 CEST] <cehoyos> And I did wonder since you wrote that email why you didn't push it
[21:29:32 CEST] <jamrial> it was not my intention to make your life more dificult. i just wanted that fix committed on time for the release
[21:29:39 CEST] <cehoyos> ok, sorry
[21:30:45 CEST] <cehoyos> Note that it is of course no fix but a work-around that is often called a "hack"
[21:31:38 CEST] <cehoyos> JEEB: Did you read the commit message?
[21:32:24 CEST] <JEEB> only the part that was mentioned/logged on IRC. I'll go read the full one on gitweb if you think I misunderstood it
[21:32:35 CEST] <JEEB> ahhh
[21:32:36 CEST] <JEEB> ok
[21:32:40 CEST] <JEEB> roger that
[21:35:16 CEST] <durandal_1707> relax,  be happy
[21:36:02 CEST] <JEEB> well, I did note that I would consider the change to be correct even if rawvideo would only take in video AVStreams :)
[21:36:13 CEST] <JEEB> given that after all, it says "video"
[21:36:41 CEST] <JEEB> me pushing unknown or subtitle AVStreams into it would be the part I would consider being a "hack" given its name :)
[21:37:23 CEST] <JEEB> so no real unhappiness regarding that vOv
[21:37:58 CEST] <durandal_1707> you can still use -f u8
[21:38:36 CEST] <JEEB> I did find it cute that if you knew the right incantation with the segment muxer you'd get a csv with the timestamps and file names of segments, and then the packets dumpe
[21:38:39 CEST] <JEEB> *dumped
[21:39:55 CEST] <JEEB> lets you then parse the csv and validate that the timestamps of the entities within the samples (because someone decided that whole XML documents should be samples, with timing inside of course) are as you'd expect
[21:41:03 CEST] <JEEB> also I noticed someone wanted to add extra stream support into webvtt muxer just because instead of doing HLS with the HLS muxer he wanted to use the segment muxer. that was interesting
[21:41:27 CEST] <JEEB> although the webvtt support in the hls muxer is kind of weird since it seems like the master playlist support never thought of it
[21:41:40 CEST] <JEEB> might poke that if I have the time to add subtitle streams into the master playlist as well
[21:41:46 CEST] <jamrial> nevcairiel: your msvc fate clients haven't run in like two weeks
[21:52:16 CEST] <durandal_1707> it is end
[21:52:42 CEST] <durandal_1707> no more fate machines
[21:54:08 CEST] <kepstin> i could put fate on my amd k6-3 box, but i'm not sure that would be particularly useful ;)
[22:07:41 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:722fd4696583: avcodec/vorbisdec: fix FASTDIV usage for vr_type == 2
[22:07:42 CEST] <cone-928> ffmpeg 03Linjie Fu 07master:ef1342650f97: swscale/output: fix some code indentations
[22:07:43 CEST] <cone-928> ffmpeg 03Limin Wang 07master:ede15bf2e359: avfilter/vf_showinfo: display GOP timecode side data
[22:29:55 CEST] <cehoyos> jamrial: I cannot reproduce with clang either.
[22:30:53 CEST] <jamrial> cehoyos: i only have gcc and can't reproduce it
[22:31:05 CEST] <cehoyos> ... after testing with gcc
[22:31:08 CEST] <cehoyos> first
[22:31:15 CEST] <jamrial> so i does sound like a bug in whatever version of clang he's using
[22:31:34 CEST] <cehoyos> I already spent far too much time on an invalid ticket;-(
[22:37:02 CEST] <durandal_1707> cehoyos: do not be angry
[22:41:29 CEST] <iive> the sign at the end indicates sadness, not anger. afaik
[22:45:44 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:413e0f2516ee: avcodec/vc1_pred: Fix refdist in scaleforopp()
[22:45:45 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:55d4e22d71ca: avformat/vividas: check for tiny blocks using alignment
[22:45:46 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:12b909ba319d: avcodec/qdm2: Check frame size
[23:38:51 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:ee89d9e3d6ee: avcodec/apedec: Fix 32bit int overflow in do_apply_filter()
[23:38:52 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:6c583ec9bdcc: avcodec/iff: Check for overlap in cmap_read_palette()
[23:38:53 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:5b44aec196e6: avcodec/vorbisdec: fix FASTDIV usage for vr_type == 2
[23:38:54 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:d3b45f137882: avcodec/vc1_pred: Fix refdist in scaleforopp()
[23:38:55 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:5a1e0cae2f78: avformat/vividas: check for tiny blocks using alignment
[23:38:56 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:e66d4725c718: avcodec/qdm2: Check frame size
[23:38:57 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07release/4.2:1529dfb73a51: Update for 4.2.1
[23:55:06 CEST] <cone-928> ffmpeg 03Carl Eugen Hoyos 07release/4.1:60094fc2f552: lavf/rawenc: Only accept the appropriate stream type for raw muxers.
[00:00:00 CEST] --- Sat Sep  7 2019