[Ffmpeg-devel-irc] ffmpeg-devel.log.20180303

burek burek021 at gmail.com
Sun Mar 4 03:05:03 EET 2018


[00:00:45 CET] <BtbN> https://clips.twitch.tv/NicePopularPorcupineCurseLit in case you haven't heard it
[00:00:53 CET] <BtbN> that was done by the Microsoft AAC encoder
[00:01:12 CET] <nevcairiel> that sounds like a packet loss =p
[00:01:35 CET] <BtbN> a packet loss triggered by high pitched noises, accross thousands of streams?
[00:17:11 CET] <TD-Linux> (also twitch uses reliable transport)
[00:19:19 CET] <BtbN> "reliable"
[00:19:21 CET] <BtbN> They use rtmp
[00:19:40 CET] <nevcairiel> didnt they switch to html5 yet
[00:19:54 CET] <BtbN> not for ingest, no
[00:20:03 CET] <nevcairiel> oh on that end
[00:20:38 CET] <nevcairiel> rtmp isnt too bad though, especially if its single point to point
[00:20:59 CET] <BtbN> it has weird frame dropping logic
[00:24:32 CET] <wm4> I couldn't stream rtp reliably from ffmpeg to ffmpeg-using-player on a local machine when I tried today
[00:24:43 CET] <wm4> not sure which rtp variant that was
[00:24:48 CET] <wm4> because there are too many
[00:26:06 CET] <BtbN> rtp and rtmp are two totally different things
[00:26:13 CET] <BtbN> rtmp is basically flv over tcp
[00:29:02 CET] <wm4> no clue
[00:29:15 CET] <wm4> though isn't the flv thing only because of a libavformat hack that remuxes on the fly?
[00:30:17 CET] <BtbN> rtmp originated from flme back in the day
[00:30:24 CET] <BtbN> so I can see it being similar to flv
[00:30:32 CET] <BtbN> *fmle
[00:31:08 CET] <classict> hi, I want to know if possible create an ffplay with vaapi support
[00:31:09 CET] <classict> ??
[00:31:23 CET] <JEEB> you will want an actual player
[00:31:28 CET] <JEEB> like mpv, for example
[00:31:58 CET] <classict> yes, but....only ffplay get lowest latency, and smooth play
[00:32:13 CET] <BtbN> I highly doubt that
[00:32:29 CET] <classict> really?
[00:32:50 CET] <JEEB> very similar latency should be then possible with a lot of stuff using FFmpeg's libraries
[00:32:53 CET] <JEEB> :P
[00:33:29 CET] <classict> ok, i try a lot of option's but some flags isn't support by mpv
[00:34:10 CET] <classict>  mpv --no-cache -no-demuxer-thread   --vd-lavc-threads=1  --demuxer-lavf-o=fflags=+nobuffer  --demuxer-lavf-buffersize=1 --demuxer-lavf-allow-mimetype=no  --no-audio    --hwdec=vaapi --vo=vaapi rtmp input
[00:34:37 CET] <classict> vs ffplay  -fflags nobuffer -flags  low_delay -framedrop rtmp-input
[00:35:09 CET] <wm4> does that -flags low_delay actually do anything
[00:35:30 CET] <classict> well, framedrop maybe is the most important
[00:35:47 CET] <classict> really, with this command I get 500ms latency
[00:36:12 CET] <classict> is enough good for my neededs
[00:36:28 CET] <wm4> then use it
[00:36:55 CET] <classict> yes, but I need vaapi support, because this get near 30% of cpu
[00:37:10 CET] <classict> with vaapi, in other player get 5% or less
[00:37:55 CET] <classict> something like this https://github.com/gbeauchesne/ffvademo
[00:38:28 CET] <classict> but this branch doesn't support the flags what I need
[00:39:24 CET] <cone-146> ffmpeg 03Philip Langdale 07master:e990713ff9e3: avcodec/hevcdec: Declare that nvdec supports 12bit decoding
[00:39:25 CET] <cone-146> ffmpeg 03Philip Langdale 07master:9d5aff09a716: swscale: Add p016 output support and generalise yuv420p1x to p010
[00:39:26 CET] <cone-146> ffmpeg 03Philip Langdale 07master:6a89cdc474bc: avcodec/nvenc: Declare support for P016
[02:11:24 CET] <gagandeep> kierank/atomnuker: are there more samples for cineform encoded files on sample.ffmpeg
[05:45:54 CET] <gagandeep> guys many functions present in libavutil are made so that one doesn't need to have access to c standard library files?
[05:48:28 CET] <gagandeep> nevermind got confused about some stuff! :/
[12:53:32 CET] <atomnuker> JEEB: had a look at my meson thing?
[12:53:44 CET] <atomnuker> curious to know whether it works on windows
[13:05:45 CET] <durandal_1707> where is kierank hiding?
[13:16:49 CET] <jdarnley> He's probably asleep since I saw a message from him at 5AM
[13:17:50 CET] <jdarnley> Or maybe "THE BEAST FROM THE EAST" finally annihilated London.
[13:17:58 CET] Action: jdarnley hates himself
[13:25:34 CET] <kierank> hello
[14:02:06 CET] <atomnuker> oh wow, kate allows you to print shortcuts: https://0x0.st/smXS.pdf
[14:02:23 CET] <atomnuker> was wondering why there was a print button on the shortcuts page
[14:18:58 CET] <cone-511> ffmpeg 03Vishwanath Dixit 07master:0dd1fff1c97d: avformat/dashenc: fix for segment open issue when persistent connection is enabled
[14:18:58 CET] <cone-511> ffmpeg 03Serhii Marchuk 07master:e45d55756188: avformat/dashenc: Use HTTP DELETE for HTTP base proto
[17:50:02 CET] <BtbN> I feel like consolidating that ffnvcodec check into just cuda wasn't the best idea. Now configure just throws "nvenc requested, but dependency cuda not found" and people go and install the CUDA sdk
[18:44:36 CET] <atomnuker> why does avcodec_descriptor_get() fail to find atrac9?
[18:44:52 CET] <atomnuker> I have an entry in codec_desc.c
[18:45:03 CET] <atomnuker> I think this might be a regression, I just rebased
[18:51:18 CET] <nevcairiel> iirc the descriptor table now needs to be strictly sorted by codec id
[18:51:26 CET] <nevcairiel> for some reason s omeone thought optimizing that made sense
[19:08:05 CET] <atomnuker> do we have a test for that?
[19:09:25 CET] <atomnuker> welp, it would be fine if the person who optimized it sorted codec_desc (he did, right?)
[19:13:31 CET] <durandal_1707> atomnuker: they should be sorted, arent they?
[19:15:58 CET] <atomnuker> yep, they seem to be
[20:19:37 CET] <nevcairiel> Sounded like you added a new codec descriptor, so that may have been out of place from the rebase
[20:19:55 CET] <BtbN> Picking an ISP has never been more fun. I can choose between DOCSIS with the most horrible Backbone/Peerings one can think of, DSL from an ISP who takes their customers hostage to get money from server hosters, or DSL from ISPs who resell said other ISP.
[20:20:38 CET] <nevcairiel> You see, thats why I choose where I want to live on the availability of decent internet services
[20:20:48 CET] <BtbN> I'm not gonna move out of Germany though
[20:21:12 CET] <JEEB> I'd be surprised if Germany didn't have proper internets
[20:21:18 CET] <BtbN> be surprised then
[20:21:21 CET] <nevcairiel> Mine works fine
[20:21:27 CET] <JEEB> in my case I pay 49.99 eur/m for 100|10 fiber right now
[20:21:41 CET] <sfan5> "what is fiber?" ~ISPs in Germany
[20:21:42 CET] <JEEB> which is a bit more expensive than the VDSL2+ deal I had previously
[20:21:55 CET] <BtbN> You either go to Vodafone with their Cable and DOCSIS, and in the evenings maybe still get 10 Mbps if you're lucky.
[20:22:06 CET] <nevcairiel> I have 100/40 and zero issues
[20:22:09 CET] <BtbN> Or Telekom, and you pay extra for everything
[20:22:33 CET] <BtbN> https://wiki.hetzner.de/index.php/Double_Paid_Traffic/en that's Telekom for you
[20:22:34 CET] <nevcairiel> what more beyond basic internet services do you need? =p
[20:22:43 CET] <sfan5> BtbN: no unitymedia at your place?
[20:22:48 CET] <sfan5> not that it's any better than vodafone cable
[20:22:57 CET] <BtbN> You only ever have one Cable provider
[20:23:33 CET] <sfan5> oh
[20:24:22 CET] <BtbN> I'll probably just go to some reseller with monthly cancelation and test it out
[20:24:37 CET] <BtbN> it's also dirt cheap, 16¬/month for 100/40
[20:27:54 CET] <sfan5> surely that price is only for the 12 or 24 months
[20:27:57 CET] <sfan5> +first
[20:28:36 CET] <iive> BtbN, are you in east europe?
[20:28:42 CET] <BtbN> Germany
[20:29:04 CET] <BtbN> sfan5, it's for the first 12 months. But you can cancel every month. So once it gets expensive, you just cancel and go somewhere else, and get 12 cheap months.
[20:29:23 CET] <sfan5> sure that works, it's a little bothersome though
[20:29:30 CET] <iive> BtbN, east germany ;) It's completely reasonable price.
[20:30:07 CET] <BtbN> the normal price for 100/40 is 35~45¬ depending on where you order it
[20:32:05 CET] <nevcairiel> seems like a dumb company if they dont even bind you for a year and then offer something like that
[20:32:49 CET] <BtbN> a lot of people prefer the monthly cancelation over slightly cheaper prices
[20:33:09 CET] <iive> why? if they have the best price and service, why would somebody move?
[20:33:27 CET] <BtbN> Because some other ISP came up during that time, with better service
[20:33:45 CET] <nevcairiel> like that happens
[20:33:58 CET] <sfan5> iive: >implying Germany ISPs have good service
[20:34:01 CET] <nevcairiel> maybe if you care about more then  pure internet
[20:35:27 CET] <nevcairiel> but switching ISP is disruptive, its not like power delivery where you can just switch every year for the bonus and obviously your power doesnt go out
[20:40:06 CET] <BtbN> If you don't care about keeping your phone number it's pretty smooth
[20:52:10 CET] <BtbN> Can I freely rename configure options? I want to rename cuda to ffnvcodec, to avoid confusion
[20:55:57 CET] <atomnuker> yeah, you can
[20:56:25 CET] <atomnuker> fuck ISPs who bundle TV, phone and internet with the same package
[20:56:34 CET] <atomnuker> or who include anything but internet
[20:57:13 CET] <atomnuker> in what century do these people live in? a stationary phone? interlaced moving pictures?
[21:05:45 CET] <philipl> BtbN: as in --enable-cuda?
[21:06:09 CET] <BtbN> yes. Cause right now, if you --enable-nvenc and don't have the headers, it complains about missing cuda
[21:06:20 CET] <philipl> makes sense.
[21:20:08 CET] <atomnuker> why does aurelien even have push access? he shouldn't
[21:22:01 CET] <atomnuker> sent a patch to prevent being unable to support the suffix "m" for minutes after we do a release
[21:24:14 CET] <atomnuker> if he objects I'll just revert his shit, it was pushed too soon anyway
[21:24:20 CET] <sfan5> how long do I need to wait until someone decides to push my 1-line libtls patch btw?
[21:24:39 CET] <atomnuker> ping it
[21:24:49 CET] <JEEB> I think he did already some time ago
[21:25:06 CET] <durandal_1707> ping it again, and again, and again...
[21:25:08 CET] <JEEB> it's just like nobody answering me what they meant with their comment at the patch. maybe I'm just stupid
[21:25:09 CET] <sfan5> I did so two days ago
[21:26:11 CET] <durandal_1707> first time ping it after 7 days, next time after 6 days, after that 5 days, and so on until you start pinging it each hour :)
[21:26:39 CET] <sfan5> :D
[21:27:27 CET] <JEEB> I actually thought about just pushing it
[21:27:31 CET] <JEEB> since nobody objected to it
[21:30:41 CET] <JEEB> there, sent a LGTM at it
[21:32:30 CET] <JEEB> I'd like to know what was meant with this comment to which I asked clarification https://patchwork.ffmpeg.org/patch/7554/
[21:33:03 CET] <JEEB> and something about https://patchwork.ffmpeg.org/patch/7555/
[21:50:17 CET] <BtbN> atomnuker, the hell even was this: https://github.com/jp9000/obs-studio/commit/a89470d2ea1bdd0c6a02e1af6f57f207fbd97f20
[21:50:53 CET] <atomnuker> I'm guessing they forced the low pass filter on to make it sound better
[22:06:53 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:a079eaba8ee2: avcodec/mediacodecdec_common: refactor mediacodec_dec_parse_format()
[22:06:53 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:5d69e249c81e: avcodec/mediacodecdec_common: remove spurious space
[22:06:53 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:cc9875dc2938: avcodec/mediacodecdec_common: remove unused field from MediaCodecDecContext
[22:06:53 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:c55ba52a6a06: avcodec/mediacodec_wrapper: load and use MediaFormat.constainsKey()
[22:06:53 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:535e02022548: avcodec/mediacodecdec: add missing "libavutil/internal.h" include
[22:06:54 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:af167d970bbf: avcodec/mediacodecdec: factorize common extradata functions
[22:06:54 CET] <cone-280> ffmpeg 03Matthieu Bouron 07master:2238e54ef02b: avcodec/mediacodecdec: factorize codec declarations
[22:10:48 CET] <nevcairiel> BtbN: using a custom cutoff was recommend for a long time, however the logic in there is totally weird, like cutoff2 is never going to be used for anything
[22:12:07 CET] <BtbN> interesting, also, yeah, by logic cutoff2 will never be used
[22:12:24 CET] <BtbN> hm, I broke disabling cuda/ffnvcodec somehow
[22:12:36 CET] <nevcairiel> cutoff3 at nyqist is j ust a fallback, so the only real one is cutoff1
[22:12:48 CET] <BtbN> --disable-cuda/ffnvcodec is just straight up ignored if it's found
[22:13:31 CET] <BtbN> I guess the pkgconfig check just re-enables it blindly
[22:24:28 CET] <jamrial> BtbN: do enabled cuda && check_pkg_config ...
[22:24:31 CET] <jamrial> or something like that
[22:24:47 CET] <BtbN> more like disabled cuda ||
[22:24:49 CET] <JEEB> that reminds me of trying to add pkg-config support for zlib
[22:24:54 CET] <BtbN> since it's auto-detect after all
[22:25:15 CET] <JEEB> and I wasn't sure how2 add that one (I mean, the pkg-config part worked, but I wasn't sure if it was the right way)
[22:25:32 CET] <BtbN> It's ffmpeg configure. You're never sure about it being the right way.
[22:26:00 CET] <JEEB> I guess I will just post it on the ML for comments. since I'm getting tired of adding extra cflags just to have the automated check find it
[22:26:02 CET] <BtbN> jamrial, I'm in the process of renaming cuda to ffnvcodec though
[22:26:39 CET] <jamrial> you need to keep the cuda alias in any case, at least for a bunch of releases
[22:26:51 CET] <BtbN> Yes, cuda stays, and depends on ffnvcodec itself
[22:27:01 CET] <BtbN> There is also a lot of code that depends on CONFIG_CUDA and such
[22:27:09 CET] <jamrial> also, other autodected stuff like zlib does enabled && check()
[22:27:30 CET] <BtbN> It's just so configure doesn't present users with "missing dependency: cuda"
[22:27:39 CET] <BtbN> which is highly misleading
[22:28:23 CET] <BtbN> jamrial, the enabled zlib thing adds cppflags for it
[22:28:25 CET] <BtbN> not checks if it exists
[22:28:47 CET] <jamrial> look for the other zlib check
[22:28:54 CET] <jamrial> next to the bzlib one
[22:29:27 CET] <BtbN> Oh, those. That seems wrong to me. It will never be enabled?
[22:29:31 CET] <BtbN> Just not disabled
[22:46:15 CET] <jamrial> BtbN: careful with the dependency stuff. one more level might screw up things
[22:46:23 CET] <jamrial> even after your fix from some months ago
[22:46:50 CET] <BtbN> It works nice from what I tested
[22:47:20 CET] <jamrial> guess it doesn't set extralibs of any kind so it's safe
[22:48:22 CET] <BtbN> also, there is no other level
[22:48:28 CET] <BtbN> cuda is replaced with ffnvcodec
[22:48:35 CET] <BtbN> and cuda is just depending on ffnvcodec itself
[22:48:53 CET] <BtbN> it's not nvenc->cuda->ffnvcodec
[23:02:19 CET] <cone-280> ffmpeg 03Tianqiang Liu 07master:5df9d75b4417: avformat/mov: Add manu/modl to mov_read_udta_string.
[23:02:20 CET] <cone-280> ffmpeg 03Jun Zhao 07master:1c7f1f38c533: doc/bitstream_filters: correct dump_extra bsfs docs.
[00:00:00 CET] --- Sun Mar  4 2018


More information about the Ffmpeg-devel-irc mailing list